qapi: add generator for Golang interface
v2 cover qapi: add generator for Golang interface
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html
# What this is about?
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmp
# Issues raised in RFC v1
The leading '*' for issues I addressed in this iteration
* 1) Documentation was removed to avoid License issues, by Daniel
Thread: https://lists.nongnu.org/archive/html/qemu-devel/2022-05/msg01889.html
It is important for the generated Go module to be compatible with Licenses used by projects that would be using this. Copying the documentation of the QAPI spec might conflict with GPLv2+.
I have not proposed another license in this iteration, but I'm planning to go with MIT No Attribution, aka MIT-0 [0]. Does it make sense to bind the generated code's license to MIT-0 already at generator level?
[0] https://github.com/aws/mit-0/blob/master/MIT-0
2) Inconsistent generated Names, by Andrea + Markus
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg05026.html
Example 1:
| qapi | Go | Expected |
| @logappend | Logappend | LogAppend |
Example 2:
(acronyms) VncInfo and DisplayReloadOptionsVNC
This was not addressed in RFC v2 mainly because it seems to need more metadata support from the QAPI spec to handle specific scenarios. The solution seems either an extra metadata proposal by Andrea [1] or reviving Kevin's work [2]
[1] https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg00127.html
[2] https://lists.gnu.org/archive/html/qemu-devel/2021-09/msg04703.html
* 3) Better type safety, by Andrea + Daniel
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01906.html
Most of the 'Any' type (interface {}) has been removed. The only place it still exists is for fields that uses QAPI's any type, like with command qom-set or the struct type ObjectPropertyInfo.
* 4) QAPI enums mapped to String instead of Int type, by Daniel.
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01904.html
I'm still over the fence about using string here, mostly by the same issue reported here:
https://gitlab.com/libvirt/libvirt-go-module/-/merge_requests/30#note_975517740
* 5) Events and Commands as interface, by Daniel
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01914.html
So, instead of having a Command/Event struct with a Any type for the Arguments (which could be set with SetPasswordCommand struct type for example); now we have a Command interface which all previous structs that behaved as Arguments implement.
I've included Marshal{Command Event} and Unmarshal{Command Event} helper functions that operate on top of each interface.
* 6) Removing Any from Unions, by Daniel
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01912.html
I basically followed the above suggestion to all other types that used Any. Specifically to unions were the removal of the 'discriminator' field, as proposed also in the above link.
* 7) Flat structs by removing embed types. Discussion with Andrea
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01590.html
No one required it but I decided to give it a try. Major issue that I see with this approach is to have generated a few 'Base' structs that are now useless. Overall, less nested structs seems better to me. Opnions?
Example:
| /* This is now useless, should be removed? */ | type InetSocketAddressBase struct {
| Host string json:"host"
| Port string json:"port"
| }
|
| type InetSocketAddress struct {
| // Base fields
| Host string json:"host"
| Port string json:"port"
|
|
| Numeric *bool json:"numeric,omitempty"
| To *uint16 json:"to,omitempty"
| Ipv4 *bool json:"ipv4,omitempty"
| Ipv6 *bool json:"ipv6,omitempty"
| KeepAlive *bool json:"keep-alive,omitempty"
| Mptcp *bool json:"mptcp,omitempty"
| }
8) Supporting multiple versions
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg02147.html
I'm keen to working on the proposed solution above as it seems a good compromise to make code that can be compatible with multiple versions of qmp/qemu.
But the basis needs to be defined first, so this is for the future.
* 9) Handling { "error": { ... } }
This was missing in the RFC v1. I've added a QapiError type that is included in all CommandReturn types, following Andrea's suggestion:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg02199.html
# victortoso/qapi-go and victortoso/qemu
I'm currently hosting the generated code from this RFCv2 at:
https://gitlab.com/victortoso/qapi-go/-/tree/main/pkg/qapi
This series can also be found at:
https://gitlab.com/victortoso/qemu/-/commits/qapi-golang
Thanks for taking a look,
Victor
Victor Toso (8):
qapi: golang: Generate qapi's enum types in Go
qapi: golang: Generate qapi's alternate types in Go
qapi: golang: Generate qapi's struct types in Go
qapi: golang: Generate qapi's union types in Go
qapi: golang: Generate qapi's event types in Go
qapi: golang: Generate qapi's command types in Go
qapi: golang: Add CommandResult type to Go
qapi: golang: document skip function visit_array_types
scripts/qapi/golang.py | 765 +++++++++++++++++++++++++++++++++++++++++ scripts/qapi/main.py | 2 + 2 files changed, 767 insertions(+) create mode 100644 scripts/qapi/golang.py
Signature · 1 line
2.36.1
RFC PATCH v2 1/8 qapi: golang: Generate qapi's enum types in Go
This patch handles QAPI enum types and generates its equivalent in Go.
Basically, Enums are being handled as strings in Golang.
1. For each QAPI enum, we will define a string type in Go to be the assigned type of this specific enum.
2. Naming: CamelCase will be used in any identifier that we want to export [0], which is everything.
[0] https://go.dev/ref/spec#Exported_identifiers
Example:
qapi:
| { 'enum': 'DisplayProtocol',
| 'data': [ 'vnc', 'spice' ] }
go:
| type DisplayProtocol string | | const ( | DisplayProtocolVnc DisplayProtocol = "vnc" | DisplayProtocolSpice DisplayProtocol = "spice" | )
--- scripts/qapi/golang.py | 134 +++++++++++++++++++++++++++++++++++++++++ scripts/qapi/main.py | 2 + 2 files changed, 136 insertions(+) create mode 100644 scripts/qapi/golang.py
scripts/qapi/golang.py new file +134 -0
@@ -0,0 +1,134 @@ +""" +Golang QAPI generator +""" +# Copyright (c) 2022 Red Hat Inc. +# +# Authors: +# Victor Toso <[email protected]> +# +# This work is licensed under the terms of the GNU GPL, version 2. +# See the COPYING file in the top-level directory. + +# Just for type hint on self +from __future__ import annotations + +import os +from typing import List, Optional + +from .schema import ( + QAPISchema, + QAPISchemaType, + QAPISchemaVisitor, + QAPISchemaEnumMember, + QAPISchemaFeature, + QAPISchemaIfCond, + QAPISchemaObjectType, + QAPISchemaObjectTypeMember, + QAPISchemaVariants, +) +from .source import QAPISourceInfo + + +class QAPISchemaGenGolangVisitor(QAPISchemaVisitor): + + def __init__(self, prefix: str): + super().__init__() + self.target = {name: "" for name in ["enum"]} + self.schema = None + self.golang_package_name = "qapi" + + def visit_begin(self, schema): + self.schema = schema + + # Every Go file needs to reference its package name + for target in self.target: + self.target[target] = f"package {self.golang_package_name}\n" + + def visit_end(self): + self.schema = None + + def visit_object_type(self: QAPISchemaGenGolangVisitor, + name: str, + info: Optional[QAPISourceInfo], + ifcond: QAPISchemaIfCond, + features: List[QAPISchemaFeature], + base: Optional[QAPISchemaObjectType], + members: List[QAPISchemaObjectTypeMember], + variants: Optional[QAPISchemaVariants] + ) -> None: + pass + + def visit_alternate_type(self: QAPISchemaGenGolangVisitor, + name: str, + info: Optional[QAPISourceInfo], + ifcond: QAPISchemaIfCond, + features: List[QAPISchemaFeature], + variants: QAPISchemaVariants + ) -> None: + pass + + def visit_enum_type(self: QAPISchemaGenGolangVisitor, + name: str, + info: Optional[QAPISourceInfo], + ifcond: QAPISchemaIfCond, + features: List[QAPISchemaFeature], + members: List[QAPISchemaEnumMember], + prefix: Optional[str] + ) -> None: + + value = qapi_to_field_name_enum(members[0].name) + fields = "" + for member in members: + value = qapi_to_field_name_enum(member.name) + fields += f'''\t{name}{value} {name} = "{member.name}"\n''' + + self.target["enum"] += f''' +type {name} string +const ( +{fields[:-1]} +) +''' + + def visit_array_type(self, name, info, ifcond, element_type): + pass + + def visit_command(self, + name: str, + info: Optional[QAPISourceInfo], + ifcond: QAPISchemaIfCond, + features: List[QAPISchemaFeature], + arg_type: Optional[QAPISchemaObjectType], + ret_type: Optional[QAPISchemaType], + gen: bool, + success_response: bool, + boxed: bool, + allow_oob: bool, + allow_preconfig: bool, + coroutine: bool) -> None: + pass + + def visit_event(self, name, info, ifcond, features, arg_type, boxed): + pass + + def write(self, output_dir: str) -> None: + for module_name, content in self.target.items(): + go_module = module_name + "s.go" + go_dir = "go" + pathname = os.path.join(output_dir, go_dir, go_module) + odir = os.path.dirname(pathname) + os.makedirs(odir, exist_ok=True) + + with open(pathname, "w") as outfile: + outfile.write(content) + + +def gen_golang(schema: QAPISchema, + output_dir: str, + prefix: str) -> None: + vis = QAPISchemaGenGolangVisitor(prefix) + schema.visit(vis) + vis.write(output_dir) + + +def qapi_to_field_name_enum(name: str) -> str: + return name.title().replace("-", "")
scripts/qapi/main.py +2 -0
@@ -15,6 +15,7 @@ from .common import must_match from .error import QAPIError from .events import gen_events +from .golang import gen_golang from .introspect import gen_introspect from .schema import QAPISchema from .types import gen_types
@@ -54,6 +55,7 @@ def generate(schema_file: str, gen_events(schema, output_dir, prefix) gen_introspect(schema, output_dir, prefix, unmask) + gen_golang(schema, output_dir, prefix) def main() -> int: """
Signature · 1 line
2.36.1
RFC PATCH v2 2/8 qapi: golang: Generate qapi's alternate types in Go
In reply to Victor TosoThis patch handles QAPI alternate types and generates data structures in Go that handles it.
At this moment, there are 5 alternates in qemu/qapi, they are:
* BlockDirtyBitmapMergeSource
* Qcow2OverlapChecks
* BlockdevRef
* BlockdevRefOrNull
* StrOrNull
Alternate types are similar to Union but without a discriminator that can be used to identify the underlying value on the wire. It is needed to infer it. In Go, all the types are mapped as optional fields and Marshal and Unmarshal methods will be handling the data checks.
Example:
qapi:
| { 'alternate': 'BlockdevRef',
| 'data': { 'definition': 'BlockdevOptions',
| 'reference': 'str' } }go:
| type BlockdevRef struct {
| Definition *BlockdevOptions
| Reference *string
| }usage:
| input := `{"driver":"qcow2","data-file":"/some/place/my-image"}`
| k := BlockdevRef{}
| err := json.Unmarshal([]byte(input), &k)
| if err != nil {
| panic(err)
| }
| // *k.Definition.Qcow2.DataFile.Reference == "/some/place/my-image"--- scripts/qapi/golang.py | 119 ++++++++++++++++++++++++++++++++++++++++- 1 file changed, 117 insertions(+), 2 deletions(-)
scripts/qapi/golang.py +117 -2
@@ -29,11 +29,32 @@ from .source import QAPISourceInfo +TEMPLATE_HELPER = ''' +// Alias for go version lower than 1.18 +type Any = interface{} + +// Creates a decoder that errors on unknown Fields +// Returns true if successfully decoded @from string @into type +// Returns false without error is failed with "unknown field" +// Returns false with error is a different error was found +func StrictDecode(into interface{}, from []byte) error { + dec := json.NewDecoder(strings.NewReader(string(from))) + dec.DisallowUnknownFields() + + if err := dec.Decode(into); err != nil { + return err + } + return nil +} +''' + + class QAPISchemaGenGolangVisitor(QAPISchemaVisitor): def __init__(self, prefix: str): super().__init__() - self.target = {name: "" for name in ["enum"]} + self.target = {name: "" for name in ["alternate", "enum", "helper"]} + self.objects_seen = {} self.schema = None self.golang_package_name = "qapi"
@@ -44,6 +65,8 @@ def visit_begin(self, schema): for target in self.target: self.target[target] = f"package {self.golang_package_name}\n" + self.target["helper"] += TEMPLATE_HELPER + def visit_end(self): self.schema = None
@@ -65,7 +88,69 @@ def visit_alternate_type(self: QAPISchemaGenGolangVisitor, features: List[QAPISchemaFeature], variants: QAPISchemaVariants ) -> None: - pass + assert name not in self.objects_seen + self.objects_seen[name] = True + + marshal_return_default = f'nil, errors.New("{name} has empty fields")' + marshal_check_fields = "" + unmarshal_check_fields = "" + variant_fields = "" + + # We need to check if the Alternate type supports NULL as that + # means that JSON to Go would allow all fields to be empty. + # Alternate that don't support NULL, would fail to convert + # to JSON if all fields were empty. + return_on_null = f"errors.New(`null not supported for {name}`)" + + # Assembly the fields and all the checks for Marshal and + # Unmarshal methods + for var in variants.variants: + # Nothing to generate on null types. We update some + # variables to handle json-null on marshalling methods. + if var.type.name == "null": + marshal_return_default = '[]byte("null"), nil' + return_on_null = "nil" + continue + + var_name = qapi_to_field_name(var.name) + var_type = qapi_schema_type_to_go_type(var.type.name) + variant_fields += f"\t{var_name} *{var_type}\n" + + if len(marshal_check_fields) > 0: + marshal_check_fields += "} else " + + marshal_check_fields += f'''if s.{var_name} != nil {{ + return json.Marshal(s.{var_name}) + ''' + + unmarshal_check_fields += f'''// Check for {var_type} + {{ + s.{var_name} = new({var_type}) + if err := StrictDecode(s.{var_name}, data); err == nil {{ + return nil + }} + s.{var_name} = nil + }} +''' + + marshal_check_fields += "}" + + self.target["alternate"] += generate_struct_type(name, variant_fields) + self.target["alternate"] += f''' +func (s {name}) MarshalJSON() ([]byte, error) {{ + {marshal_check_fields} + return {marshal_return_default} +}} + +func (s *{name}) UnmarshalJSON(data []byte) error {{ + // Check for json-null first + if string(data) == "null" {{ + return {return_on_null} + }} + {unmarshal_check_fields} + return errors.New(fmt.Sprintf("Can't convert to {name}: %s", string(data))) +}} +''' def visit_enum_type(self: QAPISchemaGenGolangVisitor, name: str,
@@ -130,5 +215,35 @@ def gen_golang(schema: QAPISchema, vis.write(output_dir) +# Helper function for boxed or self contained structures. +def generate_struct_type(type_name, args="") -> str: + args = args if len(args) == 0 else f"\n{args}\n" + return f''' +type {type_name} struct {{{args}}} +''' + + +def qapi_schema_type_to_go_type(type: str) -> str: + schema_types_to_go = { + 'str': 'string', 'null': 'nil', 'bool': 'bool', 'number': + 'float64', 'size': 'uint64', 'int': 'int64', 'int8': 'int8', + 'int16': 'int16', 'int32': 'int32', 'int64': 'int64', 'uint8': + 'uint8', 'uint16': 'uint16', 'uint32': 'uint32', 'uint64': + 'uint64', 'any': 'Any', 'QType': 'QType', + } + + prefix = "" + if type.endswith("List"): + prefix = "[]" + type = type[:-4] + + type = schema_types_to_go.get(type, type) + return prefix + type + + def qapi_to_field_name_enum(name: str) -> str: return name.title().replace("-", "") + + +def qapi_to_field_name(name: str) -> str: + return name.title().replace("_", "").replace("-", "")
Signature · 1 line
2.36.1
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
Sorry it took me a while to find the time to look into this!
Overall this second iteration is a significant improvement over the initial one. There are still a few things that I think should be changed, so for the time being I'm going to comment mostly on the generated Go code and leave the details of the implementation for later.
Quoted 9 lines · Victor Toso
This patch handles QAPI alternate types and generates data structures in Go that handles it.
At this moment, there are 5 alternates in qemu/qapi, they are:
* BlockDirtyBitmapMergeSource
* Qcow2OverlapChecks
* BlockdevRef
* BlockdevRefOrNull
* StrOrNull
I personally don't think it's very useful to list all the alternate types in the commit message, or even mention how many there are. You do this for all other types too, and it seems to me that it's just an opportunity for incosistent information to make their way to the git repository - what if a new type is introduced between the time your series is queued and merged? I'd say just drop this part.
Quoted 13 lines
Example:
qapi:
| { 'alternate': 'BlockdevRef', | 'data': { 'definition': 'BlockdevOptions', | 'reference': 'str' } }go:
| type BlockdevRef struct { | Definition *BlockdevOptions | Reference *string | }
Quoted 8 lines
usage:
| input := `{"driver":"qcow2","data-file":"/some/place/my-image"}` | k := BlockdevRef{} | err := json.Unmarshal([]byte(input), &k) | if err != nil { | panic(err) | } | // *k.Definition.Qcow2.DataFile.Reference == "/some/place/my-image"
Let me just say that I absolutely *love* how you've added these bits comparing the QAPI / Go representations as well as usage. They really help a lot understanding what the generator is trying to achieve :)
scripts/qapi/golang.py line 48
Quoted 13 lines
scripts/qapi/golang.py +13
-0+// Creates a decoder that errors on unknown Fields +// Returns true if successfully decoded @from string @into type +// Returns false without error is failed with "unknown field" +// Returns false with error is a different error was found +func StrictDecode(into interface{}, from []byte) error { + dec := json.NewDecoder(strings.NewReader(string(from))) + dec.DisallowUnknownFields() + + if err := dec.Decode(into); err != nil { + return err + } + return nil +}
The documentation doesn't seem to be consistent with how the function actually works: AFAICT it returns false *with an error* for any failure, including those caused by unknown fields being present in the input JSON.
Looking at the generated code:
Quoted 5 lines
type BlockdevRef struct {
Definition *BlockdevOptions
Reference *string
}
Quoted 7 lines
func (s BlockdevRef) MarshalJSON() ([]byte, error) {
if s.Definition != nil {
return json.Marshal(s.Definition)
} else if s.Reference != nil {
return json.Marshal(s.Reference)
}
return nil, errors.New("BlockdevRef has empty fields")Returning an error if no field is set is good. Can we be more strict and returning one if more than one is set as well? That feels better than just picking the first one.
Quoted 11 lines
func (s *BlockdevRef) UnmarshalJSON(data []byte) error {
// Check for json-null first
if string(data) == "null" {
return errors.New(`null not supported for BlockdevRef`)
}
// Check for BlockdevOptions
{
s.Definition = new(BlockdevOptions)
if err := StrictDecode(s.Definition, data); err == nil {
return nil
}The use of StrictDecode() here means that we won't be able to parse an alternate produced by a version of QEMU where BlockdevOptions has gained additional fields, doesn't it?
Considering that we will happily parse such a BlockdevOptions outside of the context of BlockdevRef, I think we should be consistent and allow the same to happen here.
Quoted 4 lines
s.Definition = nil
}
// Check for string
{Quoted 8 lines
s.Reference = new(string)
if err := StrictDecode(s.Reference, data); err == nil {
return nil
}
s.Reference = nil
}
return errors.New(fmt.Sprintf("Can't convert to BlockdevRef: %s", string(data)))On a similar note to the MarshalJSON comment above, I'm not sure this is right.
If we find that more than one field of the alternate is set, we should error out - that's just invalid JSON we're dealing with. But if we couldn't find any, that might be valid JSON that's been produced by a version of QEMU that introduced additional options to the alternate. In the spirit of "be liberal in what you accept", I think we should probably not reject that? Of course then the client code will have to look like
if r.Definition != nil {
// do something with r.Definition
} else if r.Reference != nil {
// do something with r.Reference
} else {
// we don't recognize this - error out
}but the same is going to be true for enums, events and everything else that can be extended in a backwards-compatible fashion.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
Hi,
Quoted 1 line · Andrea Bolognani
Sorry it took me a while to find the time to look into this!
(1.5 month later.. what can I say!) :)
Quoted 5 lines
Overall this second iteration is a significant improvement over the initial one. There are still a few things that I think should be changed, so for the time being I'm going to comment mostly on the generated Go code and leave the details of the implementation for later.
Sure, and thanks.
Quoted 11 lines · Victor Toso
Quoted 9 lines
This patch handles QAPI alternate types and generates data structures in Go that handles it.
At this moment, there are 5 alternates in qemu/qapi, they are:
* BlockDirtyBitmapMergeSource
* Qcow2OverlapChecks
* BlockdevRef
* BlockdevRefOrNull
* StrOrNull
Quoted 6 lines
I personally don't think it's very useful to list all the alternate types in the commit message, or even mention how many there are. You do this for all other types too, and it seems to me that it's just an opportunity for incosistent information to make their way to the git repository - what if a new type is introduced between the time your series is queued and merged? I'd say just drop this part.
No issue on my side in dropping this bits.
Quoted 22 lines
Quoted 13 lines
Example:
qapi:
| { 'alternate': 'BlockdevRef', | 'data': { 'definition': 'BlockdevOptions', | 'reference': 'str' } }go:
| type BlockdevRef struct { | Definition *BlockdevOptions | Reference *string | }Quoted 8 lines
usage:
| input := `{"driver":"qcow2","data-file":"/some/place/my-image"}` | k := BlockdevRef{} | err := json.Unmarshal([]byte(input), &k) | if err != nil { | panic(err) | } | // *k.Definition.Qcow2.DataFile.Reference == "/some/place/my-image"
Quoted 3 lines
Let me just say that I absolutely *love* how you've added these bits comparing the QAPI / Go representations as well as usage. They really help a lot understanding what the generator is trying to achieve :)
Thanks!
Quoted 10 lines
Quoted 10 lines
scripts/qapi/golang.py +10
-0+// Creates a decoder that errors on unknown Fields +// Returns true if successfully decoded @from string @into type +// Returns false without error is failed with "unknown field" +// Returns false with error is a different error was found +func StrictDecode(into interface{}, from []byte) error { + dec := json.NewDecoder(strings.NewReader(string(from))) + dec.DisallowUnknownFields() + + if err := dec.Decode(into); err != nil { + return err
Quoted 8 lines
scripts/qapi/golang.py line 48
Quoted 3 lines
scripts/qapi/golang.py +3
-0+ } + return nil +}The documentation doesn't seem to be consistent with how the function actually works: AFAICT it returns false *with an error* for any failure, including those caused by unknown fields being present in the input JSON.
You are correct, documentation of this helper function needs to be fixed if we keep the helper function, as you made a good point about backwards-compatible decoding a struct that might have introduced extra fields in a newer version.
Quoted 15 lines
Looking at the generated code:
Quoted 5 lines
type BlockdevRef struct {
Definition *BlockdevOptions
Reference *string
}Quoted 7 lines
func (s BlockdevRef) MarshalJSON() ([]byte, error) { if s.Definition != nil { return json.Marshal(s.Definition) } else if s.Reference != nil { return json.Marshal(s.Reference) } return nil, errors.New("BlockdevRef has empty fields")
Quoted 3 lines
Returning an error if no field is set is good. Can we be more strict and returning one if more than one is set as well? That feels better than just picking the first one.
Sure.
Quoted 12 lines
Quoted 11 lines
func (s *BlockdevRef) UnmarshalJSON(data []byte) error { // Check for json-null first if string(data) == "null" { return errors.New(`null not supported for BlockdevRef`) } // Check for BlockdevOptions { s.Definition = new(BlockdevOptions) if err := StrictDecode(s.Definition, data); err == nil { return nil }
Quoted 3 lines
The use of
StrictDecode()here means that we won't be able to parse an alternate produced by a version of QEMU where BlockdevOptions has gained additional fields, doesn't it?
That's correct. This means that with this RFCv2 proposal, qapi-go based on qemu version 7.1 might not be able to decode a qmp message from qemu version 7.2 if it has introduced a new field.
This needs fixing, not sure yet the way to go.
Quoted 3 lines
Considering that we will happily parse such a BlockdevOptions outside of the context of BlockdevRef, I think we should be consistent and allow the same to happen here.
StrictDecode is only used with alternates because, unlike unions, Alternate types don't have a 'discriminator' field that would allow us to know what data type to expect.
With this in mind, theoretically speaking, we could have very similar struct types as Alternate fields and we have to find on runtime which type is that underlying byte stream.
So, to reply to your suggestion, if we allow BlockdevRef without StrictDecode we might find ourselves in a situation that it matched a few fields of BlockdevOptions but it the byte stream was actually another type.
So, I acknowledge that current version of this patch is not correct but we still need to look into the fields and see if there is a perfect match with the information that we have.
Quoted 11 lines
Quoted 11 lines
s.Definition = nil } // Check for string { s.Reference = new(string) if err := StrictDecode(s.Reference, data); err == nil { return nil } s.Reference = nil }
Quoted 7 lines
Quoted 1 line
return
errors.New(fmt.Sprintf("Can't convert to BlockdevRef: %s", string(data)))On a similar note to the MarshalJSON comment above, I'm not sure this is right.
If we find that more than one field of the alternate is set, we should error out - that's just invalid JSON we're dealing with.
With StrictDecode (or something similar) there shouldn't be multiple fields being set on the Go structure. Once it finds a match, it returns nil (no error)
Quoted 14 lines
But if we couldn't find any, that might be valid JSON that's been produced by a version of QEMU that introduced additional options to the alternate. In the spirit of "be liberal in what you accept", I think we should probably not reject that? Of course then the client code will have to look like
if r.Definition != nil { // do something with r.Definition } else if r.Reference != nil { // do something with r.Reference } else { // we don't recognize this - error out }
Quoted 2 lines
but the same is going to be true for enums, events and everything else that can be extended in a backwards-compatible fashion.
The client code looking like above is not much different than
if err := json.Unmarshal([]byte(input), &r); err != nil {
// error out: bad json? incompatibility issue?
} else if r.Definition != nil {
// do something with r.Definition
} else if r.Reference != nil {
// do something with r.Reference
} else {
// empty might be valid though, e.g: JSON null
}My main concern with alternate logic is not putting data that should go to r.Reference into r.Definition and vice versa.
I took note of all your suggestions, I'll be reworking this. I'll revisit the problem with StrictDecode together with addressing my reply to Daniel:
https://lists.gnu.org/archive/html/qemu-devel/2022-06/msg03140.html
Cheers,
Victor
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
Quoted 32 lines · Victor Toso
Quoted 13 lines · Andrea Bolognani
Quoted 11 lines · Victor Toso
func (s *BlockdevRef) UnmarshalJSON(data []byte) error { // Check for json-null first if string(data) == "null" { return errors.New(`null not supported for BlockdevRef`) } // Check for BlockdevOptions { s.Definition = new(BlockdevOptions) if err := StrictDecode(s.Definition, data); err == nil { return nil }Quoted 3 lines
The use of
StrictDecode()here means that we won't be able to parse an alternate produced by a version of QEMU where BlockdevOptions has gained additional fields, doesn't it?That's correct. This means that with this RFCv2 proposal, qapi-go based on qemu version 7.1 might not be able to decode a qmp message from qemu version 7.2 if it has introduced a new field.
This needs fixing, not sure yet the way to go.
Quoted 3 lines
Considering that we will happily parse such a BlockdevOptions outside of the context of BlockdevRef, I think we should be consistent and allow the same to happen here.
StrictDecode is only used with alternates because, unlike unions, Alternate types don't have a 'discriminator' field that would allow us to know what data type to expect.
Quoted 8 lines
With this in mind, theoretically speaking, we could have very similar struct types as Alternate fields and we have to find on runtime which type is that underlying byte stream.
So, to reply to your suggestion, if we allow BlockdevRef without StrictDecode we might find ourselves in a situation that it matched a few fields of BlockdevOptions but it the byte stream was actually another type.
IIUC your concern is that the QAPI schema could gain a new type, TotallyNotBlockdevOptions, which looks exactly like BlockdevOptions except for one or more extra fields.
If QEMU then produced a JSON like
{ "description": { /* a TotallyNotBlockdevOptions here */ } }
and we'd try to deserialize it with Go code like
ref := BlockdevRef{}
json.Unmarsal(&ref)
we'd end up mistakenly parsing the TotallyNotBlockdevOptions as a valid BlockdevOptions, dropping the extra fields in the process.
Does that correctly describe the reason why you feel that the use of StrictDecode is necessary?
If so, I respectfully disagree :)
If the client code is expecting a BlockdevRef as the return value of a command and QEMU is producing something that is *not* a BlockdevRef instead, that's an obvious bug in QEMU. If the client code is expecting a BlockdevRef as the return value of a command that is specified *not* to return a BlockdevRef, that's an obvious bug in the client code.
In neither case it should be the responsibility of the SDK to second-guess the declared intent, especially when it's perfectly valid for a type to be extended in a backwards-compatible way by adding fields to it.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
Hi,
Quoted 52 lines · Andrea Bolognani
Quoted 32 lines · Victor Toso
Quoted 13 lines · Andrea Bolognani
Quoted 11 lines · Victor Toso
func (s *BlockdevRef) UnmarshalJSON(data []byte) error { // Check for json-null first if string(data) == "null" { return errors.New(`null not supported for BlockdevRef`) } // Check for BlockdevOptions { s.Definition = new(BlockdevOptions) if err := StrictDecode(s.Definition, data); err == nil { return nil }Quoted 3 lines
The use of
StrictDecode()here means that we won't be able to parse an alternate produced by a version of QEMU where BlockdevOptions has gained additional fields, doesn't it?That's correct. This means that with this RFCv2 proposal, qapi-go based on qemu version 7.1 might not be able to decode a qmp message from qemu version 7.2 if it has introduced a new field.
This needs fixing, not sure yet the way to go.
Quoted 3 lines
Considering that we will happily parse such a BlockdevOptions outside of the context of BlockdevRef, I think we should be consistent and allow the same to happen here.
StrictDecode is only used with alternates because, unlike unions, Alternate types don't have a 'discriminator' field that would allow us to know what data type to expect.
Quoted 8 lines
With this in mind, theoretically speaking, we could have very similar struct types as Alternate fields and we have to find on runtime which type is that underlying byte stream.
So, to reply to your suggestion, if we allow BlockdevRef without StrictDecode we might find ourselves in a situation that it matched a few fields of BlockdevOptions but it the byte stream was actually another type.
IIUC your concern is that the QAPI schema could gain a new type, TotallyNotBlockdevOptions, which looks exactly like BlockdevOptions except for one or more extra fields.
If QEMU then produced a JSON like
{ "description": { /* a TotallyNotBlockdevOptions here */ } }
and we'd try to deserialize it with Go code like
Quoted 8 lines
ref := BlockdevRef{}
json.Unmarsal(&ref)we'd end up mistakenly parsing the TotallyNotBlockdevOptions as a valid BlockdevOptions, dropping the extra fields in the process.
Does that correctly describe the reason why you feel that the use of StrictDecode is necessary?
Not quite. The problem here is related to the Alternate types of the QAPI specification [0], just to name a simple in-use example, BlockdevRefOrNul [1].
[0] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/qapi-code-gen.rst?plain=1#L387
[1] https://gitlab.com/qemu-project/qemu/-/blob/master/qapi/block-core.json#L4349
To exemplify the problem that I try to solve with StrictDecode, let's say there is a DeviceRef alternate type that looks like:
{ 'alternate': 'DeviceRef',
'data': { 'memory': 'BlockdevRefInMemory',
'disk': 'BlockdevRefInDisk',
'cloud': 'BlockdevRefInCloud' } }Just a quick recap, at runtime we don't have data's payload name (e.g: disk). We need to check the actual data and try to find what is the payload type.
type BlockdevRefInMemory struct {
Name *string
Size uint64
Start uint64
End uint64
}
type BlockdevRefInDisk struct {
Name *string
Size uint64
Path *string}
type BlockdevRefInCloud struct {
Name *string
Size uint64
Uri *string}
All types have unique fields but they all share some fields too. Golang, without StrictDecode would happily decode a "disk" payload into either "memory" or "cloud" fields, matching only what the json provides and ignoring the rest. StrictDecode would error if the payload had fields that don't belong to that Type so we could try to find a perfect match.
While this should work reliably with current version of QEMU's qapi-schema.json, it is not future error proof... It is just a bit better than not checking at all.
The overall expectations is that, if the fields have too much in common, it is likely they should have been tagged as 'union' instead of 'alternate'. Yet, because alternate types have this flexibility, we need to be extra careful.
I'm still thinking about this in a way that would not give too much hassle when considering a generated code that talks with older/newer qemu where some fields might have been added/removed.
Quoted 9 lines
If so, I respectfully disagree :)
If the client code is expecting a BlockdevRef as the return value of a command and QEMU is producing something that is *not* a BlockdevRef instead, that's an obvious bug in QEMU. If the client code is expecting a BlockdevRef as the return value of a command that is specified *not* to return a BlockdevRef, that's an obvious bug in the client code.
Quoted 4 lines
In neither case it should be the responsibility of the SDK to second-guess the declared intent, especially when it's perfectly valid for a type to be extended in a backwards-compatible way by adding fields to it.
Yes, the SDK should consider valid QMP messages. This specific case is just a bit more complex qapi type that SDK needs to worry.
Thanks for your input!
Victor
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
Quoted 96 lines · Victor Toso
Hi,
Quoted 52 lines · Andrea Bolognani
Quoted 32 lines · Victor Toso
Quoted 13 lines · Andrea Bolognani
Quoted 11 lines · Victor Toso
func (s *BlockdevRef) UnmarshalJSON(data []byte) error { // Check for json-null first if string(data) == "null" { return errors.New(`null not supported for BlockdevRef`) } // Check for BlockdevOptions { s.Definition = new(BlockdevOptions) if err := StrictDecode(s.Definition, data); err == nil { return nil }Quoted 3 lines
The use of
StrictDecode()here means that we won't be able to parse an alternate produced by a version of QEMU where BlockdevOptions has gained additional fields, doesn't it?That's correct. This means that with this RFCv2 proposal, qapi-go based on qemu version 7.1 might not be able to decode a qmp message from qemu version 7.2 if it has introduced a new field.
This needs fixing, not sure yet the way to go.
Quoted 3 lines
Considering that we will happily parse such a BlockdevOptions outside of the context of BlockdevRef, I think we should be consistent and allow the same to happen here.
StrictDecode is only used with alternates because, unlike unions, Alternate types don't have a 'discriminator' field that would allow us to know what data type to expect.
Quoted 8 lines
With this in mind, theoretically speaking, we could have very similar struct types as Alternate fields and we have to find on runtime which type is that underlying byte stream.
So, to reply to your suggestion, if we allow BlockdevRef without StrictDecode we might find ourselves in a situation that it matched a few fields of BlockdevOptions but it the byte stream was actually another type.
IIUC your concern is that the QAPI schema could gain a new type, TotallyNotBlockdevOptions, which looks exactly like BlockdevOptions except for one or more extra fields.
If QEMU then produced a JSON like
{ "description": { /* a TotallyNotBlockdevOptions here */ } }
and we'd try to deserialize it with Go code like
Quoted 8 lines
ref := BlockdevRef{}
json.Unmarsal(&ref)we'd end up mistakenly parsing the TotallyNotBlockdevOptions as a valid BlockdevOptions, dropping the extra fields in the process.
Does that correctly describe the reason why you feel that the use of StrictDecode is necessary?
Not quite. The problem here is related to the Alternate types of the QAPI specification [0], just to name a simple in-use example, BlockdevRefOrNul [1].
[0] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/qapi-code-gen.rst?plain=1#L387
[1] https://gitlab.com/qemu-project/qemu/-/blob/master/qapi/block-core.json#L4349To exemplify the problem that I try to solve with StrictDecode, let's say there is a DeviceRef alternate type that looks like:
{ 'alternate': 'DeviceRef',
'data': { 'memory': 'BlockdevRefInMemory','disk': 'BlockdevRefInDisk', 'cloud': 'BlockdevRefInCloud' } }Just a quick recap, at runtime we don't have data's payload name (e.g: disk). We need to check the actual data and try to find what is the payload type.
type BlockdevRefInMemory struct {
Name *string
Size uint64
Start uint64
End uint64
}type BlockdevRefInDisk struct {
Name *string Size uint64 Path *string}
Quoted 7 lines
type BlockdevRefInCloud struct {
Name *string Size uint64 Uri *string}
All types have unique fields but they all share some fields too.
Quick intercession (I merely skimmed the review thread; forgive me if it's not useful or not new):
An alternate type is like a union type, except there is no discriminator on the wire. Instead, the branch to use is inferred from the value. An alternate can only express a choice between types represented differently on the wire.
This is docs/devel/qapi-code-gen.rst. Implied there: the inference is based on the JSON type *only*, i.e. no two branches can have the same JSON type on the wire. Since all complex types (struct or union) are JSON object on the wire, at most one alternate branch can be of complex type.
More sophisticated inference would be possible if we need it. So far we haven't.
Quoted 39 lines
Golang, without StrictDecode would happily decode a "disk" payload into either "memory" or "cloud" fields, matching only what the json provides and ignoring the rest. StrictDecode would error if the payload had fields that don't belong to that Type so we could try to find a perfect match.
While this should work reliably with current version of QEMU's qapi-schema.json, it is not future error proof... It is just a bit better than not checking at all.
The overall expectations is that, if the fields have too much in common, it is likely they should have been tagged as 'union' instead of 'alternate'. Yet, because alternate types have this flexibility, we need to be extra careful.
I'm still thinking about this in a way that would not give too much hassle when considering a generated code that talks with older/newer qemu where some fields might have been added/removed.
Quoted 9 lines
If so, I respectfully disagree :)
If the client code is expecting a BlockdevRef as the return value of a command and QEMU is producing something that is *not* a BlockdevRef instead, that's an obvious bug in QEMU. If the client code is expecting a BlockdevRef as the return value of a command that is specified *not* to return a BlockdevRef, that's an obvious bug in the client code.
Quoted 4 lines
In neither case it should be the responsibility of the SDK to second-guess the declared intent, especially when it's perfectly valid for a type to be extended in a backwards-compatible way by adding fields to it.
Yes, the SDK should consider valid QMP messages. This specific case is just a bit more complex qapi type that SDK needs to worry.
Thanks for your input!
Victor
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
Hi,
Quoted 114 lines · Markus Armbruster
Quoted 96 lines · Victor Toso
Hi,
Quoted 52 lines · Andrea Bolognani
Quoted 32 lines · Victor Toso
Quoted 13 lines · Andrea Bolognani
Quoted 11 lines · Victor Toso
func (s *BlockdevRef) UnmarshalJSON(data []byte) error { // Check for json-null first if string(data) == "null" { return errors.New(`null not supported for BlockdevRef`) } // Check for BlockdevOptions { s.Definition = new(BlockdevOptions) if err := StrictDecode(s.Definition, data); err == nil { return nil }Quoted 3 lines
The use of
StrictDecode()here means that we won't be able to parse an alternate produced by a version of QEMU where BlockdevOptions has gained additional fields, doesn't it?That's correct. This means that with this RFCv2 proposal, qapi-go based on qemu version 7.1 might not be able to decode a qmp message from qemu version 7.2 if it has introduced a new field.
This needs fixing, not sure yet the way to go.
Quoted 3 lines
Considering that we will happily parse such a BlockdevOptions outside of the context of BlockdevRef, I think we should be consistent and allow the same to happen here.
StrictDecode is only used with alternates because, unlike unions, Alternate types don't have a 'discriminator' field that would allow us to know what data type to expect.
Quoted 8 lines
With this in mind, theoretically speaking, we could have very similar struct types as Alternate fields and we have to find on runtime which type is that underlying byte stream.
So, to reply to your suggestion, if we allow BlockdevRef without StrictDecode we might find ourselves in a situation that it matched a few fields of BlockdevOptions but it the byte stream was actually another type.
IIUC your concern is that the QAPI schema could gain a new type, TotallyNotBlockdevOptions, which looks exactly like BlockdevOptions except for one or more extra fields.
If QEMU then produced a JSON like
{ "description": { /* a TotallyNotBlockdevOptions here */ } }
and we'd try to deserialize it with Go code like
Quoted 8 lines
ref := BlockdevRef{}
json.Unmarsal(&ref)we'd end up mistakenly parsing the TotallyNotBlockdevOptions as a valid BlockdevOptions, dropping the extra fields in the process.
Does that correctly describe the reason why you feel that the use of StrictDecode is necessary?
Not quite. The problem here is related to the Alternate types of the QAPI specification [0], just to name a simple in-use example, BlockdevRefOrNul [1].
[0] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/qapi-code-gen.rst?plain=1#L387
[1] https://gitlab.com/qemu-project/qemu/-/blob/master/qapi/block-core.json#L4349To exemplify the problem that I try to solve with StrictDecode, let's say there is a DeviceRef alternate type that looks like:
{ 'alternate': 'DeviceRef',
'data': { 'memory': 'BlockdevRefInMemory','disk': 'BlockdevRefInDisk', 'cloud': 'BlockdevRefInCloud' } }Just a quick recap, at runtime we don't have data's payload name (e.g: disk). We need to check the actual data and try to find what is the payload type.
type BlockdevRefInMemory struct {
Name *string
Size uint64
Start uint64
End uint64
}type BlockdevRefInDisk struct {
Name *string Size uint64 Path *string}
Quoted 7 lines
type BlockdevRefInCloud struct {
Name *string Size uint64 Uri *string}
All types have unique fields but they all share some fields too.
Quick intercession (I merely skimmed the review thread; forgive me if it's not useful or not new):
An alternate type is like a union type, except there is no discriminator on the wire. Instead, the branch to use is inferred from the value. An alternate can only express a choice between types represented differently on the wire.
Quoted 5 lines
This is
docs/devel/qapi-code-gen.rst. Implied there: the inference is based on the JSON type *only*, i.e. no two branches can have the same JSON type on the wire. Since all complex types (struct or union) are JSON object on the wire, at most one alternate branch can be of complex type.
Ah, I've missed this bit. Thank you, it does make it much simpler.
Quoted 2 lines
More sophisticated inference would be possible if we need it. So far we haven't.
Ack.
Cheers,
Victor
Quoted 40 lines
Quoted 39 lines
Golang, without StrictDecode would happily decode a "disk" payload into either "memory" or "cloud" fields, matching only what the json provides and ignoring the rest. StrictDecode would error if the payload had fields that don't belong to that Type so we could try to find a perfect match.
While this should work reliably with current version of QEMU's qapi-schema.json, it is not future error proof... It is just a bit better than not checking at all.
The overall expectations is that, if the fields have too much in common, it is likely they should have been tagged as 'union' instead of 'alternate'. Yet, because alternate types have this flexibility, we need to be extra careful.
I'm still thinking about this in a way that would not give too much hassle when considering a generated code that talks with older/newer qemu where some fields might have been added/removed.
Quoted 9 lines
If so, I respectfully disagree :)
If the client code is expecting a BlockdevRef as the return value of a command and QEMU is producing something that is *not* a BlockdevRef instead, that's an obvious bug in QEMU. If the client code is expecting a BlockdevRef as the return value of a command that is specified *not* to return a BlockdevRef, that's an obvious bug in the client code.
Quoted 4 lines
In neither case it should be the responsibility of the SDK to second-guess the declared intent, especially when it's perfectly valid for a type to be extended in a backwards-compatible way by adding fields to it.
Yes, the SDK should consider valid QMP messages. This specific case is just a bit more complex qapi type that SDK needs to worry.
Thanks for your input!
Victor
Re: [RFC PATCH v2 2/8] qapi: golang: Generate qapi's alternate types in Go
In reply to Victor TosoHi,
Quoted 9 lines · Victor Toso
This patch handles QAPI alternate types and generates data structures in Go that handles it.
At this moment, there are 5 alternates in qemu/qapi, they are:
* BlockDirtyBitmapMergeSource
* Qcow2OverlapChecks
* BlockdevRef
* BlockdevRefOrNull
* StrOrNull
So, things got a little bit complicated due the fact that BlockdevRefOrNull and StrOrNull can take JSON NULL value and that's completely different than an omitted field.
The last reply from Markus in another patch series make this clear with a good deal of examples too.
https://lists.gnu.org/archive/html/qemu-devel/2022-09/msg00113.html
I'll put the struggle that I'm having just in case someone have suggestions. Let's get an example:
In qemu/qapi/block-core.json:
{ 'struct': 'BlockdevOptionsQcow2',
'base': 'BlockdevOptionsGenericCOWFormat',
'data': { '*lazy-refcounts': 'bool',
'*pass-discard-request': 'bool',
'*pass-discard-snapshot': 'bool',
'*pass-discard-other': 'bool',
'*overlap-check': 'Qcow2OverlapChecks',
'*cache-size': 'int',
'*l2-cache-size': 'int',
'*l2-cache-entry-size': 'int',
'*refcount-cache-size': 'int',
'*cache-clean-interval': 'int',
'*encrypt': 'BlockdevQcow2Encryption',
'*data-file': 'BlockdevRef' } }Generates in qapi-go/pkg/qapi/structs.go:
type BlockdevOptionsQcow2 struct {
// Base fields
Backing *BlockdevRefOrNull json:"backing,omitempty"
LazyRefcounts *bool json:"lazy-refcounts,omitempty"
PassDiscardRequest *bool json:"pass-discard-request,omitempty"
PassDiscardSnapshot *bool json:"pass-discard-snapshot,omitempty"
PassDiscardOther *bool json:"pass-discard-other,omitempty"
OverlapCheck *Qcow2OverlapChecks json:"overlap-check,omitempty"
CacheSize *int64 json:"cache-size,omitempty"
L2CacheSize *int64 json:"l2-cache-size,omitempty"
L2CacheEntrySize *int64 json:"l2-cache-entry-size,omitempty"
RefcountCacheSize *int64 json:"refcount-cache-size,omitempty"
CacheCleanInterval *int64 json:"cache-clean-interval,omitempty"
Encrypt *BlockdevQcow2Encryption json:"encrypt,omitempty"
DataFile *BlockdevRef json:"data-file,omitempty"
}
One thing that we assumed for all types is that an optional type can be simply ignored. For that to work with Golang's encoding/json, we make do by making every optional field a pointer to the type it represents, plus we add the "omitempty" tag. This way, if the user did not set it, it is simply ignored. If we didn't receive anything, all should be good.
The problem happens when we receive a JSON Null value, which is one possible value type for the BlockdevRefOrNull above. A message like { "backing" : null } does not trigger UnmarshalJSON for BlockdevRefOrNull and this is silently ignored.
In Go's encoding/json, in order to { "backing" : null } trigger UnmarshalJSON, we need to remove omitempty and also the pointer from Backing. Now, on the marshalling side, we will always have 'backing' being set and in order to omit it from the output (which we might want indeed) we need to rework the byte array, which is something we were trying to avoid :)
I've looked up and there has been proposals to address this kind of issue, including an omitnil tag [0] that could be used here to workaround it but that's unlikely to move forward [1].
The first thing to do is to define when to omit *StrOrNull and *BlockdevRefOrNull and when to set it to null. The solution I thought would make sense is to have a Null boolean field that user would need to set, so:
type BlockdevRefOrNull struct {
Definition *BlockdevOptions
Reference *string
Null bool
}
func (s BlockdevRefOrNull) MarshalJSON() ([]byte, error) {
if s.Definition != nil {
return json.Marshal(s.Definition)
} else if s.Reference != nil {
return json.Marshal(s.Reference)
} else if s.Null {
return []byte("null"), nil
}
// If only QAPI spec accepted "backing": {} as the same
// as ommited.
return []byte("qapi-go-remove-this-object"), nil
}
func (s *BlockdevRefOrNull) UnmarshalJSON(data []byte) error {
// Check for json-null first
if string(data) == "null" {
s.Null = true
return nil
}
// Check for BlockdevOptions
...
}Setting BlockdevRefOrNull to null and to StrOrNull could have different meanings which is explained in the documentation itself. That's why I think Null as a boolean is fine, the user sets for a specific context when it knows what it is doing...
Not having a pointer for optional fields of this two types breaks the coherence we had that all optional members are pointers in Go. This hurts a bit but this two are truly special as they can have the magic null value.
Now, I'm thinking on a reasonable way to remove the object that wraps "qapi-go-remove-this-object" as well, likely using the json's decoder [2]...
[0] https://github.com/golang/go/issues/22480
[1] https://github.com/golang/go/issues/34701#issuecomment-544710311
[2] https://pkg.go.dev/encoding/json#Decoder.Token
Cheers,
Victor
RFC PATCH v2 3/8 qapi: golang: Generate qapi's struct types in Go
In reply to Victor TosoThis patch handles QAPI struct types and generates the equivalent types in Go.
At the time of this writing, it generates 388 structures.
The highlights of this implementation are:
1. Generating an Go struct that requires a @base type, the @base type fields are copied over to the Go struct. The advantage of this approach is to not have embed structs in any of the QAPI types. The downside are some generated Types that are likely useless now, like InetSocketAddressBase from InetSocketAddress.
2. About the Go struct's fields:
i) They can be either by Value or Reference.
ii) Every field that is marked as optional in the QAPI specification are translated to Reference fields in its Go structure. This design decision is the most straightforward way to check if a given field was set or not.
iii) Mandatory fields are always by Value with the exception of QAPI arrays, which are handled by Reference (to a block of memory) by Go.
iv) All the fields are named with Uppercase due Golang's export convention.
v) In order to avoid any kind of issues when encoding ordecoding, to or from JSON, we mark all fields with its @name and, when it is optional, member, with @omitempty
Example:
qapi:
| { 'struct': 'BlockdevCreateOptionsFile',
| 'data': { 'filename': 'str',
| 'size': 'size',
| '*preallocation': 'PreallocMode',
| '*nocow': 'bool',
| '*extent-size-hint': 'size'} }go:
| type BlockdevCreateOptionsFile struct {
| Filename string `json:"filename"`
| Size uint64 `json:"size"`
| Preallocation *PreallocMode `json:"preallocation,omitempty"`
| Nocow *bool `json:"nocow,omitempty"`
| ExtentSizeHint *uint64 `json:"extent-size-hint,omitempty"`
| }--- scripts/qapi/golang.py | 117 ++++++++++++++++++++++++++++++++++++++++- 1 file changed, 115 insertions(+), 2 deletions(-)
scripts/qapi/golang.py +115 -2
@@ -53,7 +53,7 @@ class QAPISchemaGenGolangVisitor(QAPISchemaVisitor): def __init__(self, prefix: str): super().__init__() - self.target = {name: "" for name in ["alternate", "enum", "helper"]} + self.target = {name: "" for name in ["alternate", "enum", "helper", "struct"]} self.objects_seen = {} self.schema = None self.golang_package_name = "qapi"
@@ -79,7 +79,37 @@ def visit_object_type(self: QAPISchemaGenGolangVisitor, members: List[QAPISchemaObjectTypeMember], variants: Optional[QAPISchemaVariants] ) -> None: - pass + # Do not handle anything besides structs + if (name == self.schema.the_empty_object_type.name or + not isinstance(name, str) or + info.defn_meta not in ["struct"]): + return + + # Safety checks. + assert name not in self.objects_seen + self.objects_seen[name] = True + + # visit all inner objects as well, they are not going to be + # called by python's generator. + if variants: + for var in variants.variants: + assert isinstance(var.type, QAPISchemaObjectType) + self.visit_object_type(self, + var.type.name, + var.type.info, + var.type.ifcond, + var.type.base, + var.type.local_members, + var.type.variants) + + # Save generated Go code to be written later + self.target[info.defn_meta] += qapi_to_golang_struct(name, + info, + ifcond, + features, + base, + members, + variants) def visit_alternate_type(self: QAPISchemaGenGolangVisitor, name: str,
@@ -223,6 +253,72 @@ def generate_struct_type(type_name, args="") -> str: ''' +# Helper function that is used for most of QAPI types +def qapi_to_golang_struct(name: str, + info: Optional[QAPISourceInfo], + ifcond: QAPISchemaIfCond, + features: List[QAPISchemaFeature], + base: Optional[QAPISchemaObjectType], + members: List[QAPISchemaObjectTypeMember], + variants: Optional[QAPISchemaVariants]) -> str: + + type_name = qapi_to_go_type_name(name, info.defn_meta) + + fields = "" + + if base: + base_fields = "" + for lm in base.local_members: + # We do not include the Union's discriminator + # into the generated Go struct as the selection + # of what variant was set is enough on Go side. + if variants and variants.tag_member.name == lm.name: + continue + + field = qapi_to_field_name(lm.name) + member_type = qapi_schema_type_to_go_type(lm.type.name) + + isptr = "*" if lm.optional and member_type[0] not in "*[" else "" + optional = ",omitempty" if lm.optional else "" + fieldtag = f'`json:"{lm.name}{optional}"`' + + base_fields += f"\t{field} {isptr}{member_type}{fieldtag}\n" + + if len(base_fields) > 0: + fields += f"\t// Base fields\n\t{base_fields}\n" + + if members: + for memb in members: + field = qapi_to_field_name(memb.name) + member_type = qapi_schema_type_to_go_type(memb.type.name) + + isptr = "*" if memb.optional and member_type[0] not in "*[" else "" + optional = ",omitempty" if memb.optional else "" + fieldtag = f'`json:"{memb.name}{optional}"`' + + fields += f"\t{field} {isptr}{member_type}{fieldtag}\n" + + fields += "\n" + + if variants: + fields += "\t// Variants fields\n" + for var in variants.variants: + if var.type.is_implicit(): + continue + + field = qapi_to_field_name(var.name) + member_type = qapi_schema_type_to_go_type(var.type.name) + # Variant's are handled in the Marshal/Unmarshal methods + fieldtag = '`json:"-"`' + fields += f"\t{field} *{member_type}{fieldtag}\n" + member_type = qapi_schema_type_to_go_type(var.type.name) + # Variant's are handled in the Marshal/Unmarshal methods + fieldtag = '`json:"-"`' + fields += f"\t{field} *{member_type}{fieldtag}\n" + + return generate_struct_type(type_name, fields) + + def qapi_schema_type_to_go_type(type: str) -> str: schema_types_to_go = { 'str': 'string', 'null': 'nil', 'bool': 'bool', 'number':
@@ -247,3 +343,20 @@ def qapi_to_field_name_enum(name: str) -> str: def qapi_to_field_name(name: str) -> str: return name.title().replace("_", "").replace("-", "") + + +def qapi_to_go_type_name(name: str, meta: str) -> str: + if name.startswith("q_obj_"): + name = name[6:] + + # We want to keep CamelCase for Golang types. We want to avoid removing + # already set CameCase names while fixing uppercase ones, eg: + # 1) q_obj_SocketAddress_base -> SocketAddressBase + # 2) q_obj_WATCHDOG-arg -> WatchdogArg + words = [word for word in name.replace("_", "-").split("-")] + name = words[0] + if name.islower() or name.isupper(): + name = name.title() + + name += ''.join(word.title() for word in words[1:]) + return name
Signature · 1 line
2.36.1
Re: [RFC PATCH v2 3/8] qapi: golang: Generate qapi's struct types in Go
Quoted 42 lines · Victor Toso
This patch handles QAPI struct types and generates the equivalent types in Go.
At the time of this writing, it generates 388 structures.
The highlights of this implementation are:
1. Generating an Go struct that requires a @base type, the @base type fields are copied over to the Go struct. The advantage of this approach is to not have embed structs in any of the QAPI types. The downside are some generated Types that are likely useless now, like InetSocketAddressBase from InetSocketAddress.
2. About the Go struct's fields:
i) They can be either by Value or Reference.
ii) Every field that is marked as optional in the QAPI specification are translated to Reference fields in its Go structure. This design decision is the most straightforward way to check if a given field was set or not.
iii) Mandatory fields are always by Value with the exception of QAPI arrays, which are handled by Reference (to a block of memory) by Go.
iv) All the fields are named with Uppercase due Golang's export convention.
v) In order to avoid any kind of issues when encoding ordecoding, to or from JSON, we mark all fields with its @name and, when it is optional, member, with @omitempty
Example:
qapi:
| { 'struct': 'BlockdevCreateOptionsFile', | 'data': { 'filename': 'str', | 'size': 'size', | '*preallocation': 'PreallocMode', | '*nocow': 'bool', | '*extent-size-hint': 'size'} }
Quoted 8 lines
go:
| type BlockdevCreateOptionsFile struct { | Filename string `json:"filename"` | Size uint64 `json:"size"` | Preallocation *PreallocMode `json:"preallocation,omitempty"` | Nocow *bool `json:"nocow,omitempty"` | ExtentSizeHint *uint64 `json:"extent-size-hint,omitempty"` | }
One thing to bear in mind here
At the QAPI level, changing a field from mandatory to optional has been considered a backwards compatible change by QEMU maintainers, because any existing caller can happily continue passing the optional field with no downside.
With this Go design, changing a field from mandatory to optional will be an API breakage, because the developer will need to change from passing a literal value, to a pointer to the value, when initializing the struct.
IOW, this Go impl provides weaker compat guarantees than even QAPI does, and QAPI compat guarantees were already weaker than I would like as an app developer.
If we want to make ourselves future proof, we would have to make all struct fields optional from the start, even if they are mandatory at QAPI level. This would make the code less self-documenting though, so that's not very appealing either.
If we want to avoid this, we would need the same approach I suggested wrt support multiple versions of the API concurrently. Namely have versioned structs, so every time there's a field change of any kind, we introduce a new struct version.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 3/8] qapi: golang: Generate qapi's struct types in Go
Hi,
Quoted 59 lines · Daniel P. Berrangé
Quoted 42 lines · Victor Toso
This patch handles QAPI struct types and generates the equivalent types in Go.
At the time of this writing, it generates 388 structures.
The highlights of this implementation are:
1. Generating an Go struct that requires a @base type, the @base type fields are copied over to the Go struct. The advantage of this approach is to not have embed structs in any of the QAPI types. The downside are some generated Types that are likely useless now, like InetSocketAddressBase from InetSocketAddress.
2. About the Go struct's fields:
i) They can be either by Value or Reference.
ii) Every field that is marked as optional in the QAPI specification are translated to Reference fields in its Go structure. This design decision is the most straightforward way to check if a given field was set or not.
iii) Mandatory fields are always by Value with the exception of QAPI arrays, which are handled by Reference (to a block of memory) by Go.
iv) All the fields are named with Uppercase due Golang's export convention.
v) In order to avoid any kind of issues when encoding ordecoding, to or from JSON, we mark all fields with its @name and, when it is optional, member, with @omitempty
Example:
qapi:
| { 'struct': 'BlockdevCreateOptionsFile', | 'data': { 'filename': 'str', | 'size': 'size', | '*preallocation': 'PreallocMode', | '*nocow': 'bool', | '*extent-size-hint': 'size'} }Quoted 8 lines
go:
| type BlockdevCreateOptionsFile struct { | Filename string `json:"filename"` | Size uint64 `json:"size"` | Preallocation *PreallocMode `json:"preallocation,omitempty"` | Nocow *bool `json:"nocow,omitempty"` | ExtentSizeHint *uint64 `json:"extent-size-hint,omitempty"` | }One thing to bear in mind here
At the QAPI level, changing a field from mandatory to optional has been considered a backwards compatible change by QEMU maintainers, because any existing caller can happily continue passing the optional field with no downside.
Quoted 8 lines
With this Go design, changing a field from mandatory to optional will be an API breakage, because the developer will need to change from passing a literal value, to a pointer to the value, when initializing the struct.
IOW, this Go impl provides weaker compat guarantees than even QAPI does, and QAPI compat guarantees were already weaker than I would like as an app developer.
I think the current draft should be considered an interface that can work with the QEMU version this was generated from. That is the first thing we should get right.
Quoted 10 lines
If we want to make ourselves future proof, we would have to make all struct fields optional from the start, even if they are mandatory at QAPI level. This would make the code less self-documenting though, so that's not very appealing either.
If we want to avoid this, we would need the same approach I suggested wrt support multiple versions of the API concurrently. Namely have versioned structs, so every time there's a field change of any kind, we introduce a new struct version.
That's more or less what I had in mind. I mentioned it in the item 8 of the cover-letter. I just did not want to address it at before deciding what the structs should look like first, for the version we are generating from.
Just to clarify, so far I plan to follow the suggestion:
https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg02147.html
Of course, If there are other ideas, we can discuss it too.
Cheers,
Victor
RFC PATCH v2 4/8 qapi: golang: Generate qapi's union types in Go
In reply to Victor TosoThis patch handles QAPI union types and generates the equivalent data structures and methods in Go to handle it.
At the moment of this writing, it generates 38 structures.
The QAPI union type has two types of fields: The @base and the @variants members. The @base fields can be considered common members for the union while only one field maximum is set for the @variants.
In the QAPI specification, it defines a @discriminator field, which is an Enum type. The purpose of the @discriminator is to identify which @variant type is being used. The @discriminator is not used in the generated union Go structs as it suffices to check which one of the @variants fields were set.
The union types implement the Marshaler and Unmarshaler interfaces to seamless decode from JSON objects to Golang structs and vice versa.
qapi:
| { 'union': 'SetPasswordOptions',
| 'base': { 'protocol': 'DisplayProtocol',
| 'password': 'str',
| '*connected': 'SetPasswordAction' },
| 'discriminator': 'protocol',
| 'data': { 'vnc': 'SetPasswordOptionsVnc' } }go:
| type SetPasswordOptions struct {
| // Base fields
| Password string `json:"password"`
| Connected *SetPasswordAction `json:"connected,omitempty"`
|
| // Variants fields
| Vnc *SetPasswordOptionsVnc `json:"-"`
| }--- scripts/qapi/golang.py | 112 ++++++++++++++++++++++++++++++++++++++--- 1 file changed, 105 insertions(+), 7 deletions(-)
scripts/qapi/golang.py +105 -7
@@ -53,7 +53,8 @@ class QAPISchemaGenGolangVisitor(QAPISchemaVisitor): def __init__(self, prefix: str): super().__init__() - self.target = {name: "" for name in ["alternate", "enum", "helper", "struct"]} + self.target = {name: "" for name in ["alternate", "enum", "helper", "struct", + "union"]} self.objects_seen = {} self.schema = None self.golang_package_name = "qapi"
@@ -79,10 +80,14 @@ def visit_object_type(self: QAPISchemaGenGolangVisitor, members: List[QAPISchemaObjectTypeMember], variants: Optional[QAPISchemaVariants] ) -> None: - # Do not handle anything besides structs + # Do not handle anything besides struct and unions. if (name == self.schema.the_empty_object_type.name or not isinstance(name, str) or - info.defn_meta not in ["struct"]): + info.defn_meta not in ["struct", "union"]): + return + + # Base structs are embed + if qapi_name_is_base(name): return # Safety checks.
@@ -110,6 +115,10 @@ def visit_object_type(self: QAPISchemaGenGolangVisitor, base, members, variants) + if info.defn_meta == "union": + self.target[info.defn_meta] += qapi_to_golang_methods_union(name, + info, + variants) def visit_alternate_type(self: QAPISchemaGenGolangVisitor, name: str,
@@ -311,14 +320,99 @@ def qapi_to_golang_struct(name: str, # Variant's are handled in the Marshal/Unmarshal methods fieldtag = '`json:"-"`' fields += f"\t{field} *{member_type}{fieldtag}\n" - member_type = qapi_schema_type_to_go_type(var.type.name) - # Variant's are handled in the Marshal/Unmarshal methods - fieldtag = '`json:"-"`' - fields += f"\t{field} *{member_type}{fieldtag}\n" return generate_struct_type(type_name, fields) +def qapi_to_golang_methods_union(name: str, + info: Optional[QAPISourceInfo], + variants: Optional[QAPISchemaVariants] + ) -> str: + + type_name = qapi_to_go_type_name(name, info.defn_meta) + + driverCases = "" + checkFields = "" + if variants: + for var in variants.variants: + if var.type.is_implicit(): + continue + + field = qapi_to_field_name(var.name) + member_type = qapi_schema_type_to_go_type(var.type.name) + + if len(checkFields) > 0: + checkFields += "\t} else " + checkFields += f'''if s.{field} != nil {{ + driver = "{var.name}" + payload, err = json.Marshal(s.{field}) +''' + # for Unmarshal method + driverCases += f''' + case "{var.name}": + s.{field} = new({member_type}) + if err := json.Unmarshal(data, s.{field}); err != nil {{ + s.{field} = nil + return err + }}''' + + checkFields += "}" + + return f''' +func (s {type_name}) MarshalJSON() ([]byte, error) {{ + type Alias {type_name} + alias := Alias(s) + base, err := json.Marshal(alias) + if err != nil {{ + return nil, err + }} + + driver := "" + payload := []byte{{}} + {checkFields} + + if err != nil {{ + return nil, err + }} + + if len(base) == len("{{}}") {{ + base = []byte("") + }} else {{ + base = append([]byte{{','}}, base[1:len(base)-1]...) + }} + + if len(payload) == 0 || len(payload) == len("{{}}") {{ + payload = []byte("") + }} else {{ + payload = append([]byte{{','}}, payload[1:len(payload)-1]...) + }} + + result := fmt.Sprintf(`{{"{variants.tag_member.name}":"%s"%s%s}}`, driver, base, payload) + return []byte(result), nil +}} + +func (s *{type_name}) UnmarshalJSON(data []byte) error {{ + type Alias {type_name} + peek := struct {{ + Alias + Driver string `json:"{variants.tag_member.name}"` + }}{{}} + + + if err := json.Unmarshal(data, &peek); err != nil {{ + return err + }} + *s = {type_name}(peek.Alias) + + switch peek.Driver {{ + {driverCases} + }} + // Unrecognizer drivers are silently ignored. + return nil +}} +''' + + def qapi_schema_type_to_go_type(type: str) -> str: schema_types_to_go = { 'str': 'string', 'null': 'nil', 'bool': 'bool', 'number':
@@ -345,6 +439,10 @@ def qapi_to_field_name(name: str) -> str: return name.title().replace("_", "").replace("-", "") +def qapi_name_is_base(name: str) -> bool: + return name.startswith("q_obj_") and name.endswith("-base") + + def qapi_to_go_type_name(name: str, meta: str) -> str: if name.startswith("q_obj_"): name = name[6:]
Signature · 1 line
2.36.1
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 9 lines · Victor Toso
qapi:
| { 'union': 'SetPasswordOptions', | 'base': { 'protocol': 'DisplayProtocol', | 'password': 'str', | '*connected': 'SetPasswordAction' }, | 'discriminator': 'protocol', | 'data': { 'vnc': 'SetPasswordOptionsVnc' } }go:
Quoted 8 lines
| type SetPasswordOptions struct {
| // Base fields
| Password string `json:"password"`
| Connected *SetPasswordAction `json:"connected,omitempty"`
|
| // Variants fields
| Vnc *SetPasswordOptionsVnc `json:"-"`
| }Generated code:
Quoted 34 lines
type GuestPanicInformation struct { // Variants fields HyperV *GuestPanicInformationHyperV `json:"-"` S390 *GuestPanicInformationS390 `json:"-"` }func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { type Alias GuestPanicInformation alias := Alias(s) base, err := json.Marshal(alias) if err != nil { return nil, err } driver := "" payload := []byte{} if s.HyperV != nil { driver = "hyper-v" payload, err = json.Marshal(s.HyperV) } else if s.S390 != nil { driver = "s390" payload, err = json.Marshal(s.S390) } if err != nil { return nil, err } if len(base) == len("{}") { base = []byte("") } else { base = append([]byte{','}, base[1:len(base)-1]...) }
Quoted 8 lines
if len(payload) == 0 || len(payload) == len("{}") {
payload = []byte("")
} else {
payload = append([]byte{','}, payload[1:len(payload)-1]...)
}
result := fmt.Sprintf(`{"type":"%s"%s%s}`, driver, base, payload)
return []byte(result), nilAll this string manipulation looks sketchy. Is there some reason that I'm not seeing preventing you for doing something like the untested code below?
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) {
if s.HyperV != nil {
type union struct {
Discriminator string `json:"type"`
HyperV GuestPanicInformationHyperV `json:"hyper-v"`
}
tmp := union {
Discriminator: "hyper-v",
HyperV: s.HyperV,
}
return json.Marshal(tmp)
} else if s.S390 != nil {
type union struct {
Discriminator string `json:"type"`
S390 GuestPanicInformationHyperV `json:"s390"`
}
tmp := union {
Discriminator: "s390",
S390: s.S390,
}
return json.Marshal(tmp)
}
return nil, errors.New("...")
}Quoted 21 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error {
type Alias GuestPanicInformation
peek := struct {
Alias
Driver string `json:"type"`
}{}
if err := json.Unmarshal(data, &peek); err != nil {
return err
}
*s = GuestPanicInformation(peek.Alias)
switch peek.Driver {
case "hyper-v":
s.HyperV = new(GuestPanicInformationHyperV)
if err := json.Unmarshal(data, s.HyperV); err != nil {
s.HyperV = nil
return err
}
case "s390":Quoted 8 lines
s.S390 = new(GuestPanicInformationS390)
if err := json.Unmarshal(data, s.S390); err != nil {
s.S390 = nil
return err
}
}
// Unrecognizer drivers are silently ignored.
return nilThis looks pretty reasonable, but since you're only using "peek" to look at the discriminator you should be able to leave out the Alias type entirely and perform the initial Unmarshal operation while ignoring all other fields.
The comments made elsewhere about potentially being more strict and rejecting invalid JSON also apply here.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 84 lines · Andrea Bolognani
Quoted 9 lines · Victor Toso
qapi:
| { 'union': 'SetPasswordOptions', | 'base': { 'protocol': 'DisplayProtocol', | 'password': 'str', | '*connected': 'SetPasswordAction' }, | 'discriminator': 'protocol', | 'data': { 'vnc': 'SetPasswordOptionsVnc' } }go:
Quoted 8 lines
| type SetPasswordOptions struct { | // Base fields | Password string `json:"password"` | Connected *SetPasswordAction `json:"connected,omitempty"` | | // Variants fields | Vnc *SetPasswordOptionsVnc `json:"-"` | }Generated code:
Quoted 34 lines
type GuestPanicInformation struct { // Variants fields HyperV *GuestPanicInformationHyperV `json:"-"` S390 *GuestPanicInformationS390 `json:"-"` }func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { type Alias GuestPanicInformation alias := Alias(s) base, err := json.Marshal(alias) if err != nil { return nil, err } driver := "" payload := []byte{} if s.HyperV != nil { driver = "hyper-v" payload, err = json.Marshal(s.HyperV) } else if s.S390 != nil { driver = "s390" payload, err = json.Marshal(s.S390) } if err != nil { return nil, err } if len(base) == len("{}") { base = []byte("") } else { base = append([]byte{','}, base[1:len(base)-1]...) }Quoted 8 lines
if len(payload) == 0 || len(payload) == len("{}") { payload = []byte("") } else { payload = append([]byte{','}, payload[1:len(payload)-1]...) } result := fmt.Sprintf(`{"type":"%s"%s%s}`, driver, base, payload) return []byte(result), nilAll this string manipulation looks sketchy. Is there some reason that I'm not seeing preventing you for doing something like the untested code below?
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if s.HyperV != nil { type union struct { Discriminator string `json:"type"` HyperV GuestPanicInformationHyperV `json:"hyper-v"` } tmp := union { Discriminator: "hyper-v", HyperV: s.HyperV, } return json.Marshal(tmp) } else if s.S390 != nil { type union struct { Discriminator string `json:"type"` S390 GuestPanicInformationHyperV `json:"s390"` }
Quoted 8 lines
tmp := union {
Discriminator: "s390",
S390: s.S390,
}
return json.Marshal(tmp)
}
return nil, errors.New("...")
}Using these dummy structs is the way I've approached the discriminated union issue in the libvirt Golang XML bindings and it works well. It is the bit I like the least, but it was the lesser of many evils, and on the plus side in the QEMU case it'll be auto-generated code.
Quoted 31 lines
Quoted 22 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { type Alias GuestPanicInformation peek := struct { Alias Driver string `json:"type"` }{} if err := json.Unmarshal(data, &peek); err != nil { return err } *s = GuestPanicInformation(peek.Alias) switch peek.Driver { case "hyper-v": s.HyperV = new(GuestPanicInformationHyperV) if err := json.Unmarshal(data, s.HyperV); err != nil { s.HyperV = nil return err } case "s390":Quoted 8 lines
s.S390 = new(GuestPanicInformationS390) if err := json.Unmarshal(data, s.S390); err != nil { s.S390 = nil return err } } // Unrecognizer drivers are silently ignored. return nil
Quoted 4 lines
This looks pretty reasonable, but since you're only using "peek" to look at the discriminator you should be able to leave out the Alias type entirely and perform the initial Unmarshal operation while ignoring all other fields.
Once you've defined the dummy structs for the Marshall case though, you might as well use them for Unmarshall too, so you're not parsing the JSON twice.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 30 lines · Daniel P. Berrangé
Quoted 20 lines · Andrea Bolognani
All this string manipulation looks sketchy. Is there some reason that I'm not seeing preventing you for doing something like the untested code below?
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if s.HyperV != nil { type union struct { Discriminator string `json:"type"` HyperV GuestPanicInformationHyperV `json:"hyper-v"` } tmp := union { Discriminator: "hyper-v", HyperV: s.HyperV, } return json.Marshal(tmp) } else if s.S390 != nil { type union struct { Discriminator string `json:"type"` S390 GuestPanicInformationHyperV `json:"s390"` }Quoted 8 lines
tmp := union { Discriminator: "s390", S390: s.S390, } return json.Marshal(tmp) } return nil, errors.New("...") }
Quoted 5 lines
Using these dummy structs is the way I've approached the discriminated union issue in the libvirt Golang XML bindings and it works well. It is the bit I like the least, but it was the lesser of many evils, and on the plus side in the QEMU case it'll be auto-generated code.
It appears to be the standard way to approach the problem in Go. It sort of comes naturally given how the APIs for marshal/unmarshal have been defined.
Quoted 30 lines
Quoted 30 lines
Quoted 29 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { type Alias GuestPanicInformation peek := struct { Alias Driver string `json:"type"` }{} if err := json.Unmarshal(data, &peek); err != nil { return err } *s = GuestPanicInformation(peek.Alias) switch peek.Driver { case "hyper-v": s.HyperV = new(GuestPanicInformationHyperV) if err := json.Unmarshal(data, s.HyperV); err != nil { s.HyperV = nil return err } case "s390": s.S390 = new(GuestPanicInformationS390) if err := json.Unmarshal(data, s.S390); err != nil { s.S390 = nil return err } } // Unrecognizer drivers are silently ignored. return nil
Quoted 8 lines
Quoted 4 lines
This looks pretty reasonable, but since you're only using "peek" to look at the discriminator you should be able to leave out the Alias type entirely and perform the initial Unmarshal operation while ignoring all other fields.
Once you've defined the dummy structs for the Marshall case though, you might as well use them for Unmarshall too, so you're not parsing the JSON twice.
You're right, that is undesirable. What about something like this?
type GuestPanicInformation struct {
HyperV *GuestPanicInformationHyperV
S390 *GuestPanicInformationS390
}
type jsonGuestPanicInformation struct {
Discriminator string `json:"type"`
HyperV *GuestPanicInformationHyperV `json:"hyper-v"`
S390 *GuestPanicInformationS390 `json:"s390"`
}
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) {
if (s.HyperV != nil && s.S390 != nil) ||
(s.HyperV == nil && s.S390 == nil) {
// client hasn't filled in the struct properly
return nil, errors.New("...")
} tmp := jsonGuestPanicInformation{}
if s.HyperV != nil {
tmp.Discriminator = "hyper-v"
tmp.HyperV = s.HyperV
} else if s.S390 != nil {
tmp.Discriminator = "s390"
tmp.S390 = s.S390
}
return json.Marshal(tmp)}
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error {
tmp := jsonGuestPanicInformation{} err := json.Unmarshal(data, &tmp)
if err != nil {
return err
}
switch tmp.Discriminator {
case "hyper-v":
if tmp.HyperV == nil {
return errors.New("...")
}
s.HyperV = tmp.HyperV
case "s390":
if tmp.S390 == nil {
return errors.New("...")
}
s.S390 = tmp.S390
}
// if we hit none of the cases above, that means the
// server has produced a variant we don't know about
return nil}
This avoid parsing the JSON twice as well as having to define multiple dummy structs, which keeps the code shorter and more readable.
I've also thrown in some additional error checking for good measure, ensuring that we abort when the input is completely nonsensical from a semantical standpoint.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 82 lines · Andrea Bolognani
Quoted 30 lines · Daniel P. Berrangé
Quoted 20 lines · Andrea Bolognani
All this string manipulation looks sketchy. Is there some reason that I'm not seeing preventing you for doing something like the untested code below?
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if s.HyperV != nil { type union struct { Discriminator string `json:"type"` HyperV GuestPanicInformationHyperV `json:"hyper-v"` } tmp := union { Discriminator: "hyper-v", HyperV: s.HyperV, } return json.Marshal(tmp) } else if s.S390 != nil { type union struct { Discriminator string `json:"type"` S390 GuestPanicInformationHyperV `json:"s390"` }Quoted 8 lines
tmp := union { Discriminator: "s390", S390: s.S390, } return json.Marshal(tmp) } return nil, errors.New("...") }Quoted 5 lines
Using these dummy structs is the way I've approached the discriminated union issue in the libvirt Golang XML bindings and it works well. It is the bit I like the least, but it was the lesser of many evils, and on the plus side in the QEMU case it'll be auto-generated code.
It appears to be the standard way to approach the problem in Go. It sort of comes naturally given how the APIs for marshal/unmarshal have been defined.
Quoted 30 lines
Quoted 30 lines
Quoted 29 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { type Alias GuestPanicInformation peek := struct { Alias Driver string `json:"type"` }{} if err := json.Unmarshal(data, &peek); err != nil { return err } *s = GuestPanicInformation(peek.Alias) switch peek.Driver { case "hyper-v": s.HyperV = new(GuestPanicInformationHyperV) if err := json.Unmarshal(data, s.HyperV); err != nil { s.HyperV = nil return err } case "s390": s.S390 = new(GuestPanicInformationS390) if err := json.Unmarshal(data, s.S390); err != nil { s.S390 = nil return err } } // Unrecognizer drivers are silently ignored. return nilQuoted 8 lines
Quoted 4 lines
This looks pretty reasonable, but since you're only using "peek" to look at the discriminator you should be able to leave out the Alias type entirely and perform the initial Unmarshal operation while ignoring all other fields.
Once you've defined the dummy structs for the Marshall case though, you might as well use them for Unmarshall too, so you're not parsing the JSON twice.
You're right, that is undesirable. What about something like this?
Quoted 10 lines
type GuestPanicInformation struct {
HyperV *GuestPanicInformationHyperV
S390 *GuestPanicInformationS390
}
type jsonGuestPanicInformation struct {
Discriminator string `json:"type"`
HyperV *GuestPanicInformationHyperV `json:"hyper-v"`
S390 *GuestPanicInformationS390 `json:"s390"`
}It can possibly be even simpler with just embedding the real struct
type jsonGuestPanicInformation struct {
Discriminator string
GuestPanicInformation
}Quoted 19 lines
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if (s.HyperV != nil && s.S390 != nil) || (s.HyperV == nil && s.S390 == nil) { // client hasn't filled in the struct properly return nil, errors.New("...") }tmp := jsonGuestPanicInformation{} if s.HyperV != nil { tmp.Discriminator = "hyper-v" tmp.HyperV = s.HyperV } else if s.S390 != nil { tmp.Discriminator = "s390" tmp.S390 = s.S390 } return json.Marshal(tmp)
Quoted 1 line
}
And...
var discriminator string
if s.HyperV != nil {
discriminator = "hyper-v"
} else if s.S390 != nil {
discriminator = "s390"
}
tmp := jsonGuestPanicInformation{ discriminator, s}
return json.Marshal(tmp)Quoted 13 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": if tmp.HyperV == nil { return errors.New("...")
Quoted 8 lines
}
s.HyperV = tmp.HyperV
case "s390":
if tmp.S390 == nil {
return errors.New("...")
}
s.S390 = tmp.S390
}I'm not actually sure this works, because the first json.Unmarshal call won't know which branch to try unmarhsalling. So it might be unavoidable to parse twice. With the XML parser this wouldn't be a problem as it has separated the parse phase and then fills the struct after.
Quoted 6 lines
// if we hit none of the cases above, that means the // server has produced a variant we don't know about return nil}
Quoted 7 lines
This avoid parsing the JSON twice as well as having to define multiple dummy structs, which keeps the code shorter and more readable.
I've also thrown in some additional error checking for good measure, ensuring that we abort when the input is completely nonsensical from a semantical standpoint.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 157 lines · Daniel P. Berrangé
Quoted 82 lines · Andrea Bolognani
Quoted 30 lines · Daniel P. Berrangé
Quoted 20 lines · Andrea Bolognani
All this string manipulation looks sketchy. Is there some reason that I'm not seeing preventing you for doing something like the untested code below?
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if s.HyperV != nil { type union struct { Discriminator string `json:"type"` HyperV GuestPanicInformationHyperV `json:"hyper-v"` } tmp := union { Discriminator: "hyper-v", HyperV: s.HyperV, } return json.Marshal(tmp) } else if s.S390 != nil { type union struct { Discriminator string `json:"type"` S390 GuestPanicInformationHyperV `json:"s390"` }Quoted 8 lines
tmp := union { Discriminator: "s390", S390: s.S390, } return json.Marshal(tmp) } return nil, errors.New("...") }Quoted 5 lines
Using these dummy structs is the way I've approached the discriminated union issue in the libvirt Golang XML bindings and it works well. It is the bit I like the least, but it was the lesser of many evils, and on the plus side in the QEMU case it'll be auto-generated code.
It appears to be the standard way to approach the problem in Go. It sort of comes naturally given how the APIs for marshal/unmarshal have been defined.
Quoted 30 lines
Quoted 30 lines
Quoted 29 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { type Alias GuestPanicInformation peek := struct { Alias Driver string `json:"type"` }{} if err := json.Unmarshal(data, &peek); err != nil { return err } *s = GuestPanicInformation(peek.Alias) switch peek.Driver { case "hyper-v": s.HyperV = new(GuestPanicInformationHyperV) if err := json.Unmarshal(data, s.HyperV); err != nil { s.HyperV = nil return err } case "s390": s.S390 = new(GuestPanicInformationS390) if err := json.Unmarshal(data, s.S390); err != nil { s.S390 = nil return err } } // Unrecognizer drivers are silently ignored. return nilQuoted 8 lines
Quoted 4 lines
This looks pretty reasonable, but since you're only using "peek" to look at the discriminator you should be able to leave out the Alias type entirely and perform the initial Unmarshal operation while ignoring all other fields.
Once you've defined the dummy structs for the Marshall case though, you might as well use them for Unmarshall too, so you're not parsing the JSON twice.
You're right, that is undesirable. What about something like this?
Quoted 10 lines
type GuestPanicInformation struct { HyperV *GuestPanicInformationHyperV S390 *GuestPanicInformationS390 } type jsonGuestPanicInformation struct { Discriminator string `json:"type"` HyperV *GuestPanicInformationHyperV `json:"hyper-v"` S390 *GuestPanicInformationS390 `json:"s390"` }It can possibly be even simpler with just embedding the real struct
type jsonGuestPanicInformation struct { Discriminator string GuestPanicInformation }Quoted 19 lines
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if (s.HyperV != nil && s.S390 != nil) || (s.HyperV == nil && s.S390 == nil) { // client hasn't filled in the struct properly return nil, errors.New("...") }tmp := jsonGuestPanicInformation{} if s.HyperV != nil { tmp.Discriminator = "hyper-v" tmp.HyperV = s.HyperV } else if s.S390 != nil { tmp.Discriminator = "s390" tmp.S390 = s.S390 } return json.Marshal(tmp)Quoted 1 line
}
And...
var discriminator string if s.HyperV != nil { discriminator = "hyper-v" } else if s.S390 != nil { discriminator = "s390" } tmp := jsonGuestPanicInformation{ discriminator, s} return json.Marshal(tmp)Quoted 13 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": if tmp.HyperV == nil { return errors.New("...")Quoted 8 lines
} s.HyperV = tmp.HyperV case "s390": if tmp.S390 == nil { return errors.New("...") } s.S390 = tmp.S390 }
Quoted 5 lines
I'm not actually sure this works, because the first json.Unmarshal call won't know which branch to try unmarhsalling. So it might be unavoidable to parse twice. With the XML parser this wouldn't be a problem as it has separated the parse phase and then fills the struct after.
Right afer sending, I remember how this is supposed to be done. It involves use of 'json.RawMessage' eg examples at:
https://pkg.go.dev/encoding/json#example-RawMessage-Unmarshal
So it would look like:
type GuestPanicInformation struct {
HyperV *GuestPanicInformationHyperV
S390 *GuestPanicInformationS390
}
type jsonGuestPanicInformation struct {
Discriminator string `json:"type"`
Payload *json.RawMessage
} func (s GuestPanicInformation) MarshalJSON() ([]byte, error) {
var p *json.RawMesage
var err error
if s.HyperV != nil {
d = "hyper-v"
p, err = json.Marshal(s.HyperV)
} else if s.S390 != nil {
d = "s390"
p, err = json.Marshal(s.S390)
} else {
err = fmt.Errorf("No payload defined")
}
if err != nil {
return []byte{}, err
}
return json.Marshal(jsonGuestPanicInformation{d, p}), nil
} func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error {
tmp := jsonGuestPanicInformation{} err := json.Unmarshal(data, &tmp)
if err != nil {
return err
}
switch tmp.Discriminator {
case "hyper-v":
s.HyperV := GuestPanicInformationHyperV{}
err := json.Unmarshal(tmp.Payload, s.HyperV)
if err != nil {
return err
}
case "s390":
s.S390 := GuestPanicInformationS390{}
err := json.Unmarshal(tmp.Payload, s.S390)
if err != nil {
return err
}
}
return fmt.Errorf("Unknown type '%s'", tmp.Discriminator)}
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 23 lines · Daniel P. Berrangé
Quoted 22 lines · Daniel P. Berrangé
Quoted 20 lines · Andrea Bolognani
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": if tmp.HyperV == nil { return errors.New("...") } s.HyperV = tmp.HyperV case "s390": if tmp.S390 == nil { return errors.New("...") } s.S390 = tmp.S390 }
Quoted 5 lines
Quoted 5 lines
I'm not actually sure this works, because the first json.Unmarshal call won't know which branch to try unmarhsalling. So it might be unavoidable to parse twice. With the XML parser this wouldn't be a problem as it has separated the parse phase and then fills the struct after.
It does, because the struct it's filling with data (jsonGuestPanicInformation) covers all possible cases. We then pick only the part we care about and transfer it to the user-provided return location.
Quoted 60 lines
Right afer sending, I remember how this is supposed to be done. It involves use of 'json.RawMessage' eg examples at:
https://pkg.go.dev/encoding/json#example-RawMessage-Unmarshal
So it would look like:
type GuestPanicInformation struct { HyperV *GuestPanicInformationHyperV S390 *GuestPanicInformationS390 } type jsonGuestPanicInformation struct { Discriminator string `json:"type"` Payload *json.RawMessage }func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { var p *json.RawMesage var err error if s.HyperV != nil { d = "hyper-v" p, err = json.Marshal(s.HyperV) } else if s.S390 != nil { d = "s390" p, err = json.Marshal(s.S390) } else { err = fmt.Errorf("No payload defined") } if err != nil { return []byte{}, err } return json.Marshal(jsonGuestPanicInformation{d, p}), nil }func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": s.HyperV := GuestPanicInformationHyperV{} err := json.Unmarshal(tmp.Payload, s.HyperV) if err != nil { return err } case "s390": s.S390 := GuestPanicInformationS390{} err := json.Unmarshal(tmp.Payload, s.S390) if err != nil { return err } } return fmt.Errorf("Unknown type '%s'", tmp.Discriminator)
Quoted 1 line
}
I guess this version would work too, but I think it might still perform more computation than the one I suggested?
json.RawMessage is a type alias for []byte, so I think the first call to json.Unmarshal will have to go through the message to figure out its structure and the contents of the discriminator field, then tweak the original input so that only the part that hasn't been consumed yet is returned. The second call to json.Marshal will then parse that part, which means that the "inner" chunk of JSON ends up being processed twice. In the approach I suggested, you go over the entire JSON in one go.
Things might even out when you take into account allocating and copying data around though. I'm not particularly interested in splitting hair when it comes to the most efficient approach at this point in time :) Knowing that we're not going through the entire JSON twice is good enough.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
In reply to Daniel P. BerrangéQuoted 15 lines · Daniel P. Berrangé
Quoted 14 lines · Daniel P. Berrangé
Quoted 12 lines · Andrea Bolognani
You're right, that is undesirable. What about something like this?
type GuestPanicInformation struct { HyperV *GuestPanicInformationHyperV S390 *GuestPanicInformationS390 } type jsonGuestPanicInformation struct { Discriminator string `json:"type"` HyperV *GuestPanicInformationHyperV `json:"hyper-v"` S390 *GuestPanicInformationS390 `json:"s390"` }
Quoted 7 lines
Quoted 7 lines
It can possibly be even simpler with just embedding the real struct
type jsonGuestPanicInformation struct { Discriminator string GuestPanicInformation }
Similar to what I said in previous email (same thread) to Andrea, this would not work because the end result does not match with QAPI spec, where HyperV or S390 fields should be at the same level as 'type'.
If we embed either HyperV or S390, then it should work, like:
tmp := struct {
GuestPanicInformationHyperV
Discriminator string "type"
}{}But I intend to try the json.RawMessage too as with description it seems like we can avoid looking the whole json data twice.
Quoted 123 lines
Quoted 54 lines
Quoted 18 lines
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if (s.HyperV != nil && s.S390 != nil) || (s.HyperV == nil && s.S390 == nil) { // client hasn't filled in the struct properly return nil, errors.New("...") }tmp := jsonGuestPanicInformation{} if s.HyperV != nil { tmp.Discriminator = "hyper-v" tmp.HyperV = s.HyperV } else if s.S390 != nil { tmp.Discriminator = "s390" tmp.S390 = s.S390 } return json.Marshal(tmp)Quoted 1 line
}
And...
var discriminator string if s.HyperV != nil { discriminator = "hyper-v" } else if s.S390 != nil { discriminator = "s390" } tmp := jsonGuestPanicInformation{ discriminator, s} return json.Marshal(tmp)Quoted 13 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": if tmp.HyperV == nil { return errors.New("...")Quoted 8 lines
} s.HyperV = tmp.HyperV case "s390": if tmp.S390 == nil { return errors.New("...") } s.S390 = tmp.S390 }Quoted 5 lines
I'm not actually sure this works, because the first json.Unmarshal call won't know which branch to try unmarhsalling. So it might be unavoidable to parse twice. With the XML parser this wouldn't be a problem as it has separated the parse phase and then fills the struct after.
Right afer sending, I remember how this is supposed to be done. It involves use of 'json.RawMessage' eg examples at:
https://pkg.go.dev/encoding/json#example-RawMessage-Unmarshal
So it would look like:
type GuestPanicInformation struct { HyperV *GuestPanicInformationHyperV S390 *GuestPanicInformationS390 } type jsonGuestPanicInformation struct { Discriminator string `json:"type"` Payload *json.RawMessage }func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { var p *json.RawMesage var err error if s.HyperV != nil { d = "hyper-v" p, err = json.Marshal(s.HyperV) } else if s.S390 != nil { d = "s390" p, err = json.Marshal(s.S390) } else { err = fmt.Errorf("No payload defined") } if err != nil { return []byte{}, err } return json.Marshal(jsonGuestPanicInformation{d, p}), nil }func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": s.HyperV := GuestPanicInformationHyperV{} err := json.Unmarshal(tmp.Payload, s.HyperV) if err != nil { return err } case "s390": s.S390 := GuestPanicInformationS390{} err := json.Unmarshal(tmp.Payload, s.S390) if err != nil { return err } } return fmt.Errorf("Unknown type '%s'", tmp.Discriminator)
Quoted 8 lines
}
With regards,
DanielSignature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Cheers,
Victor
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Hi,
Quoted 31 lines · Victor Toso
Quoted 15 lines · Daniel P. Berrangé
Quoted 14 lines · Daniel P. Berrangé
Quoted 12 lines · Andrea Bolognani
You're right, that is undesirable. What about something like this?
type GuestPanicInformation struct { HyperV *GuestPanicInformationHyperV S390 *GuestPanicInformationS390 } type jsonGuestPanicInformation struct { Discriminator string `json:"type"` HyperV *GuestPanicInformationHyperV `json:"hyper-v"` S390 *GuestPanicInformationS390 `json:"s390"` }Quoted 7 lines
Quoted 7 lines
It can possibly be even simpler with just embedding the real struct
type jsonGuestPanicInformation struct { Discriminator string GuestPanicInformation }Similar to what I said in previous email (same thread) to Andrea, this would not work because the end result does not match with QAPI spec, where HyperV or S390 fields should be at the same level as 'type'.
If we embed either HyperV or S390, then it should work, like:
Quoted 7 lines
tmp := struct { GuestPanicInformationHyperV Discriminator string "type" }{}But I intend to try the json.RawMessage too as with description it seems like we can avoid looking the whole json data twice.
For the same reason, I could not use json.RawMessage, sadly.
As Andrea pointed out before, json.RawMessage is just an alias for []byte. We need a field for that (so, it can't be embed) and the contents of the JSON need to match that field's name.
I've removed the string manipulation and used a few anonymous structs instead. I'm attaching a main.go program that tests this behavior together with input checks that Andrea suggested. This is more or less how the generated code will look like in the next iteration but in case one want's to find a nicer/shorter solution, I'm all ears :)
Cheers,
Victor
package main
import (
"encoding/json" "errors" "fmt" "strings"
)
type QCryptoBlockOptionsQCow struct {
KeySecret *string json:"key-secret,omitempty"
}
type QCryptoBlockOptionsLUKS struct {
KeySecret *string json:"key-secret,omitempty"
}
type QCryptoBlockOpenOptions struct {
// Variants fields
Qcow *QCryptoBlockOptionsQCow json:"-"
Luks *QCryptoBlockOptionsLUKS json:"-"
}
func (s QCryptoBlockOpenOptions) MarshalJSON() ([]byte, error) {
var bytes []byte
var err error
if s.Qcow != nil {
tmp := struct {
QCryptoBlockOptionsQCow
Discriminator string `json:"format"`
}{
QCryptoBlockOptionsQCow: *s.Qcow,
Discriminator: "qcow",
}
if bytes, err = json.Marshal(tmp); err != nil {
return nil, err
}
}
if s.Luks != nil {
if len(bytes) != 0 {
return nil, errors.New(`multiple fields set for QCryptoBlockOpenOptions`)
}
tmp := struct {
QCryptoBlockOptionsLUKS
Discriminator string `json:"format"`
}{
QCryptoBlockOptionsLUKS: *s.Luks,
Discriminator: "luks",
}
if bytes, err = json.Marshal(tmp); err != nil {
return nil, err
}
}
if len(bytes) == 0 {
return nil, errors.New(`null not supported for QCryptoBlockOpenOptions`)
}
return bytes, nil
}func (s *QCryptoBlockOpenOptions) UnmarshalJSON(data []byte) error {
tmp := struct {
Discriminator string `json:"format"`
}{}
if err := json.Unmarshal(data, &tmp); err != nil {
return err
}
switch tmp.Discriminator {
case "qcow":
s.Qcow = &QCryptoBlockOptionsQCow{}
if err := json.Unmarshal(data, s.Qcow); err != nil {
s.Qcow = nil
return err
}
case "luks":
s.Luks = &QCryptoBlockOptionsLUKS{}
if err := json.Unmarshal(data, s.Luks); err != nil {
s.Luks = nil
return err
}
}
return nil
}func main() {
jsonLuks := `{"key-secret":"my luks secret is here","format":"luks"}`
jsonQcow := `{"key-secret":"my qcow secret is here","format":"qcow"}`
r := QCryptoBlockOpenOptions{}
if err := json.Unmarshal([]byte(jsonLuks), &r); err != nil {
panic(err)
} else if r.Luks == nil || r.Qcow != nil {
panic(fmt.Sprintf("Wrong: %v", r))
} else if b, err := json.Marshal(&r); err != nil {
panic(err)
} else if string(b) != jsonLuks {
panic(string(b))
}
r = QCryptoBlockOpenOptions{}
if err := json.Unmarshal([]byte(jsonQcow), &r); err != nil {
panic(err)
} else if r.Luks != nil || r.Qcow == nil {
panic(fmt.Sprintf("Wrong: %v", r))
} else if b, err := json.Marshal(&r); err != nil {
panic(err)
} else if string(b) != jsonQcow {
panic(string(b))
}
r = QCryptoBlockOpenOptions{}
if _, err := json.Marshal(&r); err == nil {
panic("No fields set should be an error")
} else if !strings.Contains(err.Error(), "null not supported") {
panic(err)
}
qcowSecret := "my-qcow-secret-is-here"
luksSecret := "my-luks-secret-is-here"
r = QCryptoBlockOpenOptions{
Qcow: &QCryptoBlockOptionsQCow{
KeySecret: &qcowSecret,
},
Luks: &QCryptoBlockOptionsLUKS{
KeySecret: &luksSecret,
},
}
if _, err := json.Marshal(&r); err == nil {
panic("multiple fields set should be an error")
} else if !strings.Contains(err.Error(), "multiple fields set for") {
panic(err)
}
}Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Quoted 23 lines · Victor Toso
Quoted 23 lines
Quoted 22 lines · Daniel P. Berrangé
Quoted 21 lines · Andrea Bolognani
Quoted 12 lines
You're right, that is undesirable. What about something like this?
type GuestPanicInformation struct { HyperV *GuestPanicInformationHyperV S390 *GuestPanicInformationS390 } type jsonGuestPanicInformation struct { Discriminator string `json:"type"` HyperV *GuestPanicInformationHyperV `json:"hyper-v"` S390 *GuestPanicInformationS390 `json:"s390"` }It can possibly be even simpler with just embedding the real struct
type jsonGuestPanicInformation struct { Discriminator string GuestPanicInformation }
Quoted 4 lines
Quoted 4 lines
Similar to what I said in previous email (same thread) to Andrea, this would not work because the end result does not match with QAPI spec, where HyperV or S390 fields should be at the same level as 'type'.
Yeah, you're absolutely correct, I was misreading the schema and suggested an implementation that couldn't possibly work.
Quoted 7 lines
Quoted 6 lines
If we embed either HyperV or S390, then it should work, like:
tmp := struct { GuestPanicInformationHyperV Discriminator string "type" }{}
Quoted 5 lines
For the same reason, I could not use json.RawMessage, sadly.
As Andrea pointed out before, json.RawMessage is just an alias for []byte. We need a field for that (so, it can't be embed) and the contents of the JSON need to match that field's name.
Right. It would work if unions looked like
{ "type": "...", "data": { ... }}
with the "data" part having different fields based on the value of "type", but not for the flat JSON dictionaries that are used for QAPI unions.
Quoted 11 lines
func (s QCryptoBlockOpenOptions) MarshalJSON() ([]byte, error) {
var bytes []byte
var err error
if s.Qcow != nil {
tmp := struct {
QCryptoBlockOptionsQCow
Discriminator string `json:"format"`
}{
QCryptoBlockOptionsQCow: *s.Qcow,
Discriminator: "qcow",
}Quoted 8 lines
if bytes, err = json.Marshal(tmp); err != nil {
return nil, err
}
}
if s.Luks != nil {
if len(bytes) != 0 {
return nil, errors.New(`multiple fields set for QCryptoBlockOpenOptions`)
}Why are you checking this here? I would just have a check like
if s.Qcow != nil && s.Luks != nil {
return nil, errors.New(`multiple fields set for QCryptoBlockOpenOptions`)
}at the top of the function, so that you can abort early and cheaply if the user has provided invalid input instead of having to wait after one of the fields has already been serialized.
Quoted 6 lines
tmp := struct {
QCryptoBlockOptionsLUKS
Discriminator string `json:"format"`
}{
QCryptoBlockOptionsLUKS: *s.Luks,
Discriminator: "luks",Quoted 8 lines
}
if bytes, err = json.Marshal(tmp); err != nil {
return nil, err
}
}
if len(bytes) == 0 {
return nil, errors.New(`null not supported for QCryptoBlockOpenOptions`)
}Similarly, this should be checked early. So the complete check could be turned into
if (s.Qcow != nil && s.Luks != nil) || (s.Qcow == nil && s.Luks == nil) {
return nil, errors.New("must set exactly one field")
}You should have enough information to produce such a check when generating the function, right? If making sure all possible combinations are covered up ends up being too complicated, something like
var setFields int
if s.Field1 != nil {
setFields += 1
}
if s.Field2 != nil {
setFields += 1
}
// ...
if setFields != 1 {
return nil, errors.New("must set exactly one field")
}would also work.
Quoted 15 lines
func (s *QCryptoBlockOpenOptions) UnmarshalJSON(data []byte) error {
tmp := struct {
Discriminator string `json:"format"`
}{}
if err := json.Unmarshal(data, &tmp); err != nil {
return err
}
switch tmp.Discriminator {
case "qcow":
s.Qcow = &QCryptoBlockOptionsQCow{}
if err := json.Unmarshal(data, s.Qcow); err != nil {
s.Qcow = nil
return err
}
case "luks":Quoted 8 lines
s.Luks = &QCryptoBlockOptionsLUKS{}
if err := json.Unmarshal(data, s.Luks); err != nil {
s.Luks = nil
return err
}
}
return nil
}Alternatively, you could define a struct that covers all possible fields and deserialize everything in one go, then copy the various parts over to the appropriate struct:
func (s *QCryptoBlockOpenOptions) UnmarshalJSON(data []byte) error {
tmp := struct {
Discriminator string `json:"format"`
KeySecret *string `json:"key-secret,omitempty"`
}{}
if err := json.Unmarshal(data, &tmp); err != nil {
return err
}
switch tmp.Discriminator {
case "qcow":
s.Qcow = &QCryptoBlockOptionsQCow{}
s.Qcow.KeySecret = tmp.KeySecret
case "luks":
s.Luks = &QCryptoBlockOptionsLUKS{}
s.Luks.KeySecret = tmp.KeySecret
}
return nil
}Honestly I'm unclear on which option is better. Parsing the JSON twice, as you're doing in your version, could be problematic when the document is large; on the other hand, my approach will probably result in more copies of the same data being created.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
Hi,
Quoted 22 lines · Andrea Bolognani
Quoted 11 lines
func (s QCryptoBlockOpenOptions) MarshalJSON() ([]byte, error) { var bytes []byte var err error if s.Qcow != nil { tmp := struct { QCryptoBlockOptionsQCow Discriminator string `json:"format"` }{ QCryptoBlockOptionsQCow: *s.Qcow, Discriminator: "qcow", }Quoted 8 lines
if bytes, err = json.Marshal(tmp); err != nil { return nil, err } } if s.Luks != nil { if len(bytes) != 0 { return nil, errors.New(`multiple fields set for QCryptoBlockOpenOptions`) }Why are you checking this here? I would just have a check like
Quoted 8 lines
if s.Qcow != nil && s.Luks != nil { return nil, errors.New(`multiple fields set for QCryptoBlockOpenOptions`) }at the top of the function, so that you can abort early and cheaply if the user has provided invalid input instead of having to wait after one of the fields has already been serialized.
In general, I too prefer to return early! So I agree with you that it is nicer. At the same time, I was thinking a bit from the code generator point of view and the overall expected diffs when generating new code. More below.
Quoted 15 lines
Quoted 6 lines
tmp := struct { QCryptoBlockOptionsLUKS Discriminator string `json:"format"` }{ QCryptoBlockOptionsLUKS: *s.Luks, Discriminator: "luks",Quoted 8 lines
} if bytes, err = json.Marshal(tmp); err != nil { return nil, err } } if len(bytes) == 0 { return nil, errors.New(`null not supported for QCryptoBlockOpenOptions`) }
Quoted 6 lines
Similarly, this should be checked early. So the complete check could be turned into
if (s.Qcow != nil && s.Luks != nil) || (s.Qcow == nil && s.Luks == nil) { return nil, errors.New("must set exactly one field") }
Main problem with this approach is that there is some big unions like BlockdevOptions, this means that we would have to repeat all fields by the number fields times (40+ in BlockdevOptions' case).
Quoted 2 lines
You should have enough information to produce such a check when generating the function, right?
We do!
Quoted 16 lines
If making sure all possible combinations are covered up ends up being too complicated, something
likevar setFields int if s.Field1 != nil { setFields += 1 } if s.Field2 != nil { setFields += 1 } // ... if setFields != 1 { return nil, errors.New("must set exactly one field") }
Quoted 1 line
would also work.
I started with this approach actually but then I realized that we can just keep the if checks and instead of counter, do check the variable bytes []byte. So, do you think that the counter is actually preferred instead of inner check? I don't mind changing it.
Quoted 28 lines
Quoted 15 lines
func (s *QCryptoBlockOpenOptions) UnmarshalJSON(data []byte) error { tmp := struct { Discriminator string `json:"format"` }{} if err := json.Unmarshal(data, &tmp); err != nil { return err } switch tmp.Discriminator { case "qcow": s.Qcow = &QCryptoBlockOptionsQCow{} if err := json.Unmarshal(data, s.Qcow); err != nil { s.Qcow = nil return err } case "luks":Quoted 8 lines
s.Luks = &QCryptoBlockOptionsLUKS{} if err := json.Unmarshal(data, s.Luks); err != nil { s.Luks = nil return err } } return nil }Alternatively, you could define a struct that covers all possible fields and deserialize everything in one go, then copy the various parts over to the appropriate struct:
Quoted 15 lines
func (s *QCryptoBlockOpenOptions) UnmarshalJSON(data []byte) error {
tmp := struct {
Discriminator string `json:"format"`
KeySecret *string `json:"key-secret,omitempty"`
}{}
if err := json.Unmarshal(data, &tmp); err != nil {
return err
}
switch tmp.Discriminator {
case "qcow":
s.Qcow = &QCryptoBlockOptionsQCow{}
s.Qcow.KeySecret = tmp.KeySecret
case "luks":
s.Luks = &QCryptoBlockOptionsLUKS{}
s.Luks.KeySecret = tmp.KeySecretI think this one adds a bit more complexity and I'm not convinced that the gains are worth.
For example, with types that differ over fields with the same name? We would need to move that inside the anonymous struct somehow;
For example,if Luks's KeySecret was KeyScretType we would have to use KeySecretLuks *KeySecretType. Still, both of them would likely be inside of "key-secret" json object (so, same key for two different fields, not really sure how that would work out)
Quoted 9 lines
} return nil}
Honestly I'm unclear on which option is better. Parsing the JSON twice, as you're doing in your version, could be problematic when the document is large; on the other hand, my approach will probably result in more copies of the same data being created.
It does sound nice to parse it once but if we really want to do that, we would need to work with Golang's JSON decoder [0]. IMHO, parsing twice with in-memory data might be okay for the moment while we are trying to keep things simple and later we could benchmark against a custom decoder [0] in the future.
[0] https://pkg.go.dev/encoding/json#Decoder
Cheers,
Victor
Re: [RFC PATCH v2 4/8] qapi: golang: Generate qapi's union types in Go
In reply to Andrea BolognaniHi,
Quoted 37 lines · Andrea Bolognani
Quoted 30 lines · Daniel P. Berrangé
Quoted 20 lines · Andrea Bolognani
All this string manipulation looks sketchy. Is there some reason that I'm not seeing preventing you for doing something like the untested code below?
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if s.HyperV != nil { type union struct { Discriminator string `json:"type"` HyperV GuestPanicInformationHyperV `json:"hyper-v"` } tmp := union { Discriminator: "hyper-v", HyperV: s.HyperV, } return json.Marshal(tmp) } else if s.S390 != nil { type union struct { Discriminator string `json:"type"` S390 GuestPanicInformationHyperV `json:"s390"` }Quoted 8 lines
tmp := union { Discriminator: "s390", S390: s.S390, } return json.Marshal(tmp) } return nil, errors.New("...") }Quoted 5 lines
Using these dummy structs is the way I've approached the discriminated union issue in the libvirt Golang XML bindings and it works well. It is the bit I like the least, but it was the lesser of many evils, and on the plus side in the QEMU case it'll be auto-generated code.
Quoted 3 lines
It appears to be the standard way to approach the problem in Go. It sort of comes naturally given how the APIs for marshal/unmarshal have been defined.
Yep, string manipulation was bad choice. Some sort of anonymous struct is a better fit. So I'll be changing this ...
Quoted 41 lines
Quoted 30 lines
Quoted 30 lines
Quoted 29 lines
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { type Alias GuestPanicInformation peek := struct { Alias Driver string `json:"type"` }{} if err := json.Unmarshal(data, &peek); err != nil { return err } *s = GuestPanicInformation(peek.Alias) switch peek.Driver { case "hyper-v": s.HyperV = new(GuestPanicInformationHyperV) if err := json.Unmarshal(data, s.HyperV); err != nil { s.HyperV = nil return err } case "s390": s.S390 = new(GuestPanicInformationS390) if err := json.Unmarshal(data, s.S390); err != nil { s.S390 = nil return err } } // Unrecognizer drivers are silently ignored. return nilQuoted 8 lines
Quoted 4 lines
This looks pretty reasonable, but since you're only using "peek" to look at the discriminator you should be able to leave out the Alias type entirely and perform the initial Unmarshal operation while ignoring all other fields.
Once you've defined the dummy structs for the Marshall case though, you might as well use them for Unmarshall too, so you're not parsing the JSON twice.
You're right, that is undesirable. What about something like this?
Quoted 10 lines
type GuestPanicInformation struct {
HyperV *GuestPanicInformationHyperV
S390 *GuestPanicInformationS390
}
type jsonGuestPanicInformation struct {
Discriminator string `json:"type"`
HyperV *GuestPanicInformationHyperV `json:"hyper-v"`
S390 *GuestPanicInformationS390 `json:"s390"`
}I didn't test this so I could be wrong but, I think this should not work in case you want to remove the string manipulation.
The marshalling of either HyperV or S390 fields would return a JSON Object but QAPI spec expects the fields at the same level as the discriminator's type [0]. So, with this you still need some string manipulation to remove the extra {}, like I did poorly without any comment 0:-)
[0] https://gitlab.com/qemu-project/qemu/-/blob/master/docs/devel/qapi-code-gen.rst?plain=1#L358
Quoted 44 lines
func (s GuestPanicInformation) MarshalJSON() ([]byte, error) { if (s.HyperV != nil && s.S390 != nil) || (s.HyperV == nil && s.S390 == nil) { // client hasn't filled in the struct properly return nil, errors.New("...") }tmp := jsonGuestPanicInformation{} if s.HyperV != nil { tmp.Discriminator = "hyper-v" tmp.HyperV = s.HyperV } else if s.S390 != nil { tmp.Discriminator = "s390" tmp.S390 = s.S390 } return json.Marshal(tmp)}
func (s *GuestPanicInformation) UnmarshalJSON(data []byte) error { tmp := jsonGuestPanicInformation{}err := json.Unmarshal(data, &tmp) if err != nil { return err } switch tmp.Discriminator { case "hyper-v": if tmp.HyperV == nil { return errors.New("...") } s.HyperV = tmp.HyperV case "s390": if tmp.S390 == nil { return errors.New("...") } s.S390 = tmp.S390 } // if we hit none of the cases above, that means the // server has produced a variant we don't know about return nil
Quoted 5 lines
}
This avoid parsing the JSON twice as well as having to define multiple dummy structs, which keeps the code shorter and more readable.
If not too verbose, I'd still use anonymous structs whenever possible. They are created for a given scope, they are basically self documented in that given context and don't polluted package's namespace.
Quoted 3 lines
I've also thrown in some additional error checking for good measure, ensuring that we abort when the input is completely nonsensical from a semantical standpoint.
Thanks!
Victor
RFC PATCH v2 5/8 qapi: golang: Generate qapi's event types in Go
In reply to Victor TosoThis patch handles QAPI event types and generates data structures in Go that handles it.
We also define a Event interface and two helper functions MarshalEvent and UnmarshalEvent.
At the moment of this writing, this patch generates 51 structures (50 events)
Example:
qapi:
| { 'event': 'MEMORY_DEVICE_SIZE_CHANGE',
| 'data': { '*id': 'str', 'size': 'size', 'qom-path' : 'str'} }
go:
| type MemoryDeviceSizeChangeEvent struct {
| EventTimestamp Timestamp `json:"-"`
| Id *string `json:"id,omitempty"`
| Size uint64 `json:"size"`
| QomPath string `json:"qom-path"`
| }usage:
| input := `{"event":"MEMORY_DEVICE_SIZE_CHANGE",` +
| `"timestamp":{"seconds":1588168529,"microseconds":201316},` +
| `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}`
| e, err := UnmarshalEvent([]byte(input)
| if err != nil {
| panic(err)
| }
| if e.GetName() == `MEMORY_DEVICE_SIZE_CHANGE` {
| m := e.(*MemoryDeviceSizeChangeEvent)
| // m.QomPath == "/machine/unattached/device[2]"
| }--- scripts/qapi/golang.py | 120 ++++++++++++++++++++++++++++++++++++++++- 1 file changed, 118 insertions(+), 2 deletions(-)
scripts/qapi/golang.py +118 -2
@@ -28,7 +28,66 @@ ) from .source import QAPISourceInfo +# Only variable is @unm_cases to handle +# all events's names and associated types. +TEMPLATE_EVENT = ''' +type Timestamp struct {{ + Seconds int64 `json:"seconds"` + Microseconds int64 `json:"microseconds"` +}} + +type Event interface {{ + GetName() string + GetTimestamp() Timestamp +}} +func MarshalEvent(e Event) ([]byte, error) {{ + baseStruct := struct {{ + Name string `json:"event"` + EventTimestamp Timestamp `json:"timestamp"` + }}{{ + Name: e.GetName(), + EventTimestamp: e.GetTimestamp(), + }} + base, err := json.Marshal(baseStruct) + if err != nil {{ + return []byte{{}}, err + }} + + dataStruct := struct {{ + Payload Event `json:"data"` + }}{{ + Payload: e, + }} + data, err := json.Marshal(dataStruct) + if err != nil {{ + return []byte{{}}, err + }} + + if len(data) == len(`{{"data":{{}}}}`) {{ + return base, nil + }} + + // Combines Event's base and data in a single JSON object + result := fmt.Sprintf("%s,%s", base[:len(base)-1], data[1:]) + return []byte(result), nil +}} + +func UnmarshalEvent(data []byte) (Event, error) {{ + base := struct {{ + Name string `json:"event"` + EventTimestamp Timestamp `json:"timestamp"` + }}{{}} + if err := json.Unmarshal(data, &base); err != nil {{ + return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) + }} + + switch base.Name {{ + {unm_cases} + }} + return nil, errors.New("Failed to recognize event") +}} +''' TEMPLATE_HELPER = ''' // Alias for go version lower than 1.18 type Any = interface{}
@@ -53,10 +112,12 @@ class QAPISchemaGenGolangVisitor(QAPISchemaVisitor): def __init__(self, prefix: str): super().__init__() - self.target = {name: "" for name in ["alternate", "enum", "helper", "struct", + self.target = {name: "" for name in ["alternate", "enum", + "event", "helper", "struct", "union"]} self.objects_seen = {} self.schema = None + self.events = {} self.golang_package_name = "qapi" def visit_begin(self, schema):
@@ -71,6 +132,24 @@ def visit_begin(self, schema): def visit_end(self): self.schema = None + unm_cases = "" + for name in sorted(self.events): + case_type = self.events[name] + unm_cases += f''' + case "{name}": + event := struct {{ + Data {case_type} `json:"data"` + }}{{}} + + if err := json.Unmarshal(data, &event); err != nil {{ + return nil, errors.New(fmt.Sprintf("Failed to unmarshal: %s", string(data))) + }} + event.Data.EventTimestamp = base.EventTimestamp + return &event.Data, nil +''' + self.target["event"] += TEMPLATE_EVENT.format(unm_cases=unm_cases) + + def visit_object_type(self: QAPISchemaGenGolangVisitor, name: str, info: Optional[QAPISourceInfo],
@@ -232,7 +311,37 @@ def visit_command(self, pass def visit_event(self, name, info, ifcond, features, arg_type, boxed): - pass + assert name == info.defn_name + type_name = qapi_to_go_type_name(name, info.defn_meta) + self.events[name] = type_name + + self_contained = True + if arg_type and arg_type.name.startswith("q_obj"): + self_contained = False + + content = "" + if self_contained: + content = generate_struct_type(type_name, '''EventTimestamp Timestamp `json:"-"`''') + else: + assert isinstance(arg_type, QAPISchemaObjectType) + content = qapi_to_golang_struct(name, + arg_type.info, + arg_type.ifcond, + arg_type.features, + arg_type.base, + arg_type.members, + arg_type.variants) + + methods = f''' +func (s *{type_name}) GetName() string {{ + return "{name}" +}} + +func (s *{type_name}) GetTimestamp() Timestamp {{ + return s.EventTimestamp +}} +''' + self.target["event"] += content + methods def write(self, output_dir: str) -> None: for module_name, content in self.target.items():
@@ -274,6 +383,8 @@ def qapi_to_golang_struct(name: str, type_name = qapi_to_go_type_name(name, info.defn_meta) fields = "" + if info.defn_meta == "event": + fields += '''\tEventTimestamp Timestamp `json:"-"`\n''' if base: base_fields = ""
@@ -457,4 +568,9 @@ def qapi_to_go_type_name(name: str, meta: str) -> str: name = name.title() name += ''.join(word.title() for word in words[1:]) + + if meta in ["event"]: + name = name[:-3] if name.endswith("Arg") else name + name += meta.title().replace(" ", "") + return name
Signature · 1 line
2.36.1
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
Quoted 24 lines · Victor Toso
This patch handles QAPI event types and generates data structures in Go that handles it.
We also define a Event interface and two helper functions MarshalEvent and UnmarshalEvent.
At the moment of this writing, this patch generates 51 structures (50 events)
Example:
qapi:
| { 'event': 'MEMORY_DEVICE_SIZE_CHANGE',
| 'data': { '*id': 'str', 'size': 'size', 'qom-path' : 'str'} }go:
| type MemoryDeviceSizeChangeEvent struct { | EventTimestamp Timestamp `json:"-"` | Id *string `json:"id,omitempty"` | Size uint64 `json:"size"` | QomPath string `json:"qom-path"` | }usage:
Quoted 11 lines
| input := `{"event":"MEMORY_DEVICE_SIZE_CHANGE",` +
| `"timestamp":{"seconds":1588168529,"microseconds":201316},` +
| `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}`
| e, err := UnmarshalEvent([]byte(input)
| if err != nil {
| panic(err)
| }
| if e.GetName() == `MEMORY_DEVICE_SIZE_CHANGE` {
| m := e.(*MemoryDeviceSizeChangeEvent)
| // m.QomPath == "/machine/unattached/device[2]"
| }Generated code:
Quoted 2 lines
type AcpiDeviceOstEvent struct {
EventTimestamp Timestampjson:"-"
Any reason for naming this EventTimestamp as opposed to just Timestamp?
Quoted 6 lines
Info ACPIOSTInfo
json:"info"
}func (s *AcpiDeviceOstEvent)
GetName()string {
return "ACPI_DEVICE_OST"
}
Getters in Go don't usually start with "Get", so this could be just Name(). And pointer receivers are usually only recommended when the underlying object needs to be modified, which is not the case here.
Quoted 3 lines
func (s *AcpiDeviceOstEvent)
GetTimestamp()Timestamp {
return s.EventTimestamp
}
Does this even need a getter? The struct member is public, and Go code seems to favor direct member access.
Quoted 33 lines
type Timestamp struct {
Seconds int64json:"seconds"
Microseconds int64json:"microseconds"
}type Event interface {
GetName()string
GetTimestamp()Timestamp
}func MarshalEvent(e Event) ([]byte, error) { baseStruct := struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` }{ Name: e.GetName(), EventTimestamp: e.GetTimestamp(), } base, err := json.Marshal(baseStruct) if err != nil { return []byte{}, err } dataStruct := struct { Payload Event `json:"data"` }{ Payload: e, } data, err := json.Marshal(dataStruct) if err != nil { return []byte{}, err }
Quoted 8 lines
if len(data) == len(`{"data":{}}`) {
return base, nil
}
// Combines Event's base and data in a single JSON object
result := fmt.Sprintf("%s,%s", base[:len(base)-1], data[1:])
return []byte(result), nil
}I have the same concerns about the string manipulation going on here as I had for unions. Here's an alternative implementation, and this time I've even tested it :)
func (s AcpiDeviceOstEvent) MarshalJSON() ([]byte, error) {
type eventData struct {
Info ACPIOSTInfo `json:"info"`
}
type event struct {
Name string `json:"event"`
Timestamp Timestamp `json:"timestamp"`
Data eventData `json:"data"`
} tmp := event{
Name: "ACPI_DEVICE_OST",
Timestamp: s.EventTimestamp,
Data: eventData{
Info: s.Info,
},
}
return json.Marshal(tmp)}
Quoted 18 lines
func UnmarshalEvent(data []byte) (Event, error) {
base := struct {
Name string `json:"event"`
EventTimestamp Timestamp `json:"timestamp"`
}{}
if err := json.Unmarshal(data, &base); err != nil {
return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data)))
}
switch base.Name {
case "ACPI_DEVICE_OST":
event := struct {
Data AcpiDeviceOstEvent `json:"data"`
}{}
if err := json.Unmarshal(data, &event); err != nil {
return nil, errors.New(fmt.Sprintf("Failed to unmarshal: %s", string(data)))Quoted 8 lines
}
event.Data.EventTimestamp = base.EventTimestamp
return &event.Data, nil
// more cases here
}
return nil, errors.New("Failed to recognize event")
}While I understand the need to have some way to figure out exactly what type of event we're looking at before being able to unmarshal the JSON data, I don't like how we force users to go through a non-standard API for it.
Here's how I think we should do it:
func GetEventType(data []byte) (Event, error) {
type event struct {
Name string `json:"event"`
} tmp := event{}
if err := json.Unmarshal(data, &tmp); err != nil {
return nil, errors.New(fmt.Sprintf("Failed to decode event:
%s", string(data)))
}
switch tmp.Name {
case "ACPI_DEVICE_OST":
return &AcpiDeviceOstEvent{}, nil
// more cases here
}
return nil, errors.New("Failed to recognize event")}
func (s *AcpiDeviceOstEvent) UnmarshalJSON(data []byte) error {
type eventData struct {
Info ACPIOSTInfo `json:"info"`
}
type event struct {
Name string `json:"event"`
EventTimestamp Timestamp `json:"timestamp"`
Data eventData `json:"data"`
} tmp := event{}
err := json.Unmarshal(data, &tmp)
if err != nil {
return err
}
if tmp.Name != "ACPI_DEVICE_OST" {
return errors.New("name")
}
s.EventTimestamp = tmp.EventTimestamp
s.Info = tmp.Data.Info
return nil}
This way, client code can look like
in := {"event":"ACPI_DEVICE_OST","timestamp":{"seconds":1265044230,"microseconds":450486},"data":{"info":{"device":"d1","slot":"0","slot-type":"DIMM","source":1,"status":0}}}
e, err := qapi.GetEventType([]byte(in))
if err != nil {
panic(err)
}
// e is of type AcpiDeviceOstEvent
err = json.Unmarshal([]byte(in), &e)
if err != nil {
panic(err)
}where only the first part (figuring out the type of the event) needs custom APIs, and the remaining code looks just like your average JSON handling.
Speaking of client code, in the commit message you have
Quoted 11 lines
input :=
{"event":"MEMORY_DEVICE_SIZE_CHANGE",+`"timestamp":{"seconds":1588168529,"microseconds":201316},` + `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}`e, err := UnmarshalEvent([]byte(input)
if err != nil {
panic(err)
}
ife.GetName()==MEMORY_DEVICE_SIZE_CHANGE{
m := e.(*MemoryDeviceSizeChangeEvent) // m.QomPath == "/machine/unattached/device[2]"
}
I don't think we should encourage the use of string comparison for the purpose of going from a generic Event instance to a specific one:
a better way would be to use Go's type switch feature, such as
switch m := e.(type) {
case *MemoryDeviceSizeChangeEvent:
// m.QomPath == "/machine/unattached/device[2]"
}In fact, I don't really see a reason to provide the Name() getter outside of possibly diagnostics, and given the potential of it being misused I would argue that it might be better not to have it.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
Quoted 38 lines · Andrea Bolognani
Quoted 35 lines · Victor Toso
This patch handles QAPI event types and generates data structures in Go that handles it.
We also define a Event interface and two helper functions MarshalEvent and UnmarshalEvent.
At the moment of this writing, this patch generates 51 structures (50 events)
Example:
qapi:
| { 'event': 'MEMORY_DEVICE_SIZE_CHANGE',
| 'data': { '*id': 'str', 'size': 'size', 'qom-path' : 'str'} }go:
| type MemoryDeviceSizeChangeEvent struct { | EventTimestamp Timestamp `json:"-"` | Id *string `json:"id,omitempty"` | Size uint64 `json:"size"` | QomPath string `json:"qom-path"` | }usage:
| input := `{"event":"MEMORY_DEVICE_SIZE_CHANGE",` + | `"timestamp":{"seconds":1588168529,"microseconds":201316},` + | `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}` | e, err := UnmarshalEvent([]byte(input) | if err != nil { | panic(err) | } | if e.GetName() == `MEMORY_DEVICE_SIZE_CHANGE` { | m := e.(*MemoryDeviceSizeChangeEvent) | // m.QomPath == "/machine/unattached/device[2]" | }
Quoted 6 lines
Quoted 3 lines
func (s *AcpiDeviceOstEvent)
GetTimestamp()Timestamp {
return s.EventTimestamp
}Does this even need a getter? The struct member is public, and Go code seems to favor direct member access.
It satisfies the 'Event' interface, so you can fetch timestamp without needing to know the specific event struct you're dealing with.
Quoted 107 lines
Quoted 31 lines
type Timestamp struct {
Seconds int64json:"seconds"
Microseconds int64json:"microseconds"
}type Event interface {
GetName()string
GetTimestamp()Timestamp
}func UnmarshalEvent(data []byte) (Event, error) { base := struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` }{} if err := json.Unmarshal(data, &base); err != nil { return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) } switch base.Name { case "ACPI_DEVICE_OST": event := struct { Data AcpiDeviceOstEvent `json:"data"` }{} if err := json.Unmarshal(data, &event); err != nil { return nil, errors.New(fmt.Sprintf("Failed to unmarshal: %s", string(data)))Quoted 8 lines
} event.Data.EventTimestamp = base.EventTimestamp return &event.Data, nil // more cases here } return nil, errors.New("Failed to recognize event") }While I understand the need to have some way to figure out exactly what type of event we're looking at before being able to unmarshal the JSON data, I don't like how we force users to go through a non-standard API for it.
Here's how I think we should do it:
func GetEventType(data []byte) (Event, error) { type event struct { Name string `json:"event"` }tmp := event{} if err := json.Unmarshal(data, &tmp); err != nil { return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) } switch tmp.Name { case "ACPI_DEVICE_OST": return &AcpiDeviceOstEvent{}, nil // more cases here } return nil, errors.New("Failed to recognize event")}
func (s *AcpiDeviceOstEvent) UnmarshalJSON(data []byte) error { type eventData struct { Info ACPIOSTInfo `json:"info"` } type event struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` Data eventData `json:"data"` }tmp := event{} err := json.Unmarshal(data, &tmp) if err != nil { return err } if tmp.Name != "ACPI_DEVICE_OST" { return errors.New("name") } s.EventTimestamp = tmp.EventTimestamp s.Info = tmp.Data.Info return nil}
This way, client code can look like
in :=
{"event":"ACPI_DEVICE_OST","timestamp":{"seconds":1265044230,"microseconds":450486},"data":{"info":{"device":"d1","slot":"0","slot-type":"DIMM","source":1,"status":0}}}e, err := qapi.GetEventType([]byte(in)) if err != nil { panic(err) } // e is of type AcpiDeviceOstEvent err = json.Unmarshal([]byte(in), &e) if err != nil { panic(err) }
Quoted 3 lines
where only the first part (figuring out the type of the event) needs custom APIs, and the remaining code looks just like your average JSON handling.
I don't think this kind of detail really needs to be exposed to clients. Also parsing the same json doc twice just feels wrong.
I think using the dummy union structs is the right approach, but I'd just call it 'EventFromJSON' rather than 'UnmarshalJSON' to make it clear this is different from a normal 'UnmarshalJSON' method.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
Quoted 6 lines · Daniel P. Berrangé
Quoted 5 lines · Andrea Bolognani
Quoted 3 lines · Victor Toso
func (s *AcpiDeviceOstEvent)
GetTimestamp()Timestamp {
return s.EventTimestamp
}
Quoted 6 lines
Quoted 2 lines
Does this even need a getter? The struct member is public, and Go code seems to favor direct member access.
It satisfies the 'Event' interface, so you can fetch timestamp without needing to know the specific event struct you're dealing with.
Yeah but we're generating structs for all possible events ourselves and we don't really expect external implementations of this interface so I don't see what having this getter buys us over the guarantee, that we have by construction, that all events will have a public member with a specific name that contains the timestamp.
Quoted 7 lines
I don't think this kind of detail really needs to be exposed to clients. Also parsing the same json doc twice just feels wrong.
I think using the dummy union structs is the right approach, but I'd just call it 'EventFromJSON' rather than 'UnmarshalJSON' to make it clear this is different from a normal 'UnmarshalJSON' method.
Fair point on avoiding parsing the JSON twice.
I still don't like the fact that we have to force clients to use a non-standard API, or that the API for events has to be different from that for unions. But Go won't let you add method implementations to an interface, so we're kinda stuck.
The only alternative I can think of would be to define Event as
type Event struct {
Shutdown *ShutdownEvent
Reset *ResetEvent
// and so on
}which wouldn't be too bad from client code, as you'd only have to change from
e, err := EventFromJSON(in)
switch v := e.(type) {
case ShutdownEvent:
// do something with v
case ResetEvent:
// do something with v
// and so on
}to
err := json.UnmarshalJSON(in, &e)
if e.Shutdown != nil {
// do something with e.Shutdown
} else if e.Reset != nil {
// do something with e.Reset
} // and so onbut that would mean each time we have to parse an event we'd end up allocating room for ~50 pointers and only actually using one, which is a massive waste.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
Quoted 14 lines · Andrea Bolognani
Quoted 6 lines · Daniel P. Berrangé
Quoted 5 lines · Andrea Bolognani
Quoted 3 lines · Victor Toso
func (s *AcpiDeviceOstEvent)
GetTimestamp()Timestamp {
return s.EventTimestamp
}Quoted 6 lines
Quoted 2 lines
Does this even need a getter? The struct member is public, and Go code seems to favor direct member access.
It satisfies the 'Event' interface, so you can fetch timestamp without needing to know the specific event struct you're dealing with.
Quoted 5 lines
Yeah but we're generating structs for all possible events ourselves and we don't really expect external implementations of this interface so I don't see what having this getter buys us over the guarantee, that we have by construction, that all events will have a public member with a specific name that contains the timestamp.
Code doesn't neccessarily want to deal with the per-event type structs. It is credible to want to work with the abstract 'Event' interface in places and being able to get the Timestamp from that, without figuring out what specific event struct to cast it to first.
Quoted 8 lines
Quoted 7 lines
I don't think this kind of detail really needs to be exposed to clients. Also parsing the same json doc twice just feels wrong.
I think using the dummy union structs is the right approach, but I'd just call it 'EventFromJSON' rather than 'UnmarshalJSON' to make it clear this is different from a normal 'UnmarshalJSON' method.
Quoted 6 lines
Fair point on avoiding parsing the JSON twice.
I still don't like the fact that we have to force clients to use a non-standard API, or that the API for events has to be different from that for unions. But Go won't let you add method implementations to an interface, so we're kinda stuck.
I think this specific case is out of scope for the "standard" JSON APIs, and somewhere you'd expect APIs to do their own thing a layer above, which is what we're doing here.
More importantly though what we're generating in terms of QAPI derived APIs should be thought of as only the first step. Ultimately I would not expect clients to be marshalling / unmarshalling these structs at all. So from that POV I think the question of "non-standard" APIs is largely irrelevant.
Instead we would be providing a "QMPClient" type with APIs for sending/ receiving instances of the 'Command'/'Response' interfaces, along with callback(s) that receives instances of the 'Event' interface. Any JSON marshalling is hidden behind the scenes as an internal imlp detail.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
Quoted 11 lines · Daniel P. Berrangé
Quoted 5 lines · Andrea Bolognani
Yeah but we're generating structs for all possible events ourselves and we don't really expect external implementations of this interface so I don't see what having this getter buys us over the guarantee, that we have by construction, that all events will have a public member with a specific name that contains the timestamp.
Code doesn't neccessarily want to deal with the per-event type structs. It is credible to want to work with the abstract 'Event' interface in places and being able to get the Timestamp from that, without figuring out what specific event struct to cast it to first.
Makes sense.
Quoted 15 lines
Quoted 4 lines
I still don't like the fact that we have to force clients to use a non-standard API, or that the API for events has to be different from that for unions. But Go won't let you add method implementations to an interface, so we're kinda stuck.
I think this specific case is out of scope for the "standard" JSON APIs, and somewhere you'd expect APIs to do their own thing a layer above, which is what we're doing here.
More importantly though what we're generating in terms of QAPI derived APIs should be thought of as only the first step. Ultimately I would not expect clients to be marshalling / unmarshalling these structs at all. So from that POV I think the question of "non-standard" APIs is largely irrelevant.
Quoted 4 lines
Instead we would be providing a "QMPClient" type with APIs for sending/ receiving instances of the 'Command'/'Response' interfaces, along with callback(s) that receives instances of the 'Event' interface. Any JSON marshalling is hidden behind the scenes as an internal imlp detail.
I will note that we want the marshalling/unmarshalling part and the wire transfer part to be somewhat loosely coupled, to allow for use cases that are not covered by the high-level client API, but overall I now agree with you that for most users this shouldn't ultimately matter :)
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
In reply to Daniel P. BerrangéHi,
Quoted 161 lines · Daniel P. Berrangé
Quoted 38 lines · Andrea Bolognani
Quoted 35 lines · Victor Toso
This patch handles QAPI event types and generates data structures in Go that handles it.
We also define a Event interface and two helper functions MarshalEvent and UnmarshalEvent.
At the moment of this writing, this patch generates 51 structures (50 events)
Example:
qapi:
| { 'event': 'MEMORY_DEVICE_SIZE_CHANGE',
| 'data': { '*id': 'str', 'size': 'size', 'qom-path' : 'str'} }go:
| type MemoryDeviceSizeChangeEvent struct { | EventTimestamp Timestamp `json:"-"` | Id *string `json:"id,omitempty"` | Size uint64 `json:"size"` | QomPath string `json:"qom-path"` | }usage:
| input := `{"event":"MEMORY_DEVICE_SIZE_CHANGE",` + | `"timestamp":{"seconds":1588168529,"microseconds":201316},` + | `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}` | e, err := UnmarshalEvent([]byte(input) | if err != nil { | panic(err) | } | if e.GetName() == `MEMORY_DEVICE_SIZE_CHANGE` { | m := e.(*MemoryDeviceSizeChangeEvent) | // m.QomPath == "/machine/unattached/device[2]" | }Quoted 6 lines
Quoted 3 lines
func (s *AcpiDeviceOstEvent)
GetTimestamp()Timestamp {
return s.EventTimestamp
}Does this even need a getter? The struct member is public, and Go code seems to favor direct member access.
It satisfies the 'Event' interface, so you can fetch timestamp without needing to know the specific event struct you're dealing with.
Quoted 107 lines
Quoted 31 lines
type Timestamp struct {
Seconds int64json:"seconds"
Microseconds int64json:"microseconds"
}type Event interface {
GetName()string
GetTimestamp()Timestamp
}func UnmarshalEvent(data []byte) (Event, error) { base := struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` }{} if err := json.Unmarshal(data, &base); err != nil { return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) } switch base.Name { case "ACPI_DEVICE_OST": event := struct { Data AcpiDeviceOstEvent `json:"data"` }{} if err := json.Unmarshal(data, &event); err != nil { return nil, errors.New(fmt.Sprintf("Failed to unmarshal: %s", string(data)))Quoted 8 lines
} event.Data.EventTimestamp = base.EventTimestamp return &event.Data, nil // more cases here } return nil, errors.New("Failed to recognize event") }While I understand the need to have some way to figure out exactly what type of event we're looking at before being able to unmarshal the JSON data, I don't like how we force users to go through a non-standard API for it.
Here's how I think we should do it:
func GetEventType(data []byte) (Event, error) { type event struct { Name string `json:"event"` }tmp := event{} if err := json.Unmarshal(data, &tmp); err != nil { return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) } switch tmp.Name { case "ACPI_DEVICE_OST": return &AcpiDeviceOstEvent{}, nil // more cases here } return nil, errors.New("Failed to recognize event")}
func (s *AcpiDeviceOstEvent) UnmarshalJSON(data []byte) error { type eventData struct { Info ACPIOSTInfo `json:"info"` } type event struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` Data eventData `json:"data"` }tmp := event{} err := json.Unmarshal(data, &tmp) if err != nil { return err } if tmp.Name != "ACPI_DEVICE_OST" { return errors.New("name") } s.EventTimestamp = tmp.EventTimestamp s.Info = tmp.Data.Info return nil}
This way, client code can look like
in :=
{"event":"ACPI_DEVICE_OST","timestamp":{"seconds":1265044230,"microseconds":450486},"data":{"info":{"device":"d1","slot":"0","slot-type":"DIMM","source":1,"status":0}}}e, err := qapi.GetEventType([]byte(in)) if err != nil { panic(err) } // e is of type AcpiDeviceOstEvent err = json.Unmarshal([]byte(in), &e) if err != nil { panic(err) }Quoted 3 lines
where only the first part (figuring out the type of the event) needs custom APIs, and the remaining code looks just like your average JSON handling.
Quoted 7 lines
I don't think this kind of detail really needs to be exposed to clients. Also parsing the same json doc twice just feels wrong.
I think using the dummy union structs is the right approach, but I'd just call it 'EventFromJSON' rather than 'UnmarshalJSON' to make it clear this is different from a normal 'UnmarshalJSON' method.
The current method signatures are:
func MarshalEvent(e Event) ([]byte, error)
func UnmarshalEvent(data []byte) (Event, error)the suggestion is to change it to
func EventToJSON(e Event) ([]byte, error)
func EventFromJSON(data []byte) (Event, error)or
func JSONFromEvent(e Event) ([]byte, error)
func EventFromJSON(data []byte) (Event, error)? :)
I'm not picky with names so, what you find is best is okay for me.
I'll be changing the function to avoid string manipulation in favor of some anonymous structs.
Thanks you all for the review, really appreciate it! Victor
Re: [RFC PATCH v2 5/8] qapi: golang: Generate qapi's event types in Go
In reply to Andrea BolognaniHi,
Quoted 37 lines · Andrea Bolognani
Quoted 24 lines · Victor Toso
This patch handles QAPI event types and generates data structures in Go that handles it.
We also define a Event interface and two helper functions MarshalEvent and UnmarshalEvent.
At the moment of this writing, this patch generates 51 structures (50 events)
Example:
qapi:
| { 'event': 'MEMORY_DEVICE_SIZE_CHANGE',
| 'data': { '*id': 'str', 'size': 'size', 'qom-path' : 'str'} }go:
| type MemoryDeviceSizeChangeEvent struct { | EventTimestamp Timestamp `json:"-"` | Id *string `json:"id,omitempty"` | Size uint64 `json:"size"` | QomPath string `json:"qom-path"` | }usage:
Quoted 11 lines
| input := `{"event":"MEMORY_DEVICE_SIZE_CHANGE",` + | `"timestamp":{"seconds":1588168529,"microseconds":201316},` + | `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}` | e, err := UnmarshalEvent([]byte(input) | if err != nil { | panic(err) | } | if e.GetName() == `MEMORY_DEVICE_SIZE_CHANGE` { | m := e.(*MemoryDeviceSizeChangeEvent) | // m.QomPath == "/machine/unattached/device[2]" | }
Quoted 7 lines
Generated code:
Quoted 2 lines
type AcpiDeviceOstEvent struct {
EventTimestamp Timestampjson:"-"Any reason for naming this EventTimestamp as opposed to just Timestamp?
Yes, perhaps one that can be workaround in the next iteration. IIRC, I've added the type prefix to avoid the possibility of name clash with generated fields (like Info below).
Quoted 2 lines
Quoted 2 lines
Info ACPIOSTInfo
json:"info"
}
This happened with Command's Id that clashed with an Id for one or several other commands so I've changed it to "CommandId" and thought it would be wise to do the same for non generated fields.
Quoted 7 lines
Quoted 3 lines
func (s *AcpiDeviceOstEvent)
GetName()string {
return "ACPI_DEVICE_OST"
}Getters in Go don't usually start with "Get", so this could be just
Name(). And pointer receivers are usually only recommended when the underlying object needs to be modified, which is not the case here.
Yeah, thanks. I'll change it.
Quoted 6 lines
Quoted 3 lines
func (s *AcpiDeviceOstEvent)
GetTimestamp()Timestamp {
return s.EventTimestamp
}Does this even need a getter? The struct member is public, and Go code seems to favor direct member access.
As Daniel pointed out, just for the sake of working with a Event interface.
Quoted 37 lines
Quoted 37 lines
type Timestamp struct {
Seconds int64json:"seconds"
Microseconds int64json:"microseconds"
}type Event interface {
GetName()string
GetTimestamp()Timestamp
}func MarshalEvent(e Event) ([]byte, error) { baseStruct := struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` }{ Name: e.GetName(), EventTimestamp: e.GetTimestamp(), } base, err := json.Marshal(baseStruct) if err != nil { return []byte{}, err } dataStruct := struct { Payload Event `json:"data"` }{ Payload: e, } data, err := json.Marshal(dataStruct) if err != nil { return []byte{}, err } if len(data) == len(`{"data":{}}`) { return base, nil }
Quoted 8 lines
Quoted 4 lines
// Combines Event's base and data in a single JSON object result := fmt.Sprintf("%s,%s", base[:len(base)-1], data[1:]) return []byte(result), nil }I have the same concerns about the string manipulation going on here as I had for unions. Here's an alternative implementation, and this time I've even tested it :)
Yup. I agree that string manipulation was bad decision. I'll change it here too.
Quoted 114 lines
func (s AcpiDeviceOstEvent) MarshalJSON() ([]byte, error) { type eventData struct { Info ACPIOSTInfo `json:"info"` } type event struct { Name string `json:"event"` Timestamp Timestamp `json:"timestamp"` Data eventData `json:"data"` }tmp := event{ Name: "ACPI_DEVICE_OST", Timestamp: s.EventTimestamp, Data: eventData{ Info: s.Info, }, } return json.Marshal(tmp)}
Quoted 18 lines
func UnmarshalEvent(data []byte) (Event, error) { base := struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` }{} if err := json.Unmarshal(data, &base); err != nil { return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) } switch base.Name { case "ACPI_DEVICE_OST": event := struct { Data AcpiDeviceOstEvent `json:"data"` }{} if err := json.Unmarshal(data, &event); err != nil { return nil, errors.New(fmt.Sprintf("Failed to unmarshal: %s", string(data)))Quoted 8 lines
} event.Data.EventTimestamp = base.EventTimestamp return &event.Data, nil // more cases here } return nil, errors.New("Failed to recognize event") }While I understand the need to have some way to figure out exactly what type of event we're looking at before being able to unmarshal the JSON data, I don't like how we force users to go through a non-standard API for it.
Here's how I think we should do it:
func GetEventType(data []byte) (Event, error) { type event struct { Name string `json:"event"` }tmp := event{} if err := json.Unmarshal(data, &tmp); err != nil { return nil, errors.New(fmt.Sprintf("Failed to decode event: %s", string(data))) } switch tmp.Name { case "ACPI_DEVICE_OST": return &AcpiDeviceOstEvent{}, nil // more cases here } return nil, errors.New("Failed to recognize event")}
func (s *AcpiDeviceOstEvent) UnmarshalJSON(data []byte) error { type eventData struct { Info ACPIOSTInfo `json:"info"` } type event struct { Name string `json:"event"` EventTimestamp Timestamp `json:"timestamp"` Data eventData `json:"data"` }tmp := event{} err := json.Unmarshal(data, &tmp) if err != nil { return err } if tmp.Name != "ACPI_DEVICE_OST" { return errors.New("name") } s.EventTimestamp = tmp.EventTimestamp s.Info = tmp.Data.Info return nil}
This way, client code can look like
in :=
{"event":"ACPI_DEVICE_OST","timestamp":{"seconds":1265044230,"microseconds":450486},"data":{"info":{"device":"d1","slot":"0","slot-type":"DIMM","source":1,"status":0}}}e, err := qapi.GetEventType([]byte(in)) if err != nil { panic(err) } // e is of type AcpiDeviceOstEvent err = json.Unmarshal([]byte(in), &e) if err != nil { panic(err) }
Quoted 3 lines
where only the first part (figuring out the type of the event) needs custom APIs, and the remaining code looks just like your average JSON handling.
I'll reply to this bit in the other email of this thread.
Quoted 14 lines
Speaking of client code, in the commit message you have
Quoted 11 lines
input :=
{"event":"MEMORY_DEVICE_SIZE_CHANGE",+`"timestamp":{"seconds":1588168529,"microseconds":201316},` + `"data":{"id":"vm0","size":1073741824,"qom-path":"/machine/unattached/device[2]"}}`e, err := UnmarshalEvent([]byte(input)
if err != nil {
panic(err)
}
ife.GetName()==MEMORY_DEVICE_SIZE_CHANGE{
m := e.(*MemoryDeviceSizeChangeEvent) // m.QomPath == "/machine/unattached/device[2]"
}
Quoted 8 lines
I don't think we should encourage the use of string comparison for the purpose of going from a generic Event instance to a specific one:
a better way would be to use Go's type switch feature, such asswitch m := e.(type) { case *MemoryDeviceSizeChangeEvent: // m.QomPath == "/machine/unattached/device[2]" }
Yes, this seems much better/align with Go code. Many thanks for this!
Quoted 6 lines
In fact, I don't really see a reason to provide the
Name()getter outside of possibly diagnostics, and given the potential of it being misused I would argue that it might be better not to have it.--
Andrea Bolognani / Red Hat / Virtualization
Cheers,
Victor
RFC PATCH v2 6/8 qapi: golang: Generate qapi's command types in Go
In reply to Victor TosoThis patch handles QAPI command types and generates data structures in Go that decodes from QMP JSON Object to Go data structure and vice versa.
Simlar to Event, this patch adds a Command interface and two helper functions MarshalCommand and UnmarshalCommand.
At the time of this writing, it generates 209 structures.
Example:
qapi:
| { 'command': 'set_password',
| 'boxed': true,
| 'data': 'SetPasswordOptions' }go:
| type SetPasswordCommand struct {
| SetPasswordOptions
| CommandId string `json:"-"`
| }usage:
| input := `{"execute":"set_password",` +
| `"arguments":{"protocol":"vnc","password":"secret"}}`
| c, err := UnmarshalCommand([]byte(input))
| if err != nil {
| panic(err)
| }
| if c.GetName() == `set_password` {
| m := c.(*SetPasswordCommand)
| // m.Password == "secret"
| }--- scripts/qapi/golang.py | 123 ++++++++++++++++++++++++++++++++++++++++- 1 file changed, 120 insertions(+), 3 deletions(-)
scripts/qapi/golang.py +120 -3
@@ -88,6 +88,63 @@ return nil, errors.New("Failed to recognize event") }} ''' + +# Only variable is @unm_cases to handle all command's names and associated types. +TEMPLATE_COMMAND = ''' +type Command interface {{ + GetId() string + GetName() string + GetReturnType() CommandReturn +}} + +func MarshalCommand(c Command) ([]byte, error) {{ + baseStruct := struct {{ + CommandId string `json:"id,omitempty"` + Name string `json:"execute"` + }}{{ + CommandId: c.GetId(), + Name: c.GetName(), + }} + base, err := json.Marshal(baseStruct) + if err != nil {{ + return []byte{{}}, err + }} + + argsStruct := struct {{ + Args Command `json:"arguments,omitempty"` + }}{{ + Args: c, + }} + args, err := json.Marshal(argsStruct) + if err != nil {{ + return []byte{{}}, err + }} + + if len(args) == len(`{{"arguments":{{}}}}`) {{ + return base, nil + }} + + // Combines Event's base and data in a single JSON object + result := fmt.Sprintf("%s,%s", base[:len(base)-1], args[1:]) + return []byte(result), nil +}} + +func UnmarshalCommand(data []byte) (Command, error) {{ + base := struct {{ + CommandId string `json:"id,omitempty"` + Name string `json:"execute"` + }}{{}} + if err := json.Unmarshal(data, &base); err != nil {{ + return nil, errors.New(fmt.Sprintf("Failed to decode command: %s", string(data))) + }} + + switch base.Name {{ + {unm_cases} + }} + return nil, errors.New("Failed to recognize command") +}} +''' + TEMPLATE_HELPER = ''' // Alias for go version lower than 1.18 type Any = interface{}
@@ -112,12 +169,13 @@ class QAPISchemaGenGolangVisitor(QAPISchemaVisitor): def __init__(self, prefix: str): super().__init__() - self.target = {name: "" for name in ["alternate", "enum", + self.target = {name: "" for name in ["alternate", "command", "enum", "event", "helper", "struct", "union"]} self.objects_seen = {} self.schema = None self.events = {} + self.commands = {} self.golang_package_name = "qapi" def visit_begin(self, schema):
@@ -149,6 +207,23 @@ def visit_end(self): ''' self.target["event"] += TEMPLATE_EVENT.format(unm_cases=unm_cases) + unm_cases = "" + for name in sorted(self.commands): + case_type = self.commands[name] + unm_cases += f''' + case "{name}": + command := struct {{ + Args {case_type} `json:"arguments"` + }}{{}} + + if err := json.Unmarshal(data, &command); err != nil {{ + return nil, errors.New(fmt.Sprintf("Failed to unmarshal: %s", string(data))) + }} + command.Args.CommandId = base.CommandId + return &command.Args, nil +''' + self.target["command"] += TEMPLATE_COMMAND.format(unm_cases=unm_cases) + def visit_object_type(self: QAPISchemaGenGolangVisitor, name: str,
@@ -308,7 +383,47 @@ def visit_command(self, allow_oob: bool, allow_preconfig: bool, coroutine: bool) -> None: - pass + # Safety check + assert name == info.defn_name + + type_name = qapi_to_go_type_name(name, info.defn_meta) + self.commands[name] = type_name + command_ret = "" + init_ret_type_name = f'''EmptyCommandReturn {{ Name: "{name}" }}''' + + self_contained = True + if arg_type and arg_type.name.startswith("q_obj"): + self_contained = False + + content = "" + if boxed or self_contained: + args = "" if not arg_type else "\n" + arg_type.name + args += '''\n\tCommandId string `json:"-"`''' + content = generate_struct_type(type_name, args) + else: + assert isinstance(arg_type, QAPISchemaObjectType) + content = qapi_to_golang_struct(name, + arg_type.info, + arg_type.ifcond, + arg_type.features, + arg_type.base, + arg_type.members, + arg_type.variants) + + methods = f''' +func (c *{type_name}) GetName() string {{ + return "{name}" +}} + +func (s *{type_name}) GetId() string {{ + return s.CommandId +}} + +func (s *{type_name}) GetReturnType() CommandReturn {{ + return &{init_ret_type_name} +}} +''' + self.target["command"] += content + methods def visit_event(self, name, info, ifcond, features, arg_type, boxed): assert name == info.defn_name
@@ -385,6 +500,8 @@ def qapi_to_golang_struct(name: str, fields = "" if info.defn_meta == "event": fields += '''\tEventTimestamp Timestamp `json:"-"`\n''' + elif info.defn_meta == "command": + fields += '''\tCommandId string `json:"-"`\n''' if base: base_fields = ""
@@ -569,7 +686,7 @@ def qapi_to_go_type_name(name: str, meta: str) -> str: name += ''.join(word.title() for word in words[1:]) - if meta in ["event"]: + if meta in ["event", "command"]: name = name[:-3] if name.endswith("Arg") else name name += meta.title().replace(" ", "")
Signature · 1 line
2.36.1
RFC PATCH v2 7/8 qapi: golang: Add CommandResult type to Go
In reply to Victor TosoThis patch adds a struct type in Go that will handle return values for QAPI's command types.
The return value of a Command is, encouraged to be, QAPI's complex types or an Array of those.
Every Command has a underlying CommandResult. The EmptyCommandReturn is for those that don't expect any data e.g: { "return": {} }.
All CommandReturn types implement the CommandResult interface.
Example:
qapi:
| { 'command': 'query-sev', 'returns': 'SevInfo',
| 'if': 'TARGET_I386' }
go:
| type QuerySevCommandReturn struct {
| CommandId string `json:"id,omitempty"`
| Result *SevInfo `json:"return"`
| Error *QapiError `json:"error,omitempty"`
| }usage:
| // One can use QuerySevCommandReturn directly or
| // command's interface GetReturnType() instead.
|
| input := `{ "return": { "enabled": true, "api-major" : 0,` +
| `"api-minor" : 0, "build-id" : 0,` +
| `"policy" : 0, "state" : "running",` +
| `"handle" : 1 } } `
| ret := QuerySevCommandReturn{}
| err := json.Unmarshal([]byte(input), &ret)
| if ret.Error != nil {
| // Handle command failure {"error": { ...}}
| } else if ret.Result != nil {
| // ret.Result.Enable == true
| }--- scripts/qapi/golang.py | 73 ++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 70 insertions(+), 3 deletions(-)
scripts/qapi/golang.py +70 -3
@@ -89,7 +89,8 @@ }} ''' -# Only variable is @unm_cases to handle all command's names and associated types. +# Only variable is @unm_cases to handle +# all command's names and associated types. TEMPLATE_COMMAND = ''' type Command interface {{ GetId() string
@@ -145,10 +146,49 @@ }} ''' +TEMPLATE_COMMAND_RETURN = ''' +type CommandReturn interface { + GetId() string + GetCommandName() string + GetError() error +} + +type EmptyCommandReturn struct { + CommandId string `json:"id,omitempty"` + Error *QapiError `json:"error,omitempty"` + Name string `json:"-"` +} + +func (r EmptyCommandReturn) MarshalJSON() ([]byte, error) { + return []byte(`{"return":{}}`), nil +} + +func (r *EmptyCommandReturn) GetId() string { + return r.CommandId +} + +func (r *EmptyCommandReturn) GetCommandName() string { + return r.Name +} + +func (r *EmptyCommandReturn) GetError() error { + return r.Error +} +''' + TEMPLATE_HELPER = ''' // Alias for go version lower than 1.18 type Any = interface{} +type QapiError struct { + Class string `json:"class"` + Description string `json:"desc"` +} + +func (err *QapiError) Error() string { + return fmt.Sprintf("%s: %s", err.Class, err.Description) +} + // Creates a decoder that errors on unknown Fields // Returns true if successfully decoded @from string @into type // Returns false without error is failed with "unknown field"
@@ -176,6 +216,7 @@ def __init__(self, prefix: str): self.schema = None self.events = {} self.commands = {} + self.command_results = {} self.golang_package_name = "qapi" def visit_begin(self, schema):
@@ -224,6 +265,7 @@ def visit_end(self): ''' self.target["command"] += TEMPLATE_COMMAND.format(unm_cases=unm_cases) + self.target["command"] += TEMPLATE_COMMAND_RETURN def visit_object_type(self: QAPISchemaGenGolangVisitor, name: str,
@@ -390,6 +432,31 @@ def visit_command(self, self.commands[name] = type_name command_ret = "" init_ret_type_name = f'''EmptyCommandReturn {{ Name: "{name}" }}''' + if ret_type: + cmd_ret_name = qapi_to_go_type_name(name, "command return") + ret_type_name = qapi_schema_type_to_go_type(ret_type.name) + init_ret_type_name = f'''{cmd_ret_name}{{}}''' + isptr = "*" if ret_type_name[0] not in "*[" else "" + self.command_results[name] = ret_type_name + command_ret = f''' +type {cmd_ret_name} struct {{ + CommandId string `json:"id,omitempty"` + Result {isptr}{ret_type_name} `json:"return"` + Error *QapiError `json:"error,omitempty"` +}} + +func (r *{cmd_ret_name}) GetCommandName() string {{ + return "{name}" +}} + +func (r *{cmd_ret_name}) GetId() string {{ + return r.CommandId +}} + +func (r *{cmd_ret_name}) GetError() error {{ + return r.Error +}} +''' self_contained = True if arg_type and arg_type.name.startswith("q_obj"):
@@ -423,7 +490,7 @@ def visit_command(self, return &{init_ret_type_name} }} ''' - self.target["command"] += content + methods + self.target["command"] += content + methods + command_ret def visit_event(self, name, info, ifcond, features, arg_type, boxed): assert name == info.defn_name
@@ -686,7 +753,7 @@ def qapi_to_go_type_name(name: str, meta: str) -> str: name += ''.join(word.title() for word in words[1:]) - if meta in ["event", "command"]: + if meta in ["event", "command", "command return"]: name = name[:-3] if name.endswith("Arg") else name name += meta.title().replace(" ", "")
Signature · 1 line
2.36.1
Re: [RFC PATCH v2 7/8] qapi: golang: Add CommandResult type to Go
scripts/qapi/golang.py line 160
Quoted 5 lines · Victor Toso
scripts/qapi/golang.py +5
-0+type EmptyCommandReturn struct { + CommandId string `json:"id,omitempty"` + Error *QapiError `json:"error,omitempty"` + Name string `json:"-"` +}
Do we need a specific type for this? Can't we just generate an appropriately-named type for each of the commands that don't return any data? It's not like we would have to write that code manually :)
scripts/qapi/golang.py line 172
Quoted 3 lines
scripts/qapi/golang.py +3
-0+func (r *EmptyCommandReturn) GetCommandName() string { + return r.Name +}
Just like Event.GetName() and Command.GetName(), I'm not convinced we should have this.
Of course, all the comments about how marshalling and unmarshalling are handled made for events also apply here.
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Re: [RFC PATCH v2 7/8] qapi: golang: Add CommandResult type to Go
Quoted 10 lines · Andrea Bolognani
scripts/qapi/golang.py line 160
Quoted 5 lines · Victor Toso
scripts/qapi/golang.py +5
-0+type EmptyCommandReturn struct { + CommandId string `json:"id,omitempty"` + Error *QapiError `json:"error,omitempty"` + Name string `json:"-"` +}Do we need a specific type for this? Can't we just generate an appropriately-named type for each of the commands that don't return any data? It's not like we would have to write that code manually :)
Yes, I think having an explicit named return struct even for commands not /currently/ returning data is good, and anticipates future changes that might add extra return data fields to the QAPI schema.
Quoted 8 lines
scripts/qapi/golang.py line 172
Quoted 4 lines
scripts/qapi/golang.py +3
-0+func (r *EmptyCommandReturn) GetCommandName() string { + return r.Name +}Just like
Event.GetName()andCommand.GetName(), I'm not convinced we should have this.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Re: [RFC PATCH v2 7/8] qapi: golang: Add CommandResult type to Go
Hi,
Quoted 12 lines · Daniel P. Berrangé
Quoted 10 lines · Andrea Bolognani
scripts/qapi/golang.py line 160
Quoted 5 lines · Victor Toso
scripts/qapi/golang.py +5
-0+type EmptyCommandReturn struct { + CommandId string `json:"id,omitempty"` + Error *QapiError `json:"error,omitempty"` + Name string `json:"-"` +}Do we need a specific type for this? Can't we just generate an appropriately-named type for each of the commands that don't return any data? It's not like we would have to write that code manually :)
Quoted 3 lines
Yes, I think having an explicit named return struct even for commands not /currently/ returning data is good, and anticipates future changes that might add extra return data fields to the QAPI schema.
Sure.
Cheers,
Victor
RFC PATCH v2 8/8 qapi: golang: document skip function visit_array_types
In reply to Victor Toso--- scripts/qapi/golang.py | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
scripts/qapi/golang.py +6 -1
@@ -410,7 +410,12 @@ def visit_enum_type(self: QAPISchemaGenGolangVisitor, ''' def visit_array_type(self, name, info, ifcond, element_type): - pass + # TLDR: We don't need to any extra boilerplate in Go to handle Arrays. + # + # This function is implemented just to be sure that: + # 1. Every array type ends with List + # 2. Every array type's element is the array type without 'List' + assert name.endswith("List") and name[:-4] == element_type.name def visit_command(self, name: str,
Signature · 1 line
2.36.1
Quoted 8 lines · Victor Toso
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html# What this is about?
Quoted 7 lines
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmp
Aspects of review:
(1) Impact on common code, if any
I care, because any messes made there are likely to affect me down the road.
(2) The generated Go code
Is it (close to) what we want long term? If not, is it good enough short term, and how could we make necessary improvements?
I'd prefer to leave this to folks who actually know their Go.
(3) General Python sanity
We need eyes, but not necessarily mine. Any takers?
[...]
Quoted 4 lines
scripts/qapi/golang.py | 765 +++++++++++++++++++++++++++++++++++++++++ scripts/qapi/main.py | 2 + 2 files changed, 767 insertions(+) create mode 100644 scripts/qapi/golang.py
This adds a new generator and calls it from generate(), i.e. review aspect (1) is empty. "Empty" is a quick & easy way to get my ACK!
No tests?
No documentation?
Hi Markus,
Quoted 18 lines · Markus Armbruster
Quoted 8 lines · Victor Toso
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html# What this is about?
Quoted 7 lines
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmp
Quoted 6 lines
Aspects of review:
(1) Impact on common code, if any
I care, because any messes made there are likely to affect me down the road.
For the first version of the Go generated interface, my goal is to have something that works and can be considered alpha to other Go projects.
With this first version, I don't want to bring huge changes to the python library or to the QAPI spec and its usage in QEMU. Unless someone finds that a necessity.
So I hope (1) is simple :)
Quoted 17 lines
(2) The generated Go code
Is it (close to) what we want long term? If not, is it good enough short term, and how could we make necessary improvements?
I'd prefer to leave this to folks who actually know their Go.
(3) General Python sanityWe need eyes, but not necessarily mine. Any takers?
[...]
Quoted 4 lines
scripts/qapi/golang.py | 765 +++++++++++++++++++++++++++++++++++++++++ scripts/qapi/main.py | 2 + 2 files changed, 767 insertions(+) create mode 100644 scripts/qapi/golang.py
Quoted 5 lines
This adds a new generator and calls it from
generate(), i.e. review aspect (1) is empty. "Empty" is a quick & easy way to get my ACK!No tests?
I've added tests but on the qapi-go module, those are the files with _test.go prefix on them. Example for commands:
https://gitlab.com/victortoso/qapi-go/-/blob/main/pkg/qapi/commands_test.go
Should the generator itself have tests or offloading that to the qapi-go seems reasonable?
Quoted 1 line
No documentation?
Yes, this iteration removed the documentation of the generated types. I'm a bit sad about that. I want to consume the documentation in the QAPI files to provide the latest info from types/fields but we can't 'copy' it, the only solution is 'refer' to it with hyperlink, which I haven't done yet.
Cheers,
Victor
Quoted 61 lines · Victor Toso
Hi Markus,
Quoted 18 lines · Markus Armbruster
Quoted 8 lines · Victor Toso
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html# What this is about?
Quoted 7 lines
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmpQuoted 6 lines
Aspects of review:
(1) Impact on common code, if any
I care, because any messes made there are likely to affect me down the road.
For the first version of the Go generated interface, my goal is to have something that works and can be considered alpha to other Go projects.
With this first version, I don't want to bring huge changes to the python library or to the QAPI spec and its usage in QEMU. Unless someone finds that a necessity.
So I hope (1) is simple :)
Quoted 17 lines
(2) The generated Go code
Is it (close to) what we want long term? If not, is it good enough short term, and how could we make necessary improvements?
I'd prefer to leave this to folks who actually know their Go.
(3) General Python sanityWe need eyes, but not necessarily mine. Any takers?
[...]
Quoted 4 lines
scripts/qapi/golang.py | 765 +++++++++++++++++++++++++++++++++++++++++ scripts/qapi/main.py | 2 + 2 files changed, 767 insertions(+) create mode 100644 scripts/qapi/golang.pyQuoted 5 lines
This adds a new generator and calls it from
generate(), i.e. review aspect (1) is empty. "Empty" is a quick & easy way to get my ACK!No tests?
Quoted 7 lines
I've added tests but on the qapi-go module, those are the files with _test.go prefix on them. Example for commands:
https://gitlab.com/victortoso/qapi-go/-/blob/main/pkg/qapi/commands_test.go
Should the generator itself have tests or offloading that to the qapi-go seems reasonable?
Offloading may be reasonable, but how am I to run the tests then? Documentation should tell me.
We have roughly three kinds of tests so far:
1. Front end tests in tests/qapi-schema
2. Unit tests in tests/unit/
To find them:
$ git-grep '#include ".*qapi-.*\.h"' tests/unit/
3. Many tests talking QMP in tests/qtest/
Since you leave the front end alone, you don't need the first kind.
The other two kinds are less clear.
Quoted 7 lines
Quoted 1 line
No documentation?
Yes, this iteration removed the documentation of the generated types. I'm a bit sad about that. I want to consume the documentation in the QAPI files to provide the latest info from types/fields but we can't 'copy' it, the only solution is 'refer' to it with hyperlink, which I haven't done yet.
Two kinds of documentation: generated documentation for the generated Go code, and documentation about the generator. I was thinking of the latter. Specifically, docs/devel/qapi-code-gen.rst section "Code generation". Opinions on its usefulness differ. I like it.
Hi,
Quoted 78 lines · Markus Armbruster
Quoted 61 lines · Victor Toso
Hi Markus,
Quoted 18 lines · Markus Armbruster
Quoted 8 lines · Victor Toso
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html# What this is about?
Quoted 7 lines
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmpQuoted 6 lines
Aspects of review:
(1) Impact on common code, if any
I care, because any messes made there are likely to affect me down the road.
For the first version of the Go generated interface, my goal is to have something that works and can be considered alpha to other Go projects.
With this first version, I don't want to bring huge changes to the python library or to the QAPI spec and its usage in QEMU. Unless someone finds that a necessity.
So I hope (1) is simple :)
Quoted 17 lines
(2) The generated Go code
Is it (close to) what we want long term? If not, is it good enough short term, and how could we make necessary improvements?
I'd prefer to leave this to folks who actually know their Go.
(3) General Python sanityWe need eyes, but not necessarily mine. Any takers?
[...]
Quoted 4 lines
scripts/qapi/golang.py | 765 +++++++++++++++++++++++++++++++++++++++++ scripts/qapi/main.py | 2 + 2 files changed, 767 insertions(+) create mode 100644 scripts/qapi/golang.pyQuoted 5 lines
This adds a new generator and calls it from
generate(), i.e. review aspect (1) is empty. "Empty" is a quick & easy way to get my ACK!No tests?
Quoted 7 lines
I've added tests but on the qapi-go module, those are the files with _test.go prefix on them. Example for commands:
https://gitlab.com/victortoso/qapi-go/-/blob/main/pkg/qapi/commands_test.go
Should the generator itself have tests or offloading that to the qapi-go seems reasonable?
Offloading may be reasonable, but how am I to run the tests then? Documentation should tell me.
We have roughly three kinds of tests so far:
1. Front end tests in tests/qapi-schema
Quoted 7 lines
2. Unit tests in tests/unit/
To find them:
$ git-grep '#include ".*qapi-.*\.h"' tests/unit/3. Many tests talking QMP in tests/qtest/
I'm thinking on the tests in QEMU side. Perhaps adding something with Avocado that generates the qapi-go and communicates with a running QEMU with that generated Go module?
One thing that I try to keep in mind is to not add Go dependencies in QEMU and this Golang work is not internal to QEMU itself.
Quoted 4 lines
Since you leave the front end alone, you don't need the first kind.
The other two kinds are less clear.
I'm open for suggestions. I thought that, if we have a qapi-go Go module in Gitlab's qemu-project namespace, we could leverage most of the tests to the consumer of the actual generated code but I agree that it is necessary to have something in qemu too.
Quoted 8 lines
Quoted 7 lines
Quoted 1 line
No documentation?
Yes, this iteration removed the documentation of the generated types. I'm a bit sad about that. I want to consume the documentation in the QAPI files to provide the latest info from types/fields but we can't 'copy' it, the only solution is 'refer' to it with hyperlink, which I haven't done yet.
Quoted 4 lines
Two kinds of documentation: generated documentation for the generated Go code, and documentation about the generator. I was thinking of the latter. Specifically,
docs/devel/qapi-code-gen.rstsection "Code generation". Opinions on its usefulness differ. I like it.
Me too. I'll add documentation for the next iteration, thanks for pointing it out.
Cheers,
Victor
I've commented in detail to the single patches, just a couple of additional points.
Quoted 8 lines · Victor Toso
* 7) Flat structs by removing embed types. Discussion with Andrea
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01590.htmlNo one required it but I decided to give it a try. Major issue that I see with this approach is to have generated a few 'Base' structs that are now useless. Overall, less nested structs seems better to me. Opnions?
Quoted 6 lines
Example:
| /* This is now useless, should be removed? */ | type InetSocketAddressBase struct {
| Host stringjson:"host"
| Port stringjson:"port"
| }
Can we somehow keep track, in the generator, of types that are only used as building blocks for other types, and prevent them from showing up in the generated code?
Finally, looking at the repository containing the generated code I see that the generated type are sorted by kind, e.g. all unions are in a file, all events in another one and so on. I believe the structure should match more closely that of the QAPI schema, so e.g. block-related types should all go in one file, net-related types in another one and so on.
Looking forward to the next iteration :)
Signature · 1 line
Andrea Bolognani / Red Hat / Virtualization
Hi,
Quoted 19 lines · Andrea Bolognani
I've commented in detail to the single patches, just a couple of additional points.
Quoted 8 lines · Victor Toso
* 7) Flat structs by removing embed types. Discussion with Andrea
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01590.htmlNo one required it but I decided to give it a try. Major issue that I see with this approach is to have generated a few 'Base' structs that are now useless. Overall, less nested structs seems better to me. Opnions?
Quoted 6 lines
Example:
| /* This is now useless, should be removed? */ | type InetSocketAddressBase struct {
| Host stringjson:"host"
| Port stringjson:"port"
| }
Quoted 3 lines
Can we somehow keep track, in the generator, of types that are only used as building blocks for other types, and prevent them from showing up in the generated code?
I'm not 100% sure it is good to remove them from generated code because technically it is a valid qapi type. If all @base types are embed types and they don't show in other way or form, sure we can remove them from generated code... I'm not sure if it is possible to guarantee this.
But yes, if possible, I'd like to remove what seems useless type definitions.
Quoted 6 lines
Finally, looking at the repository containing the generated code I see that the generated type are sorted by kind, e.g. all unions are in a file, all events in another one and so on. I believe the structure should match more closely that of the QAPI schema, so e.g. block-related types should all go in one file, net-related types in another one and so on.
That's something I don't mind adding but some hardcoded mapping is needed. If you look into git history of qapi/ folder, .json files can come and go, types be moved around, etc. So, we need to proper map types in a way that the generated code would be kept stable even if qapi files would have been rearranged. What I proposed was only the simplest solution.
Also, the generator takes a qapi-schema.json as input. We are more focused in qemu/qapi/qapi-schema.json generated coded but would not hurt to think we could even use it for qemu-guest-agent from qemu/qga/qapi-schema.json -- this to say that the hardcoded mapping needs to take into account non qemu qapi schemas too.
Quoted 1 line
Looking forward to the next iteration :)
Me too, thanks again!
Cheers,
Victor
Quoted 26 lines · Victor Toso
Hi,
Quoted 19 lines · Andrea Bolognani
I've commented in detail to the single patches, just a couple of additional points.
Quoted 8 lines · Victor Toso
* 7) Flat structs by removing embed types. Discussion with Andrea
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01590.htmlNo one required it but I decided to give it a try. Major issue that I see with this approach is to have generated a few 'Base' structs that are now useless. Overall, less nested structs seems better to me. Opnions?
Quoted 6 lines
Example:
| /* This is now useless, should be removed? */ | type InetSocketAddressBase struct {
| Host stringjson:"host"
| Port stringjson:"port"
| }Quoted 3 lines
Can we somehow keep track, in the generator, of types that are only used as building blocks for other types, and prevent them from showing up in the generated code?
Quoted 8 lines
I'm not 100% sure it is good to remove them from generated code because technically it is a valid qapi type. If all @base types are embed types and they don't show in other way or form, sure we can remove them from generated code... I'm not sure if it is possible to guarantee this.
But yes, if possible, I'd like to remove what seems useless type definitions.
The existing C generators have to generate all the types, because the generated code is for QEMU's own use, where we need all the types.
The existing introspection generator generates only the types visible in QAPI/QMP introspection.
The former generate for internal use (where we want all the types), and the latter for external use (where only the types visible in the external interface are actually useful).
Quoted 14 lines
Quoted 6 lines
Finally, looking at the repository containing the generated code I see that the generated type are sorted by kind, e.g. all unions are in a file, all events in another one and so on. I believe the structure should match more closely that of the QAPI schema, so e.g. block-related types should all go in one file, net-related types in another one and so on.
That's something I don't mind adding but some hardcoded mapping is needed. If you look into git history of qapi/ folder, .json files can come and go, types be moved around, etc. So, we need to proper map types in a way that the generated code would be kept stable even if qapi files would have been rearranged. What I proposed was only the simplest solution.
Quoted 5 lines
Also, the generator takes a qapi-schema.json as input. We are more focused in
qemu/qapi/qapi-schema.jsongenerated coded but would not hurt to think we could even use it for qemu-guest-agent fromqemu/qga/qapi-schema.json-- this to say that the hardcoded mapping needs to take into account non qemu qapi schemas too.
In the beginning, the QAPI schema was monolithic. qga/qapi-schema.json still is.
When keeping everything in a single qapi-schema.json became unwieldy, we split it into "modules" tied together with a simple include directive. Generated code remained monolithic.
When monolithic generated code became too annoying (touch schema, recompile everything), we made it match the module structure: code for FOO.json goes into *-FOO.c and *-FOO.h, where the *-FOO.h #include the generated headers for the .json modules FOO.json includes.
Schema code motion hasn't been much of a problem. Moving from FOO.json to one of the modules it includes is transparent. Non-transparent moves are relatively rare as long as the split into modules actually makes sense.
Quoted 6 lines
Quoted 1 line
Looking forward to the next iteration :)
Me too, thanks again!
Cheers,
Victor
Hi,
Quoted 40 lines · Markus Armbruster
Quoted 26 lines · Victor Toso
Hi,
Quoted 19 lines · Andrea Bolognani
I've commented in detail to the single patches, just a couple of additional points.
Quoted 8 lines · Victor Toso
* 7) Flat structs by removing embed types. Discussion with Andrea
Thread: https://lists.gnu.org/archive/html/qemu-devel/2022-05/msg01590.htmlNo one required it but I decided to give it a try. Major issue that I see with this approach is to have generated a few 'Base' structs that are now useless. Overall, less nested structs seems better to me. Opnions?
Quoted 6 lines
Example:
| /* This is now useless, should be removed? */ | type InetSocketAddressBase struct {
| Host stringjson:"host"
| Port stringjson:"port"
| }Quoted 3 lines
Can we somehow keep track, in the generator, of types that are only used as building blocks for other types, and prevent them from showing up in the generated code?
Quoted 8 lines
I'm not 100% sure it is good to remove them from generated code because technically it is a valid qapi type. If all @base types are embed types and they don't show in other way or form, sure we can remove them from generated code... I'm not sure if it is possible to guarantee this.
But yes, if possible, I'd like to remove what seems useless type definitions.
The existing C generators have to generate all the types, because the generated code is for QEMU's own use, where we need all the types.
Quoted 6 lines
The existing introspection generator generates only the types visible in QAPI/QMP introspection.
The former generate for internal use (where we want all the types), and the latter for external use (where only the types visible in the external interface are actually useful).
My doubt are on types that might be okay to be hidden because they are embed in other types, like InetSocketAddressBase.
Note that what I mean with the struct being embed is that the actual fields of InetSocketAddressBase being added to the type which uses it, like InetSocketAddress.
| type InetSocketAddressBase struct {
| Host string `json:"host"`
| Port string `json:"port"`
| }
|
| type InetSocketAddress struct {
| // Base fields
| Host string `json:"host"`
| Port string `json:"port"`
|
| Numeric *bool `json:"numeric,omitempty"`
| To *uint16 `json:"to,omitempty"`
| Ipv4 *bool `json:"ipv4,omitempty"`
| Ipv6 *bool `json:"ipv6,omitempty"`
| KeepAlive *bool `json:"keep-alive,omitempty"`
| Mptcp *bool `json:"mptcp,omitempty"`
| }Andrea's suggestions is to have the generator to track if a given type is always embed in which case we can skip generating it in the Go module.
I think that could work indeed. In the hypothetical case that hiddens structs like InetSocketAddressBase becomes a parameter on command in the future, the generator would know and start generating this type from that point onwards.
Quoted 33 lines
Quoted 14 lines
Quoted 6 lines
Finally, looking at the repository containing the generated code I see that the generated type are sorted by kind, e.g. all unions are in a file, all events in another one and so on. I believe the structure should match more closely that of the QAPI schema, so e.g. block-related types should all go in one file, net-related types in another one and so on.
That's something I don't mind adding but some hardcoded mapping is needed. If you look into git history of qapi/ folder, .json files can come and go, types be moved around, etc. So, we need to proper map types in a way that the generated code would be kept stable even if qapi files would have been rearranged. What I proposed was only the simplest solution.
Quoted 5 lines
Also, the generator takes a qapi-schema.json as input. We are more focused in
qemu/qapi/qapi-schema.jsongenerated coded but would not hurt to think we could even use it for qemu-guest-agent fromqemu/qga/qapi-schema.json-- this to say that the hardcoded mapping needs to take into account non qemu qapi schemas too.In the beginning, the QAPI schema was monolithic.
qga/qapi-schema.jsonstill is.When keeping everything in a single qapi-schema.json became unwieldy, we split it into "modules" tied together with a simple include directive. Generated code remained monolithic.
When monolithic generated code became too annoying (touch schema, recompile everything), we made it match the module
structure: code for FOO.json goes into *-FOO.c and *-FOO.h, where the *-FOO.h #include the generated headers for the .json modules FOO.json includes.
Quoted 4 lines
Schema code motion hasn't been much of a problem. Moving from FOO.json to one of the modules it includes is transparent. Non-transparent moves are relatively rare as long as the split into modules actually makes sense.
To be honest, splitting it into different files based on their module names should not be a problem if we keep them in a single Go module as do intend to for this generated go code.
So I'll go ahead and split it.
Cheers,
Victor
Bringing this thread back from the dead, since I had an in-person discussion on a key question below at KVM Forum this year and want to record it here.
Quoted 28 lines · Victor Toso
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html# What this is about?
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmp# Issues raised in RFC v1
The leading '*' for issues I addressed in this iteration
* 1) Documentation was removed to avoid License issues, by Daniel
Thread: https://lists.nongnu.org/archive/html/qemu-devel/2022-05/msg01889.htmlIt is important for the generated Go module to be compatible with Licenses used by projects that would be using this. Copying the documentation of the QAPI spec might conflict with GPLv2+.
Quoted 6 lines
I have not proposed another license in this iteration, but I'm planning to go with MIT No Attribution, aka MIT-0 [0]. Does it make sense to bind the generated code's license to MIT-0 already at generator level?
To recap the situation
* The license of the code generator itself does not determine the license of the output generated code
* The license of the inputs to the code generator, may or may not, determine the license of the output generated code depending on use context
The primary input to the code generator is the QAPI schema, which is part of QEMU and thus licensed GPL-2.0-or-later.
The QAPI schema includes both the API definitions AND the API documentation text.
We can make the case that as the QEMU public interface, consuming the API definitions in the QAPI schema for the purpose of generating code is "fair use", and thus the output generated code does NOT need to match the GPL-2.0-or-later license of the QAPI schema. We can choose the code license, and a maximally permissive license looks appropriate.
We want to have API documentation for the Golang bindings and the obvious way to achieve this is to copy the API docs from the QAPI schema into the Golang code. It is NOT reasonable to class such *direct copying* of docs "fair use". IOW, copied docs will be under GPL-2.0-or-later.
Thus if we pick MIT-0 for the Golang code, and copy across the QAPI docs, the resulting Golang QAPI project code would be under a compound license term "MIT-0 AND GPL-2.0-or-later".
The concern was that this will limit the ability of downstream applications to consume the Golang bindings, if they don't want their combined work to contain GPL-2.0-or-later.
Ignoring whether this fear of GPL-2.0-or-later is sane or not, in retrospect I believe that this concern in fact has no legal basis.
The license of a compiled application binary is the union of all the object files linked into it.
Notice I said "Object file" there, **NOT** "Source file".
This is the crucial distinction that makes the presense of GPL-2.0-or-later docs a non-issue from a licensing POV.
When the compiler takes the "MIT-0 and GPL-2.0-or-later" license .go source file, and builds an object file, it will be discarding all the API documentation comments. IOW, all the parts that were under GPL-2.0-or-later are discarded. The only parts of the source file that get "compiled" are the Go language constructs which were MIT-0 licensed [1].
IOW, we have a "MIT-0 and GPL-2.0-or-later" .go source file, and an "MIT-0" object file.
Thus while there may be a human perception problem with the Golang bindings being "MIT-0 and GPL-2.0-or-later", there are no legal licensing limitations, as the combined work for a library or application linking the bindings will only contain the MIT-0 part.
The GPL-2.0-or-later docs won't influence the license of the combined work.
Note, this interpretation applies only to languages which are compiled, not interpreted.
If we are generating python code for example, the there is no "source file" vs "object file" distinction for licensing. The combined work in a python app is under the union of all the source file licenses.
TL;DR: I think you can re-add the documentation comments to the Golang code generator, declare the resulting code as being "MIT-0 AND GPL-2.0-or-later".
In the README.md file we need todo two important things:
* Document our interpretation of the "combined work" license situation for applications consuming the project.
* Declare that *ALL* manually written code contributions are exclusively MIT-0.
With regards,
Daniel
[1] NB, there are languages where some code comments can have semantic impacts on the compiled out. I don't believe that to be the case for any human targetted API docs that we would be copying over from the QAPI schema though in the Golang case.
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Quoted 3 lines · Daniel P. Berrangé
Bringing this thread back from the dead, since I had an in-person discussion on a key question below at KVM Forum this year and want to record it here.
[...]
Quoted 4 lines
To recap the situation
* The license of the code generator itself does not determine the license of the output generated code
For instance, GNU Bison is GPLv3+, but the parts the generator emits (the parser skeleton) come with a special exception.
Such exceptions need to be granted by the copyright holder. As long as the code generating Go is not a derived work, the copyright holder situation should be simple enough to make this practical.
Quoted 7 lines
* The license of the inputs to the code generator, may or may not, determine the license of the output generated code depending on use context
The primary input to the code generator is the QAPI schema, which is part of QEMU and thus licensed GPL-2.0-or-later.
Quoted 8 lines
The QAPI schema includes both the API definitions AND the API documentation text.
We can make the case that as the QEMU public interface, consuming the API definitions in the QAPI schema for the purpose of generating code is "fair use", and thus the output generated code does NOT need to match the GPL-2.0-or-later license of the QAPI schema. We can choose the code license, and a maximally permissive license looks appropriate.
Having this argument confirmed by an actual expert seems advisable.
Quoted 57 lines
We want to have API documentation for the Golang bindings and the obvious way to achieve this is to copy the API docs from the QAPI schema into the Golang code. It is NOT reasonable to class such *direct copying* of docs "fair use". IOW, copied docs will be under GPL-2.0-or-later.
Thus if we pick MIT-0 for the Golang code, and copy across the QAPI docs, the resulting Golang QAPI project code would be under a compound license term "MIT-0 AND GPL-2.0-or-later".
The concern was that this will limit the ability of downstream applications to consume the Golang bindings, if they don't want their combined work to contain GPL-2.0-or-later.
Ignoring whether this fear of GPL-2.0-or-later is sane or not, in retrospect I believe that this concern in fact has no legal basis.
The license of a compiled application binary is the union of all the object files linked into it.
Notice I said "Object file" there, **NOT** "Source file".
This is the crucial distinction that makes the presense of GPL-2.0-or-later docs a non-issue from a licensing POV.
When the compiler takes the "MIT-0 and GPL-2.0-or-later" license .go source file, and builds an object file, it will be discarding all the API documentation comments. IOW, all the parts that were under GPL-2.0-or-later are discarded. The only parts of the source file that get "compiled" are the Go language constructs which were MIT-0 licensed [1].
IOW, we have a "MIT-0 and GPL-2.0-or-later" .go source file, and an "MIT-0" object file.
Thus while there may be a human perception problem with the Golang bindings being "MIT-0 and GPL-2.0-or-later", there are no legal licensing limitations, as the combined work for a library or application linking the bindings will only contain the MIT-0 part.
The GPL-2.0-or-later docs won't influence the license of the combined work.
Note, this interpretation applies only to languages which are compiled, not interpreted.
If we are generating python code for example, the there is no "source file" vs "object file" distinction for licensing. The combined work in a python app is under the union of all the source file licenses.
TL;DR: I think you can re-add the documentation comments to the Golang code generator, declare the resulting code as being "MIT-0 AND GPL-2.0-or-later".
Quoted 6 lines
In the README.md file we need todo two important things:
* Document our interpretation of the "combined work" license situation for applications consuming the project.
* Declare that *ALL* manually written code contributions are exclusively MIT-0.
What code do you have in mind? Can you provide an example?
Quoted 7 lines
With regards,
Daniel[1] NB, there are languages where some code comments can have semantic impacts on the compiled out. I don't believe that to be the case for any human targetted API docs that we would be copying over from the QAPI schema though in the Golang case.
Quoted 30 lines · Markus Armbruster
Quoted 3 lines · Daniel P. Berrangé
Bringing this thread back from the dead, since I had an in-person discussion on a key question below at KVM Forum this year and want to record it here.
[...]
Quoted 4 lines
To recap the situation
* The license of the code generator itself does not determine the license of the output generated code
For instance, GNU Bison is GPLv3+, but the parts the generator emits (the parser skeleton) come with a special exception.
Such exceptions need to be granted by the copyright holder. As long as the code generating Go is not a derived work, the copyright holder situation should be simple enough to make this practical.
Quoted 10 lines
* The license of the inputs to the code generator, may or may not, determine the license of the output generated code depending on use context
The primary input to the code generator is the QAPI schema, which is part of QEMU and thus licensed GPL-2.0-or-later.
The QAPI schema includes both the API definitions AND the API documentation text.
Quoted 7 lines
Quoted 5 lines
We can make the case that as the QEMU public interface, consuming the API definitions in the QAPI schema for the purpose of generating code is "fair use", and thus the output generated code does NOT need to match the GPL-2.0-or-later license of the QAPI schema. We can choose the code license, and a maximally permissive license looks appropriate.
Having this argument confirmed by an actual expert seems advisable.
IANAL, but .... :-)
The distinction of "source code" license vs "combined work" license, is a quite normal concept. As an example, it is something that is applied in Fedora RPM packaging when deciding the RPM spec "License:" tag. The latter is considered to refer to the licenses on the subset of source is said to be part of the distributed binaries, so we can exclude licenses of any part of the source that is not compiled, and of supporting code - eg license of makefiles, meson file, autoconf, code is excluded from the RPM license tag.
The difference is that in most cases, the excluded part of the source is in separate files, or less frequently could be in #ifdef sections, whereas in this case we have two differently licensed parts of work in the same file.
The question is thus what can you say is "compiled".
Is the whole file "compiled", or are only the souce code constructs "compiled".
My proposition is that, in most cases, comments are not used by the "compilation", or "code generation" phase. They are seen by the "parsing" phase only and thus dont contribute to the contents of the output binary code.
With C code this is easy to illustrate. Run 'gcc -save-temps' and look at the '.i' file which is what the compiler uses as input for code generation after cpp has run. The '.i' file is devoid of all code comments.
I've not proved the same for golang compilers yet, but my argument hinges on that being the true.
None the less, we have raised this position/viewpoint with experts for a second opinion.
Quoted 57 lines
Quoted 57 lines
We want to have API documentation for the Golang bindings and the obvious way to achieve this is to copy the API docs from the QAPI schema into the Golang code. It is NOT reasonable to class such *direct copying* of docs "fair use". IOW, copied docs will be under GPL-2.0-or-later.
Thus if we pick MIT-0 for the Golang code, and copy across the QAPI docs, the resulting Golang QAPI project code would be under a compound license term "MIT-0 AND GPL-2.0-or-later".
The concern was that this will limit the ability of downstream applications to consume the Golang bindings, if they don't want their combined work to contain GPL-2.0-or-later.
Ignoring whether this fear of GPL-2.0-or-later is sane or not, in retrospect I believe that this concern in fact has no legal basis.
The license of a compiled application binary is the union of all the object files linked into it.
Notice I said "Object file" there, **NOT** "Source file".
This is the crucial distinction that makes the presense of GPL-2.0-or-later docs a non-issue from a licensing POV.
When the compiler takes the "MIT-0 and GPL-2.0-or-later" license .go source file, and builds an object file, it will be discarding all the API documentation comments. IOW, all the parts that were under GPL-2.0-or-later are discarded. The only parts of the source file that get "compiled" are the Go language constructs which were MIT-0 licensed [1].
IOW, we have a "MIT-0 and GPL-2.0-or-later" .go source file, and an "MIT-0" object file.
Thus while there may be a human perception problem with the Golang bindings being "MIT-0 and GPL-2.0-or-later", there are no legal licensing limitations, as the combined work for a library or application linking the bindings will only contain the MIT-0 part.
The GPL-2.0-or-later docs won't influence the license of the combined work.
Note, this interpretation applies only to languages which are compiled, not interpreted.
If we are generating python code for example, the there is no "source file" vs "object file" distinction for licensing. The combined work in a python app is under the union of all the source file licenses.
TL;DR: I think you can re-add the documentation comments to the Golang code generator, declare the resulting code as being "MIT-0 AND GPL-2.0-or-later".
Quoted 8 lines
Quoted 6 lines
In the README.md file we need todo two important things:
* Document our interpretation of the "combined work" license situation for applications consuming the project.
* Declare that *ALL* manually written code contributions are exclusively MIT-0.What code do you have in mind? Can you provide an example?
I don't have any particular examples. Perhaps 100% of the .go code in the repo in question will be auto-generated. I can't predict the future, so there's a decent chance we might end up with some basic scaffolding and/or layered framework code that is manually written. If such a situation arises, it is important to have anticipated this, to make clear it is intended to be under a specific license.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Quoted 41 lines · Daniel P. Berrangé
Quoted 30 lines · Markus Armbruster
Quoted 3 lines · Daniel P. Berrangé
Bringing this thread back from the dead, since I had an in-person discussion on a key question below at KVM Forum this year and want to record it here.
[...]
Quoted 4 lines
To recap the situation
* The license of the code generator itself does not determine the license of the output generated code
For instance, GNU Bison is GPLv3+, but the parts the generator emits (the parser skeleton) come with a special exception.
Such exceptions need to be granted by the copyright holder. As long as the code generating Go is not a derived work, the copyright holder situation should be simple enough to make this practical.
Quoted 10 lines
* The license of the inputs to the code generator, may or may not, determine the license of the output generated code depending on use context
The primary input to the code generator is the QAPI schema, which is part of QEMU and thus licensed GPL-2.0-or-later.
The QAPI schema includes both the API definitions AND the API documentation text.
Quoted 7 lines
Quoted 5 lines
We can make the case that as the QEMU public interface, consuming the API definitions in the QAPI schema for the purpose of generating code is "fair use", and thus the output generated code does NOT need to match the GPL-2.0-or-later license of the QAPI schema. We can choose the code license, and a maximally permissive license looks appropriate.
Having this argument confirmed by an actual expert seems advisable.
IANAL, but .... :-)
Quoted 8 lines
My proposition is that, in most cases, comments are not used by the "compilation", or "code generation" phase. They are seen by the "parsing" phase only and thus dont contribute to the contents of the output binary code.
None the less, we have raised this position/viewpoint with experts for a second opinion.
Another thought or two...
From a QEMU community POV we have no license problem no matter what, so from a legal POV this risk free for us.
The licensing question is purely one for application developers consuming our work who want to avoid GPL for some reason.
Lets say my interpretation is wrong, or even that it is right, but the app developers disagree, or are none the less unhappy in some way. What impact would this have for QEMU ?
Code generation is practically free once the generator is written, so we just run it twice, storing the output in git on different branches with parallel tags, once with comments and once without. eg
The "main" branch (and associated versioned vX.Y.Z tags) in qemu-go-qapi.git provide the full QAPI go source + copied QAPI comments under MIT-0 AND GPL-2.0-or-later
The "minimal" branch (and associated vX.Y.Z-minimal tags) in qemu-go-qapi.git provide only the QAPI go source, without any comments under MIT-0 only.
With regards,
Daniel
Signature · 3 lines
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
Quoted 2 lines · Daniel P. Berrangé
Quoted 1 line · Markus Armbruster
Daniel P. Berrangé <[email protected]> writes:
[...]
Quoted 11 lines
Quoted 11 lines
Quoted 10 lines
TL;DR: I think you can re-add the documentation comments to the Golang code generator, declare the resulting code as being "MIT-0 AND GPL-2.0-or-later".
In the README.md file we need todo two important things:
* Document our interpretation of the "combined work" license situation for applications consuming the project.
* Declare that *ALL* manually written code contributions are exclusively MIT-0.
Quoted 6 lines
Quoted 1 line
What code do you have in mind? Can you provide an example?
I don't have any particular examples. Perhaps 100% of the .go code in the repo in question will be auto-generated. I can't predict the future, so there's a decent chance we might end up with some basic scaffolding and/or layered framework code that is manually written.
Yes.
Quoted 2 lines
If such a situation arises, it is important to have anticipated this, to make clear it is intended to be under a specific license.
Makes sense.
Hi,
Quoted 3 lines · Daniel P. Berrangé
Bringing this thread back from the dead, since I had an in-person discussion on a key question below at KVM Forum this year and want to record it here.
Thanks for that.
Quoted 110 lines · Victor Toso
Quoted 28 lines
Hi,
This is the second iteration of RFC v1:
https://lists.gnu.org/archive/html/qemu-devel/2022-04/msg00226.html# What this is about?
To generate a simple Golang interface that could communicate with QEMU over QMP. The Go code that is generated is meant to be used as the bare bones to exchange QMP messages.
The goal is to have this as a Go module in QEMU gitlab namespace, similar to what have been done to pyhon-qemu-qmp
https://gitlab.com/qemu-project/python-qemu-qmp# Issues raised in RFC v1
The leading '*' for issues I addressed in this iteration
* 1) Documentation was removed to avoid License issues, by Daniel
Thread: https://lists.nongnu.org/archive/html/qemu-devel/2022-05/msg01889.htmlIt is important for the generated Go module to be compatible with Licenses used by projects that would be using this. Copying the documentation of the QAPI spec might conflict with GPLv2+.
Quoted 6 lines
I have not proposed another license in this iteration, but I'm planning to go with MIT No Attribution, aka MIT-0 [0]. Does it make sense to bind the generated code's license to MIT-0 already at generator level?
To recap the situation
* The license of the code generator itself does not determine the license of the output generated code
* The license of the inputs to the code generator, may or may not, determine the license of the output generated code depending on use context
The primary input to the code generator is the QAPI schema, which is part of QEMU and thus licensed GPL-2.0-or-later.
The QAPI schema includes both the API definitions AND the API documentation text.
We can make the case that as the QEMU public interface, consuming the API definitions in the QAPI schema for the purpose of generating code is "fair use", and thus the output generated code does NOT need to match the GPL-2.0-or-later license of the QAPI schema. We can choose the code license, and a maximally permissive license looks appropriate.
We want to have API documentation for the Golang bindings and the obvious way to achieve this is to copy the API docs from the QAPI schema into the Golang code. It is NOT reasonable to class such *direct copying* of docs "fair use". IOW, copied docs will be under GPL-2.0-or-later.
Thus if we pick MIT-0 for the Golang code, and copy across the QAPI docs, the resulting Golang QAPI project code would be under a compound license term "MIT-0 AND GPL-2.0-or-later".
The concern was that this will limit the ability of downstream applications to consume the Golang bindings, if they don't want their combined work to contain GPL-2.0-or-later.
Ignoring whether this fear of GPL-2.0-or-later is sane or not, in retrospect I believe that this concern in fact has no legal basis.
The license of a compiled application binary is the union of all the object files linked into it.
Notice I said "Object file" there, **NOT** "Source file".
This is the crucial distinction that makes the presense of GPL-2.0-or-later docs a non-issue from a licensing POV.
When the compiler takes the "MIT-0 and GPL-2.0-or-later" license .go source file, and builds an object file, it will be discarding all the API documentation comments. IOW, all the parts that were under GPL-2.0-or-later are discarded. The only parts of the source file that get "compiled" are the Go language constructs which were MIT-0 licensed [1].
IOW, we have a "MIT-0 and GPL-2.0-or-later" .go source file, and an "MIT-0" object file.
Thus while there may be a human perception problem with the Golang bindings being "MIT-0 and GPL-2.0-or-later", there are no legal licensing limitations, as the combined work for a library or application linking the bindings will only contain the MIT-0 part.
The GPL-2.0-or-later docs won't influence the license of the combined work.
Note, this interpretation applies only to languages which are compiled, not interpreted.
If we are generating python code for example, the there is no "source file" vs "object file" distinction for licensing. The combined work in a python app is under the union of all the source file licenses.
Quoted 3 lines
TL;DR: I think you can re-add the documentation comments to the Golang code generator, declare the resulting code as being "MIT-0 AND GPL-2.0-or-later".
Yes. I think so too. I plan to add it back and address Andrea's last review and submit it in the next week or so.
Quoted 7 lines
In the README.md file we need todo two important things:
* Document our interpretation of the "combined work" license situation for applications consuming the project.
* Declare that *ALL* manually written code contributions are exclusively MIT-0.
Quoted 7 lines
With regards,
Daniel[1] NB, there are languages where some code comments can have semantic impacts on the compiled out. I don't believe that to be the case for any human targetted API docs that we would be copying over from the QAPI schema though in the Golang case.
Cheers,
Victor