Skip to contentSkip to conversation

reftable: improvements for the `binsearch()` mechanism

PATCH Superseded Patrick Steinhardt 2 participants 45 messages Mar 22, 2024 – Apr 3, 2024

lore
Share & tools
lore
#0 lore raw

cover reftable: improvements for the `binsearch()` mechanism

Hi,

this patch series contains various improvements for the binsearch() mechanism used by the reftable library:

- Callsites are refactored to make it more obvious what exactly they are doing. This mostly involves improved names for callback functions.

- Error handling is improved such that binsearch() knows to abort when the callback returns a negative value.

- Error handling for binary searching over restart counters is fixed so that we check for errors at the correct point in time.

- Decoding of record keys is refactored when searching over restart counters to not require allocations anymore.

Thanks!

Patrick

Patrick Steinhardt (7):
reftable/basics: fix return type of binsearch() to be size_t reftable/basics: improve binsearch() test reftable/refname: refactor binary search over refnames reftable/block: refactor binary search over restart points reftable/block: fix error handling when searching restart points reftable/record: extract function to decode key lengths reftable/block: avoid decoding keys when searching restart points

 reftable/basics.c      |   7 ++-
 reftable/basics.h      |   7 +--
 reftable/basics_test.c |  55 +++++++++++---------
 reftable/block.c       | 114 ++++++++++++++++++++++++++++++-----------
 reftable/record.c      |  34 ++++++++----
 reftable/record.h      |   6 +++
 reftable/refname.c     |  53 +++++++++----------
 7 files changed, 179 insertions(+), 97 deletions(-)
Signature · 1 line
2.44.0
#1 lore raw

PATCH 1/7 reftable/basics: fix return type of `binsearch()` to be `size_t`

The binsearch() function can be used to find the first element for which a callback functions returns a truish value. But while the array size is of type size_t, the function in fact returns an int that is supposed to index into that array.

Fix the function signature to return a size_t. This conversion does not change any semantics given that the function would only ever return a value in the range [0, sz] anyway.

---
 reftable/basics.c      |  2 +-
 reftable/basics.h      |  2 +-
 reftable/basics_test.c |  6 +++---
 reftable/block.c       |  3 ++-
 reftable/refname.c     | 17 +++++++----------
 5 files changed, 14 insertions(+), 16 deletions(-)
reftable/basics.c +1 -1
index 0785aff941..2c5f34b39e 100644
@@ -27,7 +27,7 @@ void put_be16(uint8_t *out, uint16_t i)
 	out[1] = (uint8_t)(i & 0xff);
 }
 
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 {
 	size_t lo = 0;
 	size_t hi = sz;
reftable/basics.h +1 -1
index 91f3533efe..2672520e76 100644
@@ -28,7 +28,7 @@ void put_be16(uint8_t *out, uint16_t i);
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
  */
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
 
 /*
  * Frees a NULL terminated array of malloced strings. The array itself is also
reftable/basics_test.c +3 -3
index 1fcd229725..dc1c87c5df 100644
@@ -34,15 +34,15 @@ static void test_binsearch(void)
 
 	int i = 0;
 	for (i = 1; i < 11; i++) {
-		int res;
+		size_t res;
+
 		args.key = i;
 		res = binsearch(sz, &binsearch_func, &args);
 
 		if (res < sz) {
 			EXPECT(args.key < arr[res]);
-			if (res > 0) {
+			if (res > 0)
 				EXPECT(args.key >= arr[res - 1]);
-			}
 		} else {
 			EXPECT(args.key == 10 || args.key == 11);
 		}
reftable/block.c +2 -1
index e2a2cee58d..422885bddb 100644
@@ -382,7 +382,8 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
-	int err = 0, i;
+	int err = 0;
+	size_t i;
 
 	if (args.error) {
 		err = REFTABLE_FORMAT_ERROR;
reftable/refname.c +7 -10
index 7570e4acf9..64eba1b886 100644
@@ -33,10 +33,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->add,
 			.want = name,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
-		if (idx < mod->add_len && !strcmp(mod->add[idx], name)) {
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
-		}
 	}
 
 	if (mod->del_len > 0) {
@@ -44,10 +43,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->del,
 			.want = name,
 		};
-		int idx = binsearch(mod->del_len, find_name, &arg);
-		if (idx < mod->del_len && !strcmp(mod->del[idx], name)) {
+		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
-		}
 	}
 
 	err = reftable_table_read_ref(&mod->tab, name, &ref);
@@ -77,7 +75,7 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			.names = mod->add,
 			.want = prefix,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -96,11 +94,10 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 				.names = mod->del,
 				.want = ref.refname,
 			};
-			int idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, find_name, &arg);
 			if (idx < mod->del_len &&
-			    !strcmp(ref.refname, mod->del[idx])) {
+			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
-			}
 		}
 
 		if (strncmp(ref.refname, prefix, strlen(prefix))) {
Signature · 1 line
2.44.0
#2 lore raw

PATCH 2/7 reftable/basics: improve `binsearch()` test

In reply to Patrick Steinhardt

The binsearch() test is somewhat weird in that it doesn't explicitly spell out its expectations. Instead it does so in a rather ad-hoc way with some hard-to-understand computations.

Refactor the test to spell out the needle as well as expected index for all testcases. This refactoring highlights that the binsearch_func() is written somewhat weirdly to find the first integer smaller than the needle, not smaller or equal to it. Adjust the function accordingly.

While at it, rename the callback function to better convey its meaning.

---
 reftable/basics_test.c | 55 ++++++++++++++++++++++++------------------
 1 file changed, 31 insertions(+), 24 deletions(-)
reftable/basics_test.c +31 -24
index dc1c87c5df..85c4d1621c 100644
@@ -12,40 +12,47 @@ license that can be found in the LICENSE file or at
 #include "test_framework.h"
 #include "reftable-tests.h"
 
-struct binsearch_args {
-	int key;
-	int *arr;
+struct integer_needle_lesseq {
+	int needle;
+	int *haystack;
 };
 
-static int binsearch_func(size_t i, void *void_args)
+static int integer_needle_lesseq(size_t i, void *_args)
 {
-	struct binsearch_args *args = void_args;
-
-	return args->key < args->arr[i];
+	struct integer_needle_lesseq *args = _args;
+	return args->needle <= args->haystack[i];
 }
 
 static void test_binsearch(void)
 {
-	int arr[] = { 2, 4, 6, 8, 10 };
-	size_t sz = ARRAY_SIZE(arr);
-	struct binsearch_args args = {
-		.arr = arr,
+	int haystack[] = { 2, 4, 6, 8, 10 };
+	struct {
+		int needle;
+		size_t expected_idx;
+	} testcases[] = {
+		{-9000, 0},
+		{-1, 0},
+		{0, 0},
+		{2, 0},
+		{3, 1},
+		{4, 1},
+		{7, 3},
+		{9, 4},
+		{10, 4},
+		{11, 5},
+		{9000, 5},
 	};
+	size_t i = 0;
 
-	int i = 0;
-	for (i = 1; i < 11; i++) {
-		size_t res;
-
-		args.key = i;
-		res = binsearch(sz, &binsearch_func, &args);
+	for (i = 0; i < ARRAY_SIZE(testcases); i++) {
+		struct integer_needle_lesseq args = {
+			.haystack = haystack,
+			.needle = testcases[i].needle,
+		};
+		size_t idx;
 
-		if (res < sz) {
-			EXPECT(args.key < arr[res]);
-			if (res > 0)
-				EXPECT(args.key >= arr[res - 1]);
-		} else {
-			EXPECT(args.key == 10 || args.key == 11);
-		}
+		idx = binsearch(ARRAY_SIZE(haystack), &integer_needle_lesseq, &args);
+		EXPECT(idx == testcases[i].expected_idx);
 	}
 }
 
Signature · 1 line
2.44.0
#3 lore raw

Re: [PATCH 2/7] reftable/basics: improve `binsearch()` test

Quoted 21 lines · Patrick Steinhardt

The binsearch() test is somewhat weird in that it doesn't explicitly spell out its expectations. Instead it does so in a rather ad-hoc way with some hard-to-understand computations.

Refactor the test to spell out the needle as well as expected index for all testcases. This refactoring highlights that the binsearch_func() is written somewhat weirdly to find the first integer smaller than the needle, not smaller or equal to it. Adjust the function accordingly.

While at it, rename the callback function to better convey its meaning.

---
 reftable/basics_test.c | 55 ++++++++++++++++++++++++------------------
 1 file changed, 31 insertions(+), 24 deletions(-)
reftable/basics_test.c +0 -0
index dc1c87c5df..85c4d1621c 100644

reftable/basics_test.c license that can be found in the LICENSE file or at · line 18

Quoted 10 lines
reftable/basics_test.c +3 -3
@@ -12,40 +12,47 @@ license that can be found in the LICENSE file or at
 #include "test_framework.h"
 #include "reftable-tests.h"
 
-struct binsearch_args {
-	int key;
-	int *arr;
+struct integer_needle_lesseq {
+	int needle;
+	int *haystack;
 };

This is probably just personal preference, but I think key and arr in this case are a bit more straightforward. I do like that we rename the args to be more specific. Do we want to also append _args to denote that it is an argument set? Maybe integer_lesseq_args?

reftable/basics_test.c license that can be found in the LICENSE file or at · line 24

Quoted 10 lines
reftable/basics_test.c +3 -4
-static int binsearch_func(size_t i, void *void_args)
+static int integer_needle_lesseq(size_t i, void *_args)
 {
-	struct binsearch_args *args = void_args;
-
-	return args->key < args->arr[i];
+	struct integer_needle_lesseq *args = _args;
+	return args->needle <= args->haystack[i];
 }

-Justin

#4 lore raw

Re: [PATCH 2/7] reftable/basics: improve `binsearch()` test

Quoted 25 lines · Justin Tobler
Quoted 24 lines · Patrick Steinhardt

The binsearch() test is somewhat weird in that it doesn't explicitly spell out its expectations. Instead it does so in a rather ad-hoc way with some hard-to-understand computations.

Refactor the test to spell out the needle as well as expected index for all testcases. This refactoring highlights that the binsearch_func() is written somewhat weirdly to find the first integer smaller than the needle, not smaller or equal to it. Adjust the function accordingly.

While at it, rename the callback function to better convey its meaning.

---
 reftable/basics_test.c | 55 ++++++++++++++++++++++++------------------
 1 file changed, 31 insertions(+), 24 deletions(-)
reftable/basics_test.c +0 -0
index dc1c87c5df..85c4d1621c 100644
@@ -12,40 +12,47 @@ license that can be found in the LICENSE file or at
 #include "test_framework.h"
 #include "reftable-tests.h"
 
Quoted 10 lines

reftable/basics_test.c license that can be found in the LICENSE file or at · line 18

Quoted 7 lines
reftable/basics_test.c +3 -3
-struct binsearch_args {
-	int key;
-	int *arr;
+struct integer_needle_lesseq {
+	int needle;
+	int *haystack;
 };

This is probably just personal preference, but I think key and arr in this case are a bit more straightforward.

I was trying to make this consistent across all the callsites, and here I personally think that haystack and needle are well understood by most folks and generic enough.

Quoted 3 lines

I do like that we rename
the args to be more specific. Do we want to also append _args to denote that it is an argument set? Maybe integer_lesseq_args?

Oh, yeah, that's an oversight indeed.

Patrick

#5 lore raw

PATCH 3/7 reftable/refname: refactor binary search over refnames

In reply to Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +22 -22
index 64eba1b886..9ec488d727 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };
 
-static int find_name(size_t k, void *arg)
+static int refname_needle_lesseq(size_t k, void *arg)
 {
-	struct find_arg *f_arg = arg;
-	return strcmp(f_arg->names[k], f_arg->want) >= 0;
+	struct refname_needle_lesseq_args *f_arg = arg;
+	return strcmp(f_arg->needle, f_arg->haystack[k]) <= 0;
 }
 
 static int modification_has_ref(struct modification *mod, const char *name)
@@ -29,21 +29,21 @@ static int modification_has_ref(struct modification *mod, const char *name)
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
 	}
 
 	if (mod->del_len > 0) {
-		struct find_arg arg = {
-			.names = mod->del,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->del,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
 	}
@@ -71,11 +71,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = prefix,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = prefix,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -90,11 +90,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			goto done;
 
 		if (mod->del_len > 0) {
-			struct find_arg arg = {
-				.names = mod->del,
-				.want = ref.refname,
+			struct refname_needle_lesseq_args arg = {
+				.haystack = mod->del,
+				.needle = ref.refname,
 			};
-			size_t idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 			if (idx < mod->del_len &&
 			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
Signature · 1 line
2.44.0
#6 lore raw

Re: [PATCH 3/7] reftable/refname: refactor binary search over refnames

Quoted 15 lines · Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +0 -0
index 64eba1b886..9ec488d727 100644

reftable/refname.c line 18

Quoted 10 lines
reftable/refname.c +3 -3
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };

I agree that the previous names and want are a bit ambiguous. What do you think about refnames and target_refname instead?

-Justin

#7 lore raw

Re: [PATCH 3/7] reftable/refname: refactor binary search over refnames

Quoted 19 lines · Justin Tobler
Quoted 18 lines · Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +0 -0
index 64eba1b886..9ec488d727 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
Quoted 10 lines

reftable/refname.c line 18

Quoted 7 lines
reftable/refname.c +3 -3
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };

I agree that the previous names and want are a bit ambiguous. What do you think about refnames and target_refname instead?

As said in the preceding commit I was rather aiming for consistency across the callsites, so I'll keep this as-is for now. I'm happy to be overruled though if others feel the same way.

Patrick

#8 lore raw

PATCH 4/7 reftable/block: refactor binary search over restart points

In reply to Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---
 reftable/block.c | 97 ++++++++++++++++++++++++++++++++++--------------
 1 file changed, 70 insertions(+), 27 deletions(-)
reftable/block.c +70 -27
index 422885bddb..6cd4c053df 100644
@@ -273,34 +273,36 @@ void block_reader_start(struct block_reader *br, struct block_iter *it)
 	it->next_off = br->header_off + 4;
 }
 
-struct restart_find_args {
+struct restart_needle_less_args {
 	int error;
-	struct strbuf key;
-	struct block_reader *r;
+	struct strbuf needle;
+	struct block_reader *reader;
 };
 
-static int restart_key_less(size_t idx, void *args)
+static int restart_needle_less(size_t idx, void *_args)
 {
-	struct restart_find_args *a = args;
-	uint32_t off = block_reader_restart_offset(a->r, idx);
+	struct restart_needle_less_args *args = _args;
+	uint32_t off = block_reader_restart_offset(args->reader, idx);
 	struct string_view in = {
-		.buf = a->r->block.data + off,
-		.len = a->r->block_len - off,
+		.buf = args->reader->block.data + off,
+		.len = args->reader->block_len - off,
 	};
-
-	/* the restart key is verbatim in the block, so this could avoid the
-	   alloc for decoding the key */
-	struct strbuf rkey = STRBUF_INIT;
+	struct strbuf kth_restart_key = STRBUF_INIT;
 	uint8_t unused_extra;
-	int n = reftable_decode_key(&rkey, &unused_extra, in);
-	int result;
+	int result, n;
+
+	/*
+	 * TODO: The restart key is verbatim in the block, so we can in theory
+	 * avoid decoding the key and thus save some allocations.
+	 */
+	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
 	if (n < 0) {
-		a->error = 1;
+		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&a->key, &rkey);
-	strbuf_release(&rkey);
+	result = strbuf_cmp(&args->needle, &kth_restart_key);
+	strbuf_release(&kth_restart_key);
 	return result < 0;
 }
 
@@ -376,9 +378,9 @@ void block_iter_close(struct block_iter *it)
 int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		      struct strbuf *want)
 {
-	struct restart_find_args args = {
-		.key = *want,
-		.r = br,
+	struct restart_needle_less_args args = {
+		.needle = *want,
+		.reader = br,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
@@ -390,7 +392,35 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		goto done;
 	}
 
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.
+	 *
+	 *   - `i == restart_count`: The wanted record was not found at any of
+	 *     the restart points. As there is no restart point at the end of
+	 *     the section the record may thus be contained in the last block.
+	 *
+	 *   - `i > 0`: The wanted record must be contained in the section
+	 *     before the found restart point. We thus do a linear search
+	 *     starting from the preceding restart point.
+	 */
 	if (i > 0)
 		it->next_off = block_reader_restart_offset(br, i - 1);
 	else
@@ -399,21 +429,34 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 
 	reftable_record_init(&rec, block_reader_type(br));
 
-	/* We're looking for the last entry less/equal than the wanted key, so
-	   we have to go one entry too far and then back up.
-	*/
+	/*
+	 * We're looking for the last entry less than the wanted key so that
+	 * the next call to `block_reader_next()` would yield the wanted
+	 * record. We thus don't want to position our reader at the sought
+	 * after record, but one before. To do so, we have to go one entry too
+	 * far and then back up.
+	 */
 	while (1) {
 		block_iter_copy_from(&next, it);
 		err = block_iter_next(&next, &rec);
 		if (err < 0)
 			goto done;
-
-		reftable_record_key(&rec, &it->last_key);
-		if (err > 0 || strbuf_cmp(&it->last_key, want) >= 0) {
+		if (err > 0) {
 			err = 0;
 			goto done;
 		}
 
+		/*
+		 * Check whether the current key is greater or equal to the
+		 * sought-after key. In case it is greater we know that the
+		 * record does not exist in the block and can thus abort early.
+		 * In case it is equal to the sought-after key we have found
+		 * the desired record.
+		 */
+		reftable_record_key(&rec, &it->last_key);
+		if (strbuf_cmp(&it->last_key, want) >= 0)
+			goto done;
+
 		block_iter_copy_from(it, &next);
 	}
 
Signature · 1 line
2.44.0
#9 lore raw

PATCH 5/7 reftable/block: fix error handling when searching restart points

In reply to Patrick Steinhardt

When doing the binary search over restart points in a block we need to decode the record keys. This decoding step can result in an error when the block is corrupted, which we indicate to the caller of the binary search by setting args.error = 1. But the only caller that exists mishandles this because it in fact performs the error check before calling binsearch().

Fix this bug by checking for errors at the right point in time. Furthermore, refactor binsearch() so that it aborts the search in case the callback function returns a negative value so that we don't needlessly continue to search the block.

---
 reftable/basics.c | 5 ++++-
 reftable/basics.h | 5 +++--
 reftable/block.c  | 9 ++++-----
 3 files changed, 11 insertions(+), 8 deletions(-)
reftable/basics.c +4 -1
index 2c5f34b39e..fea711db7e 100644
@@ -39,8 +39,11 @@ size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 	 */
 	while (hi - lo > 1) {
 		size_t mid = lo + (hi - lo) / 2;
+		int ret = f(mid, args);
+		if (ret < 0)
+			return sz;
 
-		if (f(mid, args))
+		if (ret > 0)
 			hi = mid;
 		else
 			lo = mid;
reftable/basics.h +3 -2
index 2672520e76..523ecd5307 100644
@@ -22,8 +22,9 @@ uint32_t get_be24(uint8_t *in);
 void put_be16(uint8_t *out, uint16_t i);
 
 /*
- * find smallest index i in [0, sz) at which f(i) is true, assuming
- * that f is ascending. Return sz if f(i) is false for all indices.
+ * find smallest index i in [0, sz) at which `f(i) > 0`, assuming that f is
+ * ascending. Return sz if `f(i) == 0` for all indices. The search is aborted
+ * and `sz` is returned in case `f(i) < 0`.
  *
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
reftable/block.c +4 -5
index 6cd4c053df..ca80a05e21 100644
@@ -387,11 +387,6 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	int err = 0;
 	size_t i;
 
-	if (args.error) {
-		err = REFTABLE_FORMAT_ERROR;
-		goto done;
-	}
-
 	/*
 	 * Perform a binary search over the block's restart points, which
 	 * avoids doing a linear scan over the whole block. Like this, we
@@ -405,6 +400,10 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	 * too many record.
 	 */
 	i = binsearch(br->restart_count, &restart_needle_less, &args);
+	if (args.error) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
 
 	/*
 	 * Now there are multiple cases:
Signature · 1 line
2.44.0
#10 lore raw

PATCH 6/7 reftable/record: extract function to decode key lengths

In reply to Patrick Steinhardt

We're about to refactor the binary search over restart points so that it does not need to fully decode the record keys anymore. To do so we will need to decode the record key lengths, which is non-trivial logic.

Extract the logic to decode these lengths from refatble_decode_key() so that we can reuse it.

---
 reftable/record.c | 34 +++++++++++++++++++++++++---------
 reftable/record.h |  6 ++++++
 2 files changed, 31 insertions(+), 9 deletions(-)
reftable/record.c +25 -9
index 23b497adab..5506f3e913 100644
@@ -159,26 +159,42 @@ int reftable_encode_key(int *restart, struct string_view dest,
 	return start.len - dest.len;
 }
 
-int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
-			struct string_view in)
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra)
 {
-	int start_len = in.len;
-	uint64_t prefix_len = 0;
-	uint64_t suffix_len = 0;
+	size_t start_len = in.len;
 	int n;
 
-	n = get_var_int(&prefix_len, &in);
+	n = get_var_int(prefix_len, &in);
 	if (n < 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	n = get_var_int(&suffix_len, &in);
+	n = get_var_int(suffix_len, &in);
 	if (n <= 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	*extra = (uint8_t)(suffix_len & 0x7);
-	suffix_len >>= 3;
+	*extra = (uint8_t)(*suffix_len & 0x7);
+	*suffix_len >>= 3;
+
+	return start_len - in.len;
+}
+
+int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
+			struct string_view in)
+{
+	int start_len = in.len;
+	uint64_t prefix_len = 0;
+	uint64_t suffix_len = 0;
+	int n;
+
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, extra);
+	if (n < 0)
+		return -1;
+	string_view_consume(&in, n);
 
 	if (in.len < suffix_len ||
 	    prefix_len > last_key->len)
reftable/record.h +6 -0
index 826ee1c55c..d778133e6e 100644
@@ -86,6 +86,12 @@ int reftable_encode_key(int *is_restart, struct string_view dest,
 			struct strbuf prev_key, struct strbuf key,
 			uint8_t extra);
 
+/* Decode a record's key lengths. */
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra);
+
 /*
  * Decode into `last_key` and `extra` from `in`. `last_key` is expected to
  * contain the decoded key of the preceding record, if any.
Signature · 1 line
2.44.0
#11 lore raw

PATCH 7/7 reftable/block: avoid decoding keys when searching restart points

In reply to Patrick Steinhardt

When searching over restart points in a block we decode the key of each of the records, which results in a memory allocation. This is quite pointless though given that records it restart points will never use prefix compression and thus store their keys verbatim in the block.

Refactor the code so that we can avoid decoding the keys, which saves us some allocations.

---
 reftable/block.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)
reftable/block.c +19 -10
index ca80a05e21..8bb4e43cec 100644
@@ -287,23 +287,32 @@ static int restart_needle_less(size_t idx, void *_args)
 		.buf = args->reader->block.data + off,
 		.len = args->reader->block_len - off,
 	};
-	struct strbuf kth_restart_key = STRBUF_INIT;
-	uint8_t unused_extra;
-	int result, n;
+	uint64_t prefix_len, suffix_len;
+	uint8_t extra;
+	int n;
 
 	/*
-	 * TODO: The restart key is verbatim in the block, so we can in theory
-	 * avoid decoding the key and thus save some allocations.
+	 * Records at restart points are stored without prefix compression, so
+	 * there is no need to fully decode the record key here. This removes
+	 * the need for allocating memory.
 	 */
-	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
-	if (n < 0) {
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, &extra);
+	if (n < 0 || prefix_len) {
 		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&args->needle, &kth_restart_key);
-	strbuf_release(&kth_restart_key);
-	return result < 0;
+	string_view_consume(&in, n);
+	if (suffix_len > in.len) {
+		args->error = 1;
+		return -1;
+	}
+
+	n = memcmp(args->needle.buf, in.buf,
+		   args->needle.len < suffix_len ? args->needle.len : suffix_len);
+	if (n)
+		return n < 0;
+	return args->needle.len < suffix_len;
 }
 
 void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
Signature · 1 line
2.44.0
#12 lore raw

v2 cover reftable: improvements for the `binsearch()` mechanism

In reply to Patrick Steinhardt

Hi,

this is the second version of my patch series that refactors the binsearch() mechanism in the reftable library. There's only a single change compared to v1, namely a renamed struct.

Thanks!

Patrick

Patrick Steinhardt (7):
reftable/basics: fix return type of binsearch() to be size_t reftable/basics: improve binsearch() test reftable/refname: refactor binary search over refnames reftable/block: refactor binary search over restart points reftable/block: fix error handling when searching restart points reftable/record: extract function to decode key lengths reftable/block: avoid decoding keys when searching restart points

 reftable/basics.c      |   7 ++-
 reftable/basics.h      |   7 +--
 reftable/basics_test.c |  55 +++++++++++---------
 reftable/block.c       | 114 ++++++++++++++++++++++++++++++-----------
 reftable/record.c      |  34 ++++++++----
 reftable/record.h      |   6 +++
 reftable/refname.c     |  53 +++++++++----------
 7 files changed, 179 insertions(+), 97 deletions(-)

Range-diff against v1:

1:  cd82ac6531 = 1:  cd82ac6531 reftable/basics: fix return type of `binsearch()` to be `size_t`
2:  7955f7983a ! 2:  a277d4fa6f reftable/basics: improve `binsearch()` test
    @@ reftable/basics_test.c: license that can be found in the LICENSE file or at
     -struct binsearch_args {
     -	int key;
     -	int *arr;
    -+struct integer_needle_lesseq {
    ++struct integer_needle_lesseq_args {
     +	int needle;
     +	int *haystack;
      };
    @@ reftable/basics_test.c: license that can be found in the LICENSE file or at
     -	struct binsearch_args *args = void_args;
     -
     -	return args->key < args->arr[i];
    -+	struct integer_needle_lesseq *args = _args;
    ++	struct integer_needle_lesseq_args *args = _args;
     +	return args->needle <= args->haystack[i];
      }
      
    @@ reftable/basics_test.c: license that can be found in the LICENSE file or at
     -		args.key = i;
     -		res = binsearch(sz, &binsearch_func, &args);
     +	for (i = 0; i < ARRAY_SIZE(testcases); i++) {
    -+		struct integer_needle_lesseq args = {
    ++		struct integer_needle_lesseq_args args = {
     +			.haystack = haystack,
     +			.needle = testcases[i].needle,
     +		};
3:  44386818ce = 3:  9ffcf45c32 reftable/refname: refactor binary search over refnames
4:  f56275f288 = 4:  5e20d93ae0 reftable/block: refactor binary search over restart points
5:  36b1ef8e5c = 5:  5bbeab114f reftable/block: fix error handling when searching restart points
6:  38666de451 = 6:  271bacb210 reftable/record: extract function to decode key lengths
7:  f716400686 = 7:  e751b3c536 reftable/block: avoid decoding keys when searching restart points
Signature · 1 line
2.44.GIT
#13 lore raw

PATCH v2 1/7 reftable/basics: fix return type of `binsearch()` to be `size_t`

The binsearch() function can be used to find the first element for which a callback functions returns a truish value. But while the array size is of type size_t, the function in fact returns an int that is supposed to index into that array.

Fix the function signature to return a size_t. This conversion does not change any semantics given that the function would only ever return a value in the range [0, sz] anyway.

---
 reftable/basics.c      |  2 +-
 reftable/basics.h      |  2 +-
 reftable/basics_test.c |  6 +++---
 reftable/block.c       |  3 ++-
 reftable/refname.c     | 17 +++++++----------
 5 files changed, 14 insertions(+), 16 deletions(-)
reftable/basics.c +1 -1
index 0785aff941..2c5f34b39e 100644
@@ -27,7 +27,7 @@ void put_be16(uint8_t *out, uint16_t i)
 	out[1] = (uint8_t)(i & 0xff);
 }
 
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 {
 	size_t lo = 0;
 	size_t hi = sz;
reftable/basics.h +1 -1
index 91f3533efe..2672520e76 100644
@@ -28,7 +28,7 @@ void put_be16(uint8_t *out, uint16_t i);
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
  */
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
 
 /*
  * Frees a NULL terminated array of malloced strings. The array itself is also
reftable/basics_test.c +3 -3
index 1fcd229725..dc1c87c5df 100644
@@ -34,15 +34,15 @@ static void test_binsearch(void)
 
 	int i = 0;
 	for (i = 1; i < 11; i++) {
-		int res;
+		size_t res;
+
 		args.key = i;
 		res = binsearch(sz, &binsearch_func, &args);
 
 		if (res < sz) {
 			EXPECT(args.key < arr[res]);
-			if (res > 0) {
+			if (res > 0)
 				EXPECT(args.key >= arr[res - 1]);
-			}
 		} else {
 			EXPECT(args.key == 10 || args.key == 11);
 		}
reftable/block.c +2 -1
index e2a2cee58d..422885bddb 100644
@@ -382,7 +382,8 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
-	int err = 0, i;
+	int err = 0;
+	size_t i;
 
 	if (args.error) {
 		err = REFTABLE_FORMAT_ERROR;
reftable/refname.c +7 -10
index 7570e4acf9..64eba1b886 100644
@@ -33,10 +33,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->add,
 			.want = name,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
-		if (idx < mod->add_len && !strcmp(mod->add[idx], name)) {
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
-		}
 	}
 
 	if (mod->del_len > 0) {
@@ -44,10 +43,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->del,
 			.want = name,
 		};
-		int idx = binsearch(mod->del_len, find_name, &arg);
-		if (idx < mod->del_len && !strcmp(mod->del[idx], name)) {
+		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
-		}
 	}
 
 	err = reftable_table_read_ref(&mod->tab, name, &ref);
@@ -77,7 +75,7 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			.names = mod->add,
 			.want = prefix,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -96,11 +94,10 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 				.names = mod->del,
 				.want = ref.refname,
 			};
-			int idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, find_name, &arg);
 			if (idx < mod->del_len &&
-			    !strcmp(ref.refname, mod->del[idx])) {
+			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
-			}
 		}
 
 		if (strncmp(ref.refname, prefix, strlen(prefix))) {
Signature · 1 line
2.44.GIT
#14 lore raw

PATCH v2 2/7 reftable/basics: improve `binsearch()` test

In reply to Patrick Steinhardt

The binsearch() test is somewhat weird in that it doesn't explicitly spell out its expectations. Instead it does so in a rather ad-hoc way with some hard-to-understand computations.

Refactor the test to spell out the needle as well as expected index for all testcases. This refactoring highlights that the binsearch_func() is written somewhat weirdly to find the first integer smaller than the needle, not smaller or equal to it. Adjust the function accordingly.

While at it, rename the callback function to better convey its meaning.

---
 reftable/basics_test.c | 55 ++++++++++++++++++++++++------------------
 1 file changed, 31 insertions(+), 24 deletions(-)
reftable/basics_test.c +31 -24
index dc1c87c5df..997c4d9e01 100644
@@ -12,40 +12,47 @@ license that can be found in the LICENSE file or at
 #include "test_framework.h"
 #include "reftable-tests.h"
 
-struct binsearch_args {
-	int key;
-	int *arr;
+struct integer_needle_lesseq_args {
+	int needle;
+	int *haystack;
 };
 
-static int binsearch_func(size_t i, void *void_args)
+static int integer_needle_lesseq(size_t i, void *_args)
 {
-	struct binsearch_args *args = void_args;
-
-	return args->key < args->arr[i];
+	struct integer_needle_lesseq_args *args = _args;
+	return args->needle <= args->haystack[i];
 }
 
 static void test_binsearch(void)
 {
-	int arr[] = { 2, 4, 6, 8, 10 };
-	size_t sz = ARRAY_SIZE(arr);
-	struct binsearch_args args = {
-		.arr = arr,
+	int haystack[] = { 2, 4, 6, 8, 10 };
+	struct {
+		int needle;
+		size_t expected_idx;
+	} testcases[] = {
+		{-9000, 0},
+		{-1, 0},
+		{0, 0},
+		{2, 0},
+		{3, 1},
+		{4, 1},
+		{7, 3},
+		{9, 4},
+		{10, 4},
+		{11, 5},
+		{9000, 5},
 	};
+	size_t i = 0;
 
-	int i = 0;
-	for (i = 1; i < 11; i++) {
-		size_t res;
-
-		args.key = i;
-		res = binsearch(sz, &binsearch_func, &args);
+	for (i = 0; i < ARRAY_SIZE(testcases); i++) {
+		struct integer_needle_lesseq_args args = {
+			.haystack = haystack,
+			.needle = testcases[i].needle,
+		};
+		size_t idx;
 
-		if (res < sz) {
-			EXPECT(args.key < arr[res]);
-			if (res > 0)
-				EXPECT(args.key >= arr[res - 1]);
-		} else {
-			EXPECT(args.key == 10 || args.key == 11);
-		}
+		idx = binsearch(ARRAY_SIZE(haystack), &integer_needle_lesseq, &args);
+		EXPECT(idx == testcases[i].expected_idx);
 	}
 }
 
Signature · 1 line
2.44.GIT
#15 lore raw

PATCH v2 3/7 reftable/refname: refactor binary search over refnames

In reply to Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +22 -22
index 64eba1b886..9ec488d727 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };
 
-static int find_name(size_t k, void *arg)
+static int refname_needle_lesseq(size_t k, void *arg)
 {
-	struct find_arg *f_arg = arg;
-	return strcmp(f_arg->names[k], f_arg->want) >= 0;
+	struct refname_needle_lesseq_args *f_arg = arg;
+	return strcmp(f_arg->needle, f_arg->haystack[k]) <= 0;
 }
 
 static int modification_has_ref(struct modification *mod, const char *name)
@@ -29,21 +29,21 @@ static int modification_has_ref(struct modification *mod, const char *name)
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
 	}
 
 	if (mod->del_len > 0) {
-		struct find_arg arg = {
-			.names = mod->del,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->del,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
 	}
@@ -71,11 +71,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = prefix,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = prefix,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -90,11 +90,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			goto done;
 
 		if (mod->del_len > 0) {
-			struct find_arg arg = {
-				.names = mod->del,
-				.want = ref.refname,
+			struct refname_needle_lesseq_args arg = {
+				.haystack = mod->del,
+				.needle = ref.refname,
 			};
-			size_t idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 			if (idx < mod->del_len &&
 			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
Signature · 1 line
2.44.GIT
#16 lore raw

Re: [PATCH v2 3/7] reftable/refname: refactor binary search over refnames

Quoted 22 lines · Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +1 -3
index 64eba1b886..9ec488d727 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {

reftable/refname.c line 22

Quoted 10 lines
reftable/refname.c +4 -3
+	char **haystack;
+	const char *needle;
 };
 
-static int find_name(size_t k, void *arg)
+static int refname_needle_lesseq(size_t k, void *arg)
 {
-	struct find_arg *f_arg = arg;
-	return strcmp(f_arg->names[k], f_arg->want) >= 0;
+	struct refname_needle_lesseq_args *f_arg = arg;

nit: Looks like the f_arg variable name is a remnant from when the type was called find_arg. Do we want to rename this to args? We could also rename void *arg to void *_args.

Quoted 16 lines
reftable/refname.c +4 -4
+	return strcmp(f_arg->needle, f_arg->haystack[k]) <= 0;
 }
 
 static int modification_has_ref(struct modification *mod, const char *name)
@@ -29,21 +29,21 @@ static int modification_has_ref(struct modification *mod, const char *name)
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);

reftable/refname.c static int modification_has_ref(struct modification *mod, const char *name) · line 42

Quoted 10 lines
reftable/refname.c +2 -3
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
 	}
 
 	if (mod->del_len > 0) {
-		struct find_arg arg = {
-			.names = mod->del,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {

nit: In the other commits we use args. Do we want to be consistent?

-Justin

Quoted 43 lines
reftable/refname.c +11 -9
+			.haystack = mod->del,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
 	}
@@ -71,11 +71,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = prefix,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = prefix,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -90,11 +90,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			goto done;
 
 		if (mod->del_len > 0) {
-			struct find_arg arg = {
-				.names = mod->del,
-				.want = ref.refname,
+			struct refname_needle_lesseq_args arg = {
+				.haystack = mod->del,
+				.needle = ref.refname,
 			};
-			size_t idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 			if (idx < mod->del_len &&
 			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
Signature · 1 line
2.44.GIT
#17 lore raw

Re: [PATCH v2 3/7] reftable/refname: refactor binary search over refnames

Quoted 27 lines · Justin Tobler
Quoted 26 lines · Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +3 -3
index 64eba1b886..9ec488d727 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };
 
Quoted 10 lines

reftable/refname.c line 22

Quoted 6 lines
reftable/refname.c +2 -3
-static int find_name(size_t k, void *arg)
+static int refname_needle_lesseq(size_t k, void *arg)
 {
-	struct find_arg *f_arg = arg;
-	return strcmp(f_arg->names[k], f_arg->want) >= 0;
+	struct refname_needle_lesseq_args *f_arg = arg;

nit: Looks like the f_arg variable name is a remnant from when the type was called find_arg. Do we want to rename this to args? We could also rename void *arg to void *_args.

Yup.

Quoted 21 lines
Quoted 21 lines
reftable/refname.c +5 -4
+	return strcmp(f_arg->needle, f_arg->haystack[k]) <= 0;
 }
 
 static int modification_has_ref(struct modification *mod, const char *name)
@@ -29,21 +29,21 @@ static int modification_has_ref(struct modification *mod, const char *name)
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
 	}
 
Quoted 7 lines

reftable/refname.c static int modification_has_ref(struct modification *mod, const char *name) · line 42

Quoted 5 lines
reftable/refname.c +1 -3
 	if (mod->del_len > 0) {
-		struct find_arg arg = {
-			.names = mod->del,
-			.want = name,
+		struct refname_needle_lesseq_args arg = {

nit: In the other commits we use args. Do we want to be consistent?

Yeah. Given that it was my own argument to be consistent across callsites I should rather make it truly consistent :)

Patrick

Quoted 47 lines

-Justin

Quoted 43 lines
reftable/refname.c +11 -9
+			.haystack = mod->del,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
 	}
@@ -71,11 +71,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = prefix,
+		struct refname_needle_lesseq_args arg = {
+			.haystack = mod->add,
+			.needle = prefix,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -90,11 +90,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			goto done;
 
 		if (mod->del_len > 0) {
-			struct find_arg arg = {
-				.names = mod->del,
-				.want = ref.refname,
+			struct refname_needle_lesseq_args arg = {
+				.haystack = mod->del,
+				.needle = ref.refname,
 			};
-			size_t idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
 			if (idx < mod->del_len &&
 			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
Signature · 1 line
2.44.GIT
#18 lore raw

PATCH v2 4/7 reftable/block: refactor binary search over restart points

In reply to Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---
 reftable/block.c | 97 ++++++++++++++++++++++++++++++++++--------------
 1 file changed, 70 insertions(+), 27 deletions(-)
reftable/block.c +70 -27
index 422885bddb..6cd4c053df 100644
@@ -273,34 +273,36 @@ void block_reader_start(struct block_reader *br, struct block_iter *it)
 	it->next_off = br->header_off + 4;
 }
 
-struct restart_find_args {
+struct restart_needle_less_args {
 	int error;
-	struct strbuf key;
-	struct block_reader *r;
+	struct strbuf needle;
+	struct block_reader *reader;
 };
 
-static int restart_key_less(size_t idx, void *args)
+static int restart_needle_less(size_t idx, void *_args)
 {
-	struct restart_find_args *a = args;
-	uint32_t off = block_reader_restart_offset(a->r, idx);
+	struct restart_needle_less_args *args = _args;
+	uint32_t off = block_reader_restart_offset(args->reader, idx);
 	struct string_view in = {
-		.buf = a->r->block.data + off,
-		.len = a->r->block_len - off,
+		.buf = args->reader->block.data + off,
+		.len = args->reader->block_len - off,
 	};
-
-	/* the restart key is verbatim in the block, so this could avoid the
-	   alloc for decoding the key */
-	struct strbuf rkey = STRBUF_INIT;
+	struct strbuf kth_restart_key = STRBUF_INIT;
 	uint8_t unused_extra;
-	int n = reftable_decode_key(&rkey, &unused_extra, in);
-	int result;
+	int result, n;
+
+	/*
+	 * TODO: The restart key is verbatim in the block, so we can in theory
+	 * avoid decoding the key and thus save some allocations.
+	 */
+	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
 	if (n < 0) {
-		a->error = 1;
+		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&a->key, &rkey);
-	strbuf_release(&rkey);
+	result = strbuf_cmp(&args->needle, &kth_restart_key);
+	strbuf_release(&kth_restart_key);
 	return result < 0;
 }
 
@@ -376,9 +378,9 @@ void block_iter_close(struct block_iter *it)
 int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		      struct strbuf *want)
 {
-	struct restart_find_args args = {
-		.key = *want,
-		.r = br,
+	struct restart_needle_less_args args = {
+		.needle = *want,
+		.reader = br,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
@@ -390,7 +392,35 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		goto done;
 	}
 
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.
+	 *
+	 *   - `i == restart_count`: The wanted record was not found at any of
+	 *     the restart points. As there is no restart point at the end of
+	 *     the section the record may thus be contained in the last block.
+	 *
+	 *   - `i > 0`: The wanted record must be contained in the section
+	 *     before the found restart point. We thus do a linear search
+	 *     starting from the preceding restart point.
+	 */
 	if (i > 0)
 		it->next_off = block_reader_restart_offset(br, i - 1);
 	else
@@ -399,21 +429,34 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 
 	reftable_record_init(&rec, block_reader_type(br));
 
-	/* We're looking for the last entry less/equal than the wanted key, so
-	   we have to go one entry too far and then back up.
-	*/
+	/*
+	 * We're looking for the last entry less than the wanted key so that
+	 * the next call to `block_reader_next()` would yield the wanted
+	 * record. We thus don't want to position our reader at the sought
+	 * after record, but one before. To do so, we have to go one entry too
+	 * far and then back up.
+	 */
 	while (1) {
 		block_iter_copy_from(&next, it);
 		err = block_iter_next(&next, &rec);
 		if (err < 0)
 			goto done;
-
-		reftable_record_key(&rec, &it->last_key);
-		if (err > 0 || strbuf_cmp(&it->last_key, want) >= 0) {
+		if (err > 0) {
 			err = 0;
 			goto done;
 		}
 
+		/*
+		 * Check whether the current key is greater or equal to the
+		 * sought-after key. In case it is greater we know that the
+		 * record does not exist in the block and can thus abort early.
+		 * In case it is equal to the sought-after key we have found
+		 * the desired record.
+		 */
+		reftable_record_key(&rec, &it->last_key);
+		if (strbuf_cmp(&it->last_key, want) >= 0)
+			goto done;
+
 		block_iter_copy_from(it, &next);
 	}
 
Signature · 1 line
2.44.GIT
#19 lore raw

Re: [PATCH v2 4/7] reftable/block: refactor binary search over restart points

Quoted 11 lines · Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---

...

Quoted 11 lines
reftable/block.c +10 -1
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning

reftable/block.c int block_reader_seek(struct block_reader *br, struct block_iter *it, · line 414

Quoted 10 lines
reftable/block.c +10 -0
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.

To clarify my understanding, does a restart_offset not exist at the beginning of a block? I was originally under the impression that the first reset point would be at the beginning of the block (or just after the header). If this was the case, the first restart point being greater would indicate that the wanted record is not contained within the block.

-Justin

Quoted 56 lines
reftable/block.c +28 -6
+	 *
+	 *   - `i == restart_count`: The wanted record was not found at any of
+	 *     the restart points. As there is no restart point at the end of
+	 *     the section the record may thus be contained in the last block.
+	 *
+	 *   - `i > 0`: The wanted record must be contained in the section
+	 *     before the found restart point. We thus do a linear search
+	 *     starting from the preceding restart point.
+	 */
 	if (i > 0)
 		it->next_off = block_reader_restart_offset(br, i - 1);
 	else
@@ -399,21 +429,34 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 
 	reftable_record_init(&rec, block_reader_type(br));
 
-	/* We're looking for the last entry less/equal than the wanted key, so
-	   we have to go one entry too far and then back up.
-	*/
+	/*
+	 * We're looking for the last entry less than the wanted key so that
+	 * the next call to `block_reader_next()` would yield the wanted
+	 * record. We thus don't want to position our reader at the sought
+	 * after record, but one before. To do so, we have to go one entry too
+	 * far and then back up.
+	 */
 	while (1) {
 		block_iter_copy_from(&next, it);
 		err = block_iter_next(&next, &rec);
 		if (err < 0)
 			goto done;
-
-		reftable_record_key(&rec, &it->last_key);
-		if (err > 0 || strbuf_cmp(&it->last_key, want) >= 0) {
+		if (err > 0) {
 			err = 0;
 			goto done;
 		}
 
+		/*
+		 * Check whether the current key is greater or equal to the
+		 * sought-after key. In case it is greater we know that the
+		 * record does not exist in the block and can thus abort early.
+		 * In case it is equal to the sought-after key we have found
+		 * the desired record.
+		 */
+		reftable_record_key(&rec, &it->last_key);
+		if (strbuf_cmp(&it->last_key, want) >= 0)
+			goto done;
+
 		block_iter_copy_from(it, &next);
 	}
 
Signature · 1 line
2.44.GIT
#20 lore raw

Re: [PATCH v2 4/7] reftable/block: refactor binary search over restart points

Quoted 32 lines · Justin Tobler
Quoted 11 lines · Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---

...

Quoted 19 lines
reftable/block.c +18 -1
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
Quoted 8 lines

reftable/block.c int block_reader_seek(struct block_reader *br, struct block_iter *it, · line 414

Quoted 2 lines
reftable/block.c +2 -0
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.

To clarify my understanding, does a restart_offset not exist at the beginning of a block? I was originally under the impression that the first reset point would be at the beginning of the block (or just after the header). If this was the case, the first restart point being greater would indicate that the wanted record is not contained within the block.

It wouldn't make much sense to include it as a restart point. A restart point is a point where the prefix compression will be reset such that the record at that point can be read without reading preceding records. This is always implicitly true for the first record in a block as it is never prefix-compressed. Consequently, writing a restart point for the first record would be a waste of disk space.

Patrick

Quoted 60 lines

-Justin

Quoted 56 lines
reftable/block.c +28 -6
+	 *
+	 *   - `i == restart_count`: The wanted record was not found at any of
+	 *     the restart points. As there is no restart point at the end of
+	 *     the section the record may thus be contained in the last block.
+	 *
+	 *   - `i > 0`: The wanted record must be contained in the section
+	 *     before the found restart point. We thus do a linear search
+	 *     starting from the preceding restart point.
+	 */
 	if (i > 0)
 		it->next_off = block_reader_restart_offset(br, i - 1);
 	else
@@ -399,21 +429,34 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 
 	reftable_record_init(&rec, block_reader_type(br));
 
-	/* We're looking for the last entry less/equal than the wanted key, so
-	   we have to go one entry too far and then back up.
-	*/
+	/*
+	 * We're looking for the last entry less than the wanted key so that
+	 * the next call to `block_reader_next()` would yield the wanted
+	 * record. We thus don't want to position our reader at the sought
+	 * after record, but one before. To do so, we have to go one entry too
+	 * far and then back up.
+	 */
 	while (1) {
 		block_iter_copy_from(&next, it);
 		err = block_iter_next(&next, &rec);
 		if (err < 0)
 			goto done;
-
-		reftable_record_key(&rec, &it->last_key);
-		if (err > 0 || strbuf_cmp(&it->last_key, want) >= 0) {
+		if (err > 0) {
 			err = 0;
 			goto done;
 		}
 
+		/*
+		 * Check whether the current key is greater or equal to the
+		 * sought-after key. In case it is greater we know that the
+		 * record does not exist in the block and can thus abort early.
+		 * In case it is equal to the sought-after key we have found
+		 * the desired record.
+		 */
+		reftable_record_key(&rec, &it->last_key);
+		if (strbuf_cmp(&it->last_key, want) >= 0)
+			goto done;
+
 		block_iter_copy_from(it, &next);
 	}
 
Signature · 1 line
2.44.GIT
#21 lore raw

Re: [PATCH v2 4/7] reftable/block: refactor binary search over restart points

Quoted 42 lines · Patrick Steinhardt
Quoted 32 lines · Justin Tobler
Quoted 11 lines · Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---

...

Quoted 19 lines
reftable/block.c +18 -1
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
Quoted 8 lines

reftable/block.c int block_reader_seek(struct block_reader *br, struct block_iter *it, · line 414

Quoted 2 lines
reftable/block.c +2 -0
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.

To clarify my understanding, does a restart_offset not exist at the beginning of a block? I was originally under the impression that the first reset point would be at the beginning of the block (or just after the header). If this was the case, the first restart point being greater would indicate that the wanted record is not contained within the block.

Quoted 6 lines

It wouldn't make much sense to include it as a restart point. A restart point is a point where the prefix compression will be reset such that the record at that point can be read without reading preceding records. This is always implicitly true for the first record in a block as it is never prefix-compressed. Consequently, writing a restart point for the first record would be a waste of disk space.

Thanks Patrick! Good to know :)

From Documentation/technical/reftable.txt:

Quoted 4 lines

As the first ref block shares the first file block with the file header, all restart_offset in the first block are relative to the start of the file (position 0), and include the file header. This forces the first restart_offset to be 28.

I'm assumming this is out-of-date.

-Justin

#22 lore raw

Re: [PATCH v2 4/7] reftable/block: refactor binary search over restart points

Quoted 52 lines · Justin Tobler
Quoted 42 lines · Patrick Steinhardt
Quoted 32 lines · Justin Tobler
Quoted 11 lines · Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---

...

Quoted 19 lines
(quoted diff) +18 -1
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
Quoted 8 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.

To clarify my understanding, does a restart_offset not exist at the beginning of a block? I was originally under the impression that the first reset point would be at the beginning of the block (or just after the header). If this was the case, the first restart point being greater would indicate that the wanted record is not contained within the block.

Quoted 6 lines

It wouldn't make much sense to include it as a restart point. A restart point is a point where the prefix compression will be reset such that the record at that point can be read without reading preceding records. This is always implicitly true for the first record in a block as it is never prefix-compressed. Consequently, writing a restart point for the first record would be a waste of disk space.

Thanks Patrick! Good to know :)

Quoted 8 lines

From Documentation/technical/reftable.txt:

Quoted 4 lines

As the first ref block shares the first file block with the file header, all restart_offset in the first block are relative to the start of the file (position 0), and include the file header. This forces the first restart_offset to be 28.

I'm assumming this is out-of-date.

That paragraph actually made me re-check my own assumptions. Turns out my claim is wrong. The function responsible for registering restarts is block_writer_register_restart(), which gets a parameter is_restart that is determined by reftable_encode_key(). And that function in turn will set restart = 1 whenever prefix_len == 0. And given that the first record always has prefix_len == 0, it will thus have a restart point, as well.

I'll update the comment like this:

reftable/block.c +6 -3
index 8bb4e43cec..298e8c56b9 100644
@@ -417,9 +417,12 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	/*
 	 * Now there are multiple cases:
 	 *
-	 *   - `i == 0`: The wanted record must be contained before the first
-	 *     restart point. We will thus start searching for the record in
-	 *     that section after accounting for the header offset.
+	 *   - `i == 0`: The wanted record is smaller than the record found at
+	 *     the first restart point. As the first restart point is the first
+	 *     record in the block, our wanted record cannot be located in this
+	 *     block at all. We still need to position the iterator so that the
+	 *     next call to `block_iter_next()` will yield an end-of-iterator
+	 *     signal.
 	 *
 	 *   - `i == restart_count`: The wanted record was not found at any of
 	 *     the restart points. As there is no restart point at the end of

Thanks for making me challenge my own assumptions!

Patrick

#23 lore raw

PATCH v2 5/7 reftable/block: fix error handling when searching restart points

In reply to Patrick Steinhardt

When doing the binary search over restart points in a block we need to decode the record keys. This decoding step can result in an error when the block is corrupted, which we indicate to the caller of the binary search by setting args.error = 1. But the only caller that exists mishandles this because it in fact performs the error check before calling binsearch().

Fix this bug by checking for errors at the right point in time. Furthermore, refactor binsearch() so that it aborts the search in case the callback function returns a negative value so that we don't needlessly continue to search the block.

---
 reftable/basics.c | 5 ++++-
 reftable/basics.h | 5 +++--
 reftable/block.c  | 9 ++++-----
 3 files changed, 11 insertions(+), 8 deletions(-)
reftable/basics.c +4 -1
index 2c5f34b39e..fea711db7e 100644
@@ -39,8 +39,11 @@ size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 	 */
 	while (hi - lo > 1) {
 		size_t mid = lo + (hi - lo) / 2;
+		int ret = f(mid, args);
+		if (ret < 0)
+			return sz;
 
-		if (f(mid, args))
+		if (ret > 0)
 			hi = mid;
 		else
 			lo = mid;
reftable/basics.h +3 -2
index 2672520e76..523ecd5307 100644
@@ -22,8 +22,9 @@ uint32_t get_be24(uint8_t *in);
 void put_be16(uint8_t *out, uint16_t i);
 
 /*
- * find smallest index i in [0, sz) at which f(i) is true, assuming
- * that f is ascending. Return sz if f(i) is false for all indices.
+ * find smallest index i in [0, sz) at which `f(i) > 0`, assuming that f is
+ * ascending. Return sz if `f(i) == 0` for all indices. The search is aborted
+ * and `sz` is returned in case `f(i) < 0`.
  *
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
reftable/block.c +4 -5
index 6cd4c053df..ca80a05e21 100644
@@ -387,11 +387,6 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	int err = 0;
 	size_t i;
 
-	if (args.error) {
-		err = REFTABLE_FORMAT_ERROR;
-		goto done;
-	}
-
 	/*
 	 * Perform a binary search over the block's restart points, which
 	 * avoids doing a linear scan over the whole block. Like this, we
@@ -405,6 +400,10 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	 * too many record.
 	 */
 	i = binsearch(br->restart_count, &restart_needle_less, &args);
+	if (args.error) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
 
 	/*
 	 * Now there are multiple cases:
Signature · 1 line
2.44.GIT
#24 lore raw

PATCH v2 6/7 reftable/record: extract function to decode key lengths

In reply to Patrick Steinhardt

We're about to refactor the binary search over restart points so that it does not need to fully decode the record keys anymore. To do so we will need to decode the record key lengths, which is non-trivial logic.

Extract the logic to decode these lengths from refatble_decode_key() so that we can reuse it.

---
 reftable/record.c | 34 +++++++++++++++++++++++++---------
 reftable/record.h |  6 ++++++
 2 files changed, 31 insertions(+), 9 deletions(-)
reftable/record.c +25 -9
index 23b497adab..5506f3e913 100644
@@ -159,26 +159,42 @@ int reftable_encode_key(int *restart, struct string_view dest,
 	return start.len - dest.len;
 }
 
-int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
-			struct string_view in)
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra)
 {
-	int start_len = in.len;
-	uint64_t prefix_len = 0;
-	uint64_t suffix_len = 0;
+	size_t start_len = in.len;
 	int n;
 
-	n = get_var_int(&prefix_len, &in);
+	n = get_var_int(prefix_len, &in);
 	if (n < 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	n = get_var_int(&suffix_len, &in);
+	n = get_var_int(suffix_len, &in);
 	if (n <= 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	*extra = (uint8_t)(suffix_len & 0x7);
-	suffix_len >>= 3;
+	*extra = (uint8_t)(*suffix_len & 0x7);
+	*suffix_len >>= 3;
+
+	return start_len - in.len;
+}
+
+int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
+			struct string_view in)
+{
+	int start_len = in.len;
+	uint64_t prefix_len = 0;
+	uint64_t suffix_len = 0;
+	int n;
+
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, extra);
+	if (n < 0)
+		return -1;
+	string_view_consume(&in, n);
 
 	if (in.len < suffix_len ||
 	    prefix_len > last_key->len)
reftable/record.h +6 -0
index 826ee1c55c..d778133e6e 100644
@@ -86,6 +86,12 @@ int reftable_encode_key(int *is_restart, struct string_view dest,
 			struct strbuf prev_key, struct strbuf key,
 			uint8_t extra);
 
+/* Decode a record's key lengths. */
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra);
+
 /*
  * Decode into `last_key` and `extra` from `in`. `last_key` is expected to
  * contain the decoded key of the preceding record, if any.
Signature · 1 line
2.44.GIT
#25 lore raw

PATCH v2 7/7 reftable/block: avoid decoding keys when searching restart points

In reply to Patrick Steinhardt

When searching over restart points in a block we decode the key of each of the records, which results in a memory allocation. This is quite pointless though given that records it restart points will never use prefix compression and thus store their keys verbatim in the block.

Refactor the code so that we can avoid decoding the keys, which saves us some allocations.

---
 reftable/block.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)
reftable/block.c +19 -10
index ca80a05e21..8bb4e43cec 100644
@@ -287,23 +287,32 @@ static int restart_needle_less(size_t idx, void *_args)
 		.buf = args->reader->block.data + off,
 		.len = args->reader->block_len - off,
 	};
-	struct strbuf kth_restart_key = STRBUF_INIT;
-	uint8_t unused_extra;
-	int result, n;
+	uint64_t prefix_len, suffix_len;
+	uint8_t extra;
+	int n;
 
 	/*
-	 * TODO: The restart key is verbatim in the block, so we can in theory
-	 * avoid decoding the key and thus save some allocations.
+	 * Records at restart points are stored without prefix compression, so
+	 * there is no need to fully decode the record key here. This removes
+	 * the need for allocating memory.
 	 */
-	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
-	if (n < 0) {
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, &extra);
+	if (n < 0 || prefix_len) {
 		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&args->needle, &kth_restart_key);
-	strbuf_release(&kth_restart_key);
-	return result < 0;
+	string_view_consume(&in, n);
+	if (suffix_len > in.len) {
+		args->error = 1;
+		return -1;
+	}
+
+	n = memcmp(args->needle.buf, in.buf,
+		   args->needle.len < suffix_len ? args->needle.len : suffix_len);
+	if (n)
+		return n < 0;
+	return args->needle.len < suffix_len;
 }
 
 void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
Signature · 1 line
2.44.GIT
#26 lore raw

Re: [PATCH v2 7/7] reftable/block: avoid decoding keys when searching restart points

Quoted 7 lines · Patrick Steinhardt

When searching over restart points in a block we decode the key of each of the records, which results in a memory allocation. This is quite pointless though given that records it restart points will never use prefix compression and thus store their keys verbatim in the block.

Refactor the code so that we can avoid decoding the keys, which saves us some allocations.

Out of curiousity, do you have any benchmarks around this change and would that be something we would want to add to the commit message?

-Justin

Quoted 56 lines
---
 reftable/block.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)
reftable/block.c +19 -10
index ca80a05e21..8bb4e43cec 100644
@@ -287,23 +287,32 @@ static int restart_needle_less(size_t idx, void *_args)
 		.buf = args->reader->block.data + off,
 		.len = args->reader->block_len - off,
 	};
-	struct strbuf kth_restart_key = STRBUF_INIT;
-	uint8_t unused_extra;
-	int result, n;
+	uint64_t prefix_len, suffix_len;
+	uint8_t extra;
+	int n;
 
 	/*
-	 * TODO: The restart key is verbatim in the block, so we can in theory
-	 * avoid decoding the key and thus save some allocations.
+	 * Records at restart points are stored without prefix compression, so
+	 * there is no need to fully decode the record key here. This removes
+	 * the need for allocating memory.
 	 */
-	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
-	if (n < 0) {
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, &extra);
+	if (n < 0 || prefix_len) {
 		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&args->needle, &kth_restart_key);
-	strbuf_release(&kth_restart_key);
-	return result < 0;
+	string_view_consume(&in, n);
+	if (suffix_len > in.len) {
+		args->error = 1;
+		return -1;
+	}
+
+	n = memcmp(args->needle.buf, in.buf,
+		   args->needle.len < suffix_len ? args->needle.len : suffix_len);
+	if (n)
+		return n < 0;
+	return args->needle.len < suffix_len;
 }
 
 void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
Signature · 1 line
2.44.GIT
#27 lore raw

Re: [PATCH v2 7/7] reftable/block: avoid decoding keys when searching restart points

Quoted 11 lines · Justin Tobler
Quoted 7 lines · Patrick Steinhardt

When searching over restart points in a block we decode the key of each of the records, which results in a memory allocation. This is quite pointless though given that records it restart points will never use prefix compression and thus store their keys verbatim in the block.

Refactor the code so that we can avoid decoding the keys, which saves us some allocations.

Out of curiousity, do you have any benchmarks around this change and would that be something we would want to add to the commit message?

I don't have a benchmark. The problem is that the difference isn't really measureable when doing a single seek, only, because seeks are simply too fast. The only usecase where I know that there are a ton of of record seeks are writes, but here the performance improvement is getting drowned out by everything else.

You can try to measure allocations and indeed see a difference. But again, this is getting drowned out by the noise for writes. With my block reader refactorings (ps/reftable-block-iteration-optim) you can see the difference when iterating through refs. Before:

HEAP SUMMARY:
in use at exit: 13,603 bytes in 125 blocks
total heap usage: 314 allocs, 189 frees, 106,035 bytes allocated

After:

HEAP SUMMARY:
in use at exit: 13,603 bytes in 125 blocks
total heap usage: 303 allocs, 178 frees, 105,763 bytes allocated

But yeah, it's nothing that'd make you go "Oh, wow!". As said, it will add up when doing many seeks, but I didn't manage to find a proper benchamrk yet that would be worthy to make it into the commit message.

Patrick

Quoted 60 lines

-Justin

Quoted 56 lines
---
 reftable/block.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)
reftable/block.c +19 -10
index ca80a05e21..8bb4e43cec 100644
@@ -287,23 +287,32 @@ static int restart_needle_less(size_t idx, void *_args)
 		.buf = args->reader->block.data + off,
 		.len = args->reader->block_len - off,
 	};
-	struct strbuf kth_restart_key = STRBUF_INIT;
-	uint8_t unused_extra;
-	int result, n;
+	uint64_t prefix_len, suffix_len;
+	uint8_t extra;
+	int n;
 
 	/*
-	 * TODO: The restart key is verbatim in the block, so we can in theory
-	 * avoid decoding the key and thus save some allocations.
+	 * Records at restart points are stored without prefix compression, so
+	 * there is no need to fully decode the record key here. This removes
+	 * the need for allocating memory.
 	 */
-	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
-	if (n < 0) {
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, &extra);
+	if (n < 0 || prefix_len) {
 		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&args->needle, &kth_restart_key);
-	strbuf_release(&kth_restart_key);
-	return result < 0;
+	string_view_consume(&in, n);
+	if (suffix_len > in.len) {
+		args->error = 1;
+		return -1;
+	}
+
+	n = memcmp(args->needle.buf, in.buf,
+		   args->needle.len < suffix_len ? args->needle.len : suffix_len);
+	if (n)
+		return n < 0;
+	return args->needle.len < suffix_len;
 }
 
 void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
Signature · 1 line
2.44.GIT
#28 lore raw

v3 cover reftable: improvements for the `binsearch()` mechanism

In reply to Patrick Steinhardt

Hi,

this is the third version of my patch series that refactors the binsearch() mechanism in the reftable library. There's only a single change compared to v2, namely a rename of some arguments when calling refname_needle_lesseq().

Thanks!

Patrick

Patrick Steinhardt (7):
reftable/basics: fix return type of binsearch() to be size_t reftable/basics: improve binsearch() test reftable/refname: refactor binary search over refnames reftable/block: refactor binary search over restart points reftable/block: fix error handling when searching restart points reftable/record: extract function to decode key lengths reftable/block: avoid decoding keys when searching restart points

 reftable/basics.c      |   7 ++-
 reftable/basics.h      |   7 +--
 reftable/basics_test.c |  55 +++++++++++---------
 reftable/block.c       | 114 ++++++++++++++++++++++++++++++-----------
 reftable/record.c      |  34 ++++++++----
 reftable/record.h      |   6 +++
 reftable/refname.c     |  53 +++++++++----------
 7 files changed, 179 insertions(+), 97 deletions(-)

Range-diff against v2:

1:  cd82ac6531 = 1:  baa07ef144 reftable/basics: fix return type of `binsearch()` to be `size_t`
2:  a277d4fa6f = 2:  cbc2a107c1 reftable/basics: improve `binsearch()` test
3:  9ffcf45c32 ! 3:  f5bf65e0dd reftable/refname: refactor binary search over refnames
    @@ reftable/refname.c
      };
      
     -static int find_name(size_t k, void *arg)
    -+static int refname_needle_lesseq(size_t k, void *arg)
    ++static int refname_needle_lesseq(size_t k, void *_args)
      {
     -	struct find_arg *f_arg = arg;
     -	return strcmp(f_arg->names[k], f_arg->want) >= 0;
    -+	struct refname_needle_lesseq_args *f_arg = arg;
    -+	return strcmp(f_arg->needle, f_arg->haystack[k]) <= 0;
    ++	struct refname_needle_lesseq_args *args = _args;
    ++	return strcmp(args->needle, args->haystack[k]) <= 0;
      }
      
      static int modification_has_ref(struct modification *mod, const char *name)
    @@ reftable/refname.c: static int modification_has_ref(struct modification *mod, co
     -		struct find_arg arg = {
     -			.names = mod->add,
     -			.want = name,
    -+		struct refname_needle_lesseq_args arg = {
    ++		struct refname_needle_lesseq_args args = {
     +			.haystack = mod->add,
     +			.needle = name,
      		};
     -		size_t idx = binsearch(mod->add_len, find_name, &arg);
    -+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
    ++		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &args);
      		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
      			return 0;
      	}
    @@ reftable/refname.c: static int modification_has_ref(struct modification *mod, co
     -		struct find_arg arg = {
     -			.names = mod->del,
     -			.want = name,
    -+		struct refname_needle_lesseq_args arg = {
    ++		struct refname_needle_lesseq_args args = {
     +			.haystack = mod->del,
     +			.needle = name,
      		};
     -		size_t idx = binsearch(mod->del_len, find_name, &arg);
    -+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
    ++		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &args);
      		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
      			return 1;
      	}
    @@ reftable/refname.c: static int modification_has_ref_with_prefix(struct modificat
     -		struct find_arg arg = {
     -			.names = mod->add,
     -			.want = prefix,
    -+		struct refname_needle_lesseq_args arg = {
    ++		struct refname_needle_lesseq_args args = {
     +			.haystack = mod->add,
     +			.needle = prefix,
      		};
     -		size_t idx = binsearch(mod->add_len, find_name, &arg);
    -+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &arg);
    ++		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &args);
      		if (idx < mod->add_len &&
      		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
      			goto done;
    @@ reftable/refname.c: static int modification_has_ref_with_prefix(struct modificat
     -			struct find_arg arg = {
     -				.names = mod->del,
     -				.want = ref.refname,
    -+			struct refname_needle_lesseq_args arg = {
    ++			struct refname_needle_lesseq_args args = {
     +				.haystack = mod->del,
     +				.needle = ref.refname,
      			};
     -			size_t idx = binsearch(mod->del_len, find_name, &arg);
    -+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &arg);
    ++			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &args);
      			if (idx < mod->del_len &&
      			    !strcmp(ref.refname, mod->del[idx]))
      				continue;
4:  5e20d93ae0 = 4:  435cd0be94 reftable/block: refactor binary search over restart points
5:  5bbeab114f = 5:  8d8abfd290 reftable/block: fix error handling when searching restart points
6:  271bacb210 = 6:  f87f7ad01a reftable/record: extract function to decode key lengths
7:  e751b3c536 = 7:  f53bf9e1cc reftable/block: avoid decoding keys when searching restart points

base-commit: 11c821f2f2a31e70fb5cc449f9a29401c333aad2

Signature · 1 line
2.44.GIT
#29 lore raw

PATCH v3 1/7 reftable/basics: fix return type of `binsearch()` to be `size_t`

The binsearch() function can be used to find the first element for which a callback functions returns a truish value. But while the array size is of type size_t, the function in fact returns an int that is supposed to index into that array.

Fix the function signature to return a size_t. This conversion does not change any semantics given that the function would only ever return a value in the range [0, sz] anyway.

---
 reftable/basics.c      |  2 +-
 reftable/basics.h      |  2 +-
 reftable/basics_test.c |  6 +++---
 reftable/block.c       |  3 ++-
 reftable/refname.c     | 17 +++++++----------
 5 files changed, 14 insertions(+), 16 deletions(-)
reftable/basics.c +1 -1
index 0785aff941..2c5f34b39e 100644
@@ -27,7 +27,7 @@ void put_be16(uint8_t *out, uint16_t i)
 	out[1] = (uint8_t)(i & 0xff);
 }
 
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 {
 	size_t lo = 0;
 	size_t hi = sz;
reftable/basics.h +1 -1
index 91f3533efe..2672520e76 100644
@@ -28,7 +28,7 @@ void put_be16(uint8_t *out, uint16_t i);
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
  */
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
 
 /*
  * Frees a NULL terminated array of malloced strings. The array itself is also
reftable/basics_test.c +3 -3
index 1fcd229725..dc1c87c5df 100644
@@ -34,15 +34,15 @@ static void test_binsearch(void)
 
 	int i = 0;
 	for (i = 1; i < 11; i++) {
-		int res;
+		size_t res;
+
 		args.key = i;
 		res = binsearch(sz, &binsearch_func, &args);
 
 		if (res < sz) {
 			EXPECT(args.key < arr[res]);
-			if (res > 0) {
+			if (res > 0)
 				EXPECT(args.key >= arr[res - 1]);
-			}
 		} else {
 			EXPECT(args.key == 10 || args.key == 11);
 		}
reftable/block.c +2 -1
index e2a2cee58d..422885bddb 100644
@@ -382,7 +382,8 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
-	int err = 0, i;
+	int err = 0;
+	size_t i;
 
 	if (args.error) {
 		err = REFTABLE_FORMAT_ERROR;
reftable/refname.c +7 -10
index 7570e4acf9..64eba1b886 100644
@@ -33,10 +33,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->add,
 			.want = name,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
-		if (idx < mod->add_len && !strcmp(mod->add[idx], name)) {
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
-		}
 	}
 
 	if (mod->del_len > 0) {
@@ -44,10 +43,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->del,
 			.want = name,
 		};
-		int idx = binsearch(mod->del_len, find_name, &arg);
-		if (idx < mod->del_len && !strcmp(mod->del[idx], name)) {
+		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
-		}
 	}
 
 	err = reftable_table_read_ref(&mod->tab, name, &ref);
@@ -77,7 +75,7 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			.names = mod->add,
 			.want = prefix,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -96,11 +94,10 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 				.names = mod->del,
 				.want = ref.refname,
 			};
-			int idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, find_name, &arg);
 			if (idx < mod->del_len &&
-			    !strcmp(ref.refname, mod->del[idx])) {
+			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
-			}
 		}
 
 		if (strncmp(ref.refname, prefix, strlen(prefix))) {
Signature · 1 line
2.44.GIT
#30 lore raw

PATCH v3 2/7 reftable/basics: improve `binsearch()` test

In reply to Patrick Steinhardt

The binsearch() test is somewhat weird in that it doesn't explicitly spell out its expectations. Instead it does so in a rather ad-hoc way with some hard-to-understand computations.

Refactor the test to spell out the needle as well as expected index for all testcases. This refactoring highlights that the binsearch_func() is written somewhat weirdly to find the first integer smaller than the needle, not smaller or equal to it. Adjust the function accordingly.

While at it, rename the callback function to better convey its meaning.

---
 reftable/basics_test.c | 55 ++++++++++++++++++++++++------------------
 1 file changed, 31 insertions(+), 24 deletions(-)
reftable/basics_test.c +31 -24
index dc1c87c5df..997c4d9e01 100644
@@ -12,40 +12,47 @@ license that can be found in the LICENSE file or at
 #include "test_framework.h"
 #include "reftable-tests.h"
 
-struct binsearch_args {
-	int key;
-	int *arr;
+struct integer_needle_lesseq_args {
+	int needle;
+	int *haystack;
 };
 
-static int binsearch_func(size_t i, void *void_args)
+static int integer_needle_lesseq(size_t i, void *_args)
 {
-	struct binsearch_args *args = void_args;
-
-	return args->key < args->arr[i];
+	struct integer_needle_lesseq_args *args = _args;
+	return args->needle <= args->haystack[i];
 }
 
 static void test_binsearch(void)
 {
-	int arr[] = { 2, 4, 6, 8, 10 };
-	size_t sz = ARRAY_SIZE(arr);
-	struct binsearch_args args = {
-		.arr = arr,
+	int haystack[] = { 2, 4, 6, 8, 10 };
+	struct {
+		int needle;
+		size_t expected_idx;
+	} testcases[] = {
+		{-9000, 0},
+		{-1, 0},
+		{0, 0},
+		{2, 0},
+		{3, 1},
+		{4, 1},
+		{7, 3},
+		{9, 4},
+		{10, 4},
+		{11, 5},
+		{9000, 5},
 	};
+	size_t i = 0;
 
-	int i = 0;
-	for (i = 1; i < 11; i++) {
-		size_t res;
-
-		args.key = i;
-		res = binsearch(sz, &binsearch_func, &args);
+	for (i = 0; i < ARRAY_SIZE(testcases); i++) {
+		struct integer_needle_lesseq_args args = {
+			.haystack = haystack,
+			.needle = testcases[i].needle,
+		};
+		size_t idx;
 
-		if (res < sz) {
-			EXPECT(args.key < arr[res]);
-			if (res > 0)
-				EXPECT(args.key >= arr[res - 1]);
-		} else {
-			EXPECT(args.key == 10 || args.key == 11);
-		}
+		idx = binsearch(ARRAY_SIZE(haystack), &integer_needle_lesseq, &args);
+		EXPECT(idx == testcases[i].expected_idx);
 	}
 }
 
Signature · 1 line
2.44.GIT
#31 lore raw

PATCH v3 3/7 reftable/refname: refactor binary search over refnames

In reply to Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +22 -22
index 64eba1b886..bbfde15754 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };
 
-static int find_name(size_t k, void *arg)
+static int refname_needle_lesseq(size_t k, void *_args)
 {
-	struct find_arg *f_arg = arg;
-	return strcmp(f_arg->names[k], f_arg->want) >= 0;
+	struct refname_needle_lesseq_args *args = _args;
+	return strcmp(args->needle, args->haystack[k]) <= 0;
 }
 
 static int modification_has_ref(struct modification *mod, const char *name)
@@ -29,21 +29,21 @@ static int modification_has_ref(struct modification *mod, const char *name)
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = name,
+		struct refname_needle_lesseq_args args = {
+			.haystack = mod->add,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &args);
 		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
 	}
 
 	if (mod->del_len > 0) {
-		struct find_arg arg = {
-			.names = mod->del,
-			.want = name,
+		struct refname_needle_lesseq_args args = {
+			.haystack = mod->del,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &args);
 		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
 	}
@@ -71,11 +71,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = prefix,
+		struct refname_needle_lesseq_args args = {
+			.haystack = mod->add,
+			.needle = prefix,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &args);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -90,11 +90,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			goto done;
 
 		if (mod->del_len > 0) {
-			struct find_arg arg = {
-				.names = mod->del,
-				.want = ref.refname,
+			struct refname_needle_lesseq_args args = {
+				.haystack = mod->del,
+				.needle = ref.refname,
 			};
-			size_t idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &args);
 			if (idx < mod->del_len &&
 			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
Signature · 1 line
2.44.GIT
#32 lore raw

PATCH v3 4/7 reftable/block: refactor binary search over restart points

In reply to Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---
 reftable/block.c | 97 ++++++++++++++++++++++++++++++++++--------------
 1 file changed, 70 insertions(+), 27 deletions(-)
reftable/block.c +70 -27
index 422885bddb..6cd4c053df 100644
@@ -273,34 +273,36 @@ void block_reader_start(struct block_reader *br, struct block_iter *it)
 	it->next_off = br->header_off + 4;
 }
 
-struct restart_find_args {
+struct restart_needle_less_args {
 	int error;
-	struct strbuf key;
-	struct block_reader *r;
+	struct strbuf needle;
+	struct block_reader *reader;
 };
 
-static int restart_key_less(size_t idx, void *args)
+static int restart_needle_less(size_t idx, void *_args)
 {
-	struct restart_find_args *a = args;
-	uint32_t off = block_reader_restart_offset(a->r, idx);
+	struct restart_needle_less_args *args = _args;
+	uint32_t off = block_reader_restart_offset(args->reader, idx);
 	struct string_view in = {
-		.buf = a->r->block.data + off,
-		.len = a->r->block_len - off,
+		.buf = args->reader->block.data + off,
+		.len = args->reader->block_len - off,
 	};
-
-	/* the restart key is verbatim in the block, so this could avoid the
-	   alloc for decoding the key */
-	struct strbuf rkey = STRBUF_INIT;
+	struct strbuf kth_restart_key = STRBUF_INIT;
 	uint8_t unused_extra;
-	int n = reftable_decode_key(&rkey, &unused_extra, in);
-	int result;
+	int result, n;
+
+	/*
+	 * TODO: The restart key is verbatim in the block, so we can in theory
+	 * avoid decoding the key and thus save some allocations.
+	 */
+	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
 	if (n < 0) {
-		a->error = 1;
+		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&a->key, &rkey);
-	strbuf_release(&rkey);
+	result = strbuf_cmp(&args->needle, &kth_restart_key);
+	strbuf_release(&kth_restart_key);
 	return result < 0;
 }
 
@@ -376,9 +378,9 @@ void block_iter_close(struct block_iter *it)
 int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		      struct strbuf *want)
 {
-	struct restart_find_args args = {
-		.key = *want,
-		.r = br,
+	struct restart_needle_less_args args = {
+		.needle = *want,
+		.reader = br,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
@@ -390,7 +392,35 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		goto done;
 	}
 
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record must be contained before the first
+	 *     restart point. We will thus start searching for the record in
+	 *     that section after accounting for the header offset.
+	 *
+	 *   - `i == restart_count`: The wanted record was not found at any of
+	 *     the restart points. As there is no restart point at the end of
+	 *     the section the record may thus be contained in the last block.
+	 *
+	 *   - `i > 0`: The wanted record must be contained in the section
+	 *     before the found restart point. We thus do a linear search
+	 *     starting from the preceding restart point.
+	 */
 	if (i > 0)
 		it->next_off = block_reader_restart_offset(br, i - 1);
 	else
@@ -399,21 +429,34 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 
 	reftable_record_init(&rec, block_reader_type(br));
 
-	/* We're looking for the last entry less/equal than the wanted key, so
-	   we have to go one entry too far and then back up.
-	*/
+	/*
+	 * We're looking for the last entry less than the wanted key so that
+	 * the next call to `block_reader_next()` would yield the wanted
+	 * record. We thus don't want to position our reader at the sought
+	 * after record, but one before. To do so, we have to go one entry too
+	 * far and then back up.
+	 */
 	while (1) {
 		block_iter_copy_from(&next, it);
 		err = block_iter_next(&next, &rec);
 		if (err < 0)
 			goto done;
-
-		reftable_record_key(&rec, &it->last_key);
-		if (err > 0 || strbuf_cmp(&it->last_key, want) >= 0) {
+		if (err > 0) {
 			err = 0;
 			goto done;
 		}
 
+		/*
+		 * Check whether the current key is greater or equal to the
+		 * sought-after key. In case it is greater we know that the
+		 * record does not exist in the block and can thus abort early.
+		 * In case it is equal to the sought-after key we have found
+		 * the desired record.
+		 */
+		reftable_record_key(&rec, &it->last_key);
+		if (strbuf_cmp(&it->last_key, want) >= 0)
+			goto done;
+
 		block_iter_copy_from(it, &next);
 	}
 
Signature · 1 line
2.44.GIT
#33 lore raw

PATCH v3 5/7 reftable/block: fix error handling when searching restart points

In reply to Patrick Steinhardt

When doing the binary search over restart points in a block we need to decode the record keys. This decoding step can result in an error when the block is corrupted, which we indicate to the caller of the binary search by setting args.error = 1. But the only caller that exists mishandles this because it in fact performs the error check before calling binsearch().

Fix this bug by checking for errors at the right point in time. Furthermore, refactor binsearch() so that it aborts the search in case the callback function returns a negative value so that we don't needlessly continue to search the block.

---
 reftable/basics.c | 5 ++++-
 reftable/basics.h | 5 +++--
 reftable/block.c  | 9 ++++-----
 3 files changed, 11 insertions(+), 8 deletions(-)
reftable/basics.c +4 -1
index 2c5f34b39e..fea711db7e 100644
@@ -39,8 +39,11 @@ size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 	 */
 	while (hi - lo > 1) {
 		size_t mid = lo + (hi - lo) / 2;
+		int ret = f(mid, args);
+		if (ret < 0)
+			return sz;
 
-		if (f(mid, args))
+		if (ret > 0)
 			hi = mid;
 		else
 			lo = mid;
reftable/basics.h +3 -2
index 2672520e76..523ecd5307 100644
@@ -22,8 +22,9 @@ uint32_t get_be24(uint8_t *in);
 void put_be16(uint8_t *out, uint16_t i);
 
 /*
- * find smallest index i in [0, sz) at which f(i) is true, assuming
- * that f is ascending. Return sz if f(i) is false for all indices.
+ * find smallest index i in [0, sz) at which `f(i) > 0`, assuming that f is
+ * ascending. Return sz if `f(i) == 0` for all indices. The search is aborted
+ * and `sz` is returned in case `f(i) < 0`.
  *
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
reftable/block.c +4 -5
index 6cd4c053df..ca80a05e21 100644
@@ -387,11 +387,6 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	int err = 0;
 	size_t i;
 
-	if (args.error) {
-		err = REFTABLE_FORMAT_ERROR;
-		goto done;
-	}
-
 	/*
 	 * Perform a binary search over the block's restart points, which
 	 * avoids doing a linear scan over the whole block. Like this, we
@@ -405,6 +400,10 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	 * too many record.
 	 */
 	i = binsearch(br->restart_count, &restart_needle_less, &args);
+	if (args.error) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
 
 	/*
 	 * Now there are multiple cases:
Signature · 1 line
2.44.GIT
#34 lore raw

PATCH v3 6/7 reftable/record: extract function to decode key lengths

In reply to Patrick Steinhardt

We're about to refactor the binary search over restart points so that it does not need to fully decode the record keys anymore. To do so we will need to decode the record key lengths, which is non-trivial logic.

Extract the logic to decode these lengths from refatble_decode_key() so that we can reuse it.

---
 reftable/record.c | 34 +++++++++++++++++++++++++---------
 reftable/record.h |  6 ++++++
 2 files changed, 31 insertions(+), 9 deletions(-)
reftable/record.c +25 -9
index 23b497adab..5506f3e913 100644
@@ -159,26 +159,42 @@ int reftable_encode_key(int *restart, struct string_view dest,
 	return start.len - dest.len;
 }
 
-int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
-			struct string_view in)
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra)
 {
-	int start_len = in.len;
-	uint64_t prefix_len = 0;
-	uint64_t suffix_len = 0;
+	size_t start_len = in.len;
 	int n;
 
-	n = get_var_int(&prefix_len, &in);
+	n = get_var_int(prefix_len, &in);
 	if (n < 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	n = get_var_int(&suffix_len, &in);
+	n = get_var_int(suffix_len, &in);
 	if (n <= 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	*extra = (uint8_t)(suffix_len & 0x7);
-	suffix_len >>= 3;
+	*extra = (uint8_t)(*suffix_len & 0x7);
+	*suffix_len >>= 3;
+
+	return start_len - in.len;
+}
+
+int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
+			struct string_view in)
+{
+	int start_len = in.len;
+	uint64_t prefix_len = 0;
+	uint64_t suffix_len = 0;
+	int n;
+
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, extra);
+	if (n < 0)
+		return -1;
+	string_view_consume(&in, n);
 
 	if (in.len < suffix_len ||
 	    prefix_len > last_key->len)
reftable/record.h +6 -0
index 826ee1c55c..d778133e6e 100644
@@ -86,6 +86,12 @@ int reftable_encode_key(int *is_restart, struct string_view dest,
 			struct strbuf prev_key, struct strbuf key,
 			uint8_t extra);
 
+/* Decode a record's key lengths. */
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra);
+
 /*
  * Decode into `last_key` and `extra` from `in`. `last_key` is expected to
  * contain the decoded key of the preceding record, if any.
Signature · 1 line
2.44.GIT
#35 lore raw

PATCH v3 7/7 reftable/block: avoid decoding keys when searching restart points

In reply to Patrick Steinhardt

When searching over restart points in a block we decode the key of each of the records, which results in a memory allocation. This is quite pointless though given that records it restart points will never use prefix compression and thus store their keys verbatim in the block.

Refactor the code so that we can avoid decoding the keys, which saves us some allocations.

---
 reftable/block.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)
reftable/block.c +19 -10
index ca80a05e21..8bb4e43cec 100644
@@ -287,23 +287,32 @@ static int restart_needle_less(size_t idx, void *_args)
 		.buf = args->reader->block.data + off,
 		.len = args->reader->block_len - off,
 	};
-	struct strbuf kth_restart_key = STRBUF_INIT;
-	uint8_t unused_extra;
-	int result, n;
+	uint64_t prefix_len, suffix_len;
+	uint8_t extra;
+	int n;
 
 	/*
-	 * TODO: The restart key is verbatim in the block, so we can in theory
-	 * avoid decoding the key and thus save some allocations.
+	 * Records at restart points are stored without prefix compression, so
+	 * there is no need to fully decode the record key here. This removes
+	 * the need for allocating memory.
 	 */
-	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
-	if (n < 0) {
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, &extra);
+	if (n < 0 || prefix_len) {
 		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&args->needle, &kth_restart_key);
-	strbuf_release(&kth_restart_key);
-	return result < 0;
+	string_view_consume(&in, n);
+	if (suffix_len > in.len) {
+		args->error = 1;
+		return -1;
+	}
+
+	n = memcmp(args->needle.buf, in.buf,
+		   args->needle.len < suffix_len ? args->needle.len : suffix_len);
+	if (n)
+		return n < 0;
+	return args->needle.len < suffix_len;
 }
 
 void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
Signature · 1 line
2.44.GIT
#36 lore raw
In reply to Patrick Steinhardt
Quoted 11 lines · Patrick Steinhardt

Hi,

this is the third version of my patch series that refactors the binsearch() mechanism in the reftable library. There's only a single change compared to v2, namely a rename of some arguments when calling refname_needle_lesseq().

Thanks!

Patrick

...

Thanks Patrick! This version looks good to me :)

-Justin

#37 lore raw

v4 cover reftable: improvements for the `binsearch()` mechanism

In reply to Patrick Steinhardt

Hi,

this is the fourth and hopefully final version of my patch series that refactors the binsearch() mechanism in the reftable library.

There's only a single change compared to v3, which is a fix to a comment that was misexplaining one of the cases when searching for the restart points.

Thanks!

Patrick

Patrick Steinhardt (7):
reftable/basics: fix return type of binsearch() to be size_t reftable/basics: improve binsearch() test reftable/refname: refactor binary search over refnames reftable/block: refactor binary search over restart points reftable/block: fix error handling when searching restart points reftable/record: extract function to decode key lengths reftable/block: avoid decoding keys when searching restart points

 reftable/basics.c      |   7 ++-
 reftable/basics.h      |   7 +--
 reftable/basics_test.c |  55 ++++++++++---------
 reftable/block.c       | 117 ++++++++++++++++++++++++++++++-----------
 reftable/record.c      |  34 ++++++++----
 reftable/record.h      |   6 +++
 reftable/refname.c     |  53 +++++++++----------
 7 files changed, 182 insertions(+), 97 deletions(-)

Range-diff against v3:

1:  baa07ef144 = 1:  baa07ef144 reftable/basics: fix return type of `binsearch()` to be `size_t`
2:  cbc2a107c1 = 2:  cbc2a107c1 reftable/basics: improve `binsearch()` test
3:  f5bf65e0dd = 3:  f5bf65e0dd reftable/refname: refactor binary search over refnames
4:  435cd0be94 ! 4:  6d4a79f3e2 reftable/block: refactor binary search over restart points
    @@ reftable/block.c: int block_reader_seek(struct block_reader *br, struct block_it
     +	/*
     +	 * Now there are multiple cases:
     +	 *
    -+	 *   - `i == 0`: The wanted record must be contained before the first
    -+	 *     restart point. We will thus start searching for the record in
    -+	 *     that section after accounting for the header offset.
    ++	 *   - `i == 0`: The wanted record is smaller than the record found at
    ++	 *     the first restart point. As the first restart point is the first
    ++	 *     record in the block, our wanted record cannot be located in this
    ++	 *     block at all. We still need to position the iterator so that the
    ++	 *     next call to `block_iter_next()` will yield an end-of-iterator
    ++	 *     signal.
     +	 *
     +	 *   - `i == restart_count`: The wanted record was not found at any of
     +	 *     the restart points. As there is no restart point at the end of
5:  8d8abfd290 = 5:  52c238ee9f reftable/block: fix error handling when searching restart points
6:  f87f7ad01a = 6:  ca41ad30f4 reftable/record: extract function to decode key lengths
7:  f53bf9e1cc = 7:  632f725dde reftable/block: avoid decoding keys when searching restart points
Signature · 1 line
2.44.GIT
#38 lore raw

PATCH v4 1/7 reftable/basics: fix return type of `binsearch()` to be `size_t`

The binsearch() function can be used to find the first element for which a callback functions returns a truish value. But while the array size is of type size_t, the function in fact returns an int that is supposed to index into that array.

Fix the function signature to return a size_t. This conversion does not change any semantics given that the function would only ever return a value in the range [0, sz] anyway.

---
 reftable/basics.c      |  2 +-
 reftable/basics.h      |  2 +-
 reftable/basics_test.c |  6 +++---
 reftable/block.c       |  3 ++-
 reftable/refname.c     | 17 +++++++----------
 5 files changed, 14 insertions(+), 16 deletions(-)
reftable/basics.c +1 -1
index 0785aff941..2c5f34b39e 100644
@@ -27,7 +27,7 @@ void put_be16(uint8_t *out, uint16_t i)
 	out[1] = (uint8_t)(i & 0xff);
 }
 
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 {
 	size_t lo = 0;
 	size_t hi = sz;
reftable/basics.h +1 -1
index 91f3533efe..2672520e76 100644
@@ -28,7 +28,7 @@ void put_be16(uint8_t *out, uint16_t i);
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
  */
-int binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
+size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args);
 
 /*
  * Frees a NULL terminated array of malloced strings. The array itself is also
reftable/basics_test.c +3 -3
index 1fcd229725..dc1c87c5df 100644
@@ -34,15 +34,15 @@ static void test_binsearch(void)
 
 	int i = 0;
 	for (i = 1; i < 11; i++) {
-		int res;
+		size_t res;
+
 		args.key = i;
 		res = binsearch(sz, &binsearch_func, &args);
 
 		if (res < sz) {
 			EXPECT(args.key < arr[res]);
-			if (res > 0) {
+			if (res > 0)
 				EXPECT(args.key >= arr[res - 1]);
-			}
 		} else {
 			EXPECT(args.key == 10 || args.key == 11);
 		}
reftable/block.c +2 -1
index e2a2cee58d..422885bddb 100644
@@ -382,7 +382,8 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
-	int err = 0, i;
+	int err = 0;
+	size_t i;
 
 	if (args.error) {
 		err = REFTABLE_FORMAT_ERROR;
reftable/refname.c +7 -10
index 7570e4acf9..64eba1b886 100644
@@ -33,10 +33,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->add,
 			.want = name,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
-		if (idx < mod->add_len && !strcmp(mod->add[idx], name)) {
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
-		}
 	}
 
 	if (mod->del_len > 0) {
@@ -44,10 +43,9 @@ static int modification_has_ref(struct modification *mod, const char *name)
 			.names = mod->del,
 			.want = name,
 		};
-		int idx = binsearch(mod->del_len, find_name, &arg);
-		if (idx < mod->del_len && !strcmp(mod->del[idx], name)) {
+		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
-		}
 	}
 
 	err = reftable_table_read_ref(&mod->tab, name, &ref);
@@ -77,7 +75,7 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			.names = mod->add,
 			.want = prefix,
 		};
-		int idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, find_name, &arg);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -96,11 +94,10 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 				.names = mod->del,
 				.want = ref.refname,
 			};
-			int idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, find_name, &arg);
 			if (idx < mod->del_len &&
-			    !strcmp(ref.refname, mod->del[idx])) {
+			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
-			}
 		}
 
 		if (strncmp(ref.refname, prefix, strlen(prefix))) {
Signature · 1 line
2.44.GIT
#39 lore raw

PATCH v4 2/7 reftable/basics: improve `binsearch()` test

In reply to Patrick Steinhardt

The binsearch() test is somewhat weird in that it doesn't explicitly spell out its expectations. Instead it does so in a rather ad-hoc way with some hard-to-understand computations.

Refactor the test to spell out the needle as well as expected index for all testcases. This refactoring highlights that the binsearch_func() is written somewhat weirdly to find the first integer smaller than the needle, not smaller or equal to it. Adjust the function accordingly.

While at it, rename the callback function to better convey its meaning.

---
 reftable/basics_test.c | 55 ++++++++++++++++++++++++------------------
 1 file changed, 31 insertions(+), 24 deletions(-)
reftable/basics_test.c +31 -24
index dc1c87c5df..997c4d9e01 100644
@@ -12,40 +12,47 @@ license that can be found in the LICENSE file or at
 #include "test_framework.h"
 #include "reftable-tests.h"
 
-struct binsearch_args {
-	int key;
-	int *arr;
+struct integer_needle_lesseq_args {
+	int needle;
+	int *haystack;
 };
 
-static int binsearch_func(size_t i, void *void_args)
+static int integer_needle_lesseq(size_t i, void *_args)
 {
-	struct binsearch_args *args = void_args;
-
-	return args->key < args->arr[i];
+	struct integer_needle_lesseq_args *args = _args;
+	return args->needle <= args->haystack[i];
 }
 
 static void test_binsearch(void)
 {
-	int arr[] = { 2, 4, 6, 8, 10 };
-	size_t sz = ARRAY_SIZE(arr);
-	struct binsearch_args args = {
-		.arr = arr,
+	int haystack[] = { 2, 4, 6, 8, 10 };
+	struct {
+		int needle;
+		size_t expected_idx;
+	} testcases[] = {
+		{-9000, 0},
+		{-1, 0},
+		{0, 0},
+		{2, 0},
+		{3, 1},
+		{4, 1},
+		{7, 3},
+		{9, 4},
+		{10, 4},
+		{11, 5},
+		{9000, 5},
 	};
+	size_t i = 0;
 
-	int i = 0;
-	for (i = 1; i < 11; i++) {
-		size_t res;
-
-		args.key = i;
-		res = binsearch(sz, &binsearch_func, &args);
+	for (i = 0; i < ARRAY_SIZE(testcases); i++) {
+		struct integer_needle_lesseq_args args = {
+			.haystack = haystack,
+			.needle = testcases[i].needle,
+		};
+		size_t idx;
 
-		if (res < sz) {
-			EXPECT(args.key < arr[res]);
-			if (res > 0)
-				EXPECT(args.key >= arr[res - 1]);
-		} else {
-			EXPECT(args.key == 10 || args.key == 11);
-		}
+		idx = binsearch(ARRAY_SIZE(haystack), &integer_needle_lesseq, &args);
+		EXPECT(idx == testcases[i].expected_idx);
 	}
 }
 
Signature · 1 line
2.44.GIT
#40 lore raw

PATCH v4 3/7 reftable/refname: refactor binary search over refnames

In reply to Patrick Steinhardt

It is comparatively hard to understand how exactly the binary search over refnames works given that the function and variable names are not exactly easy to grasp. Rename them to make this more obvious. This should not result in any change in behaviour.

---
 reftable/refname.c | 44 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 22 deletions(-)
reftable/refname.c +22 -22
index 64eba1b886..bbfde15754 100644
@@ -12,15 +12,15 @@
 #include "refname.h"
 #include "reftable-iterator.h"
 
-struct find_arg {
-	char **names;
-	const char *want;
+struct refname_needle_lesseq_args {
+	char **haystack;
+	const char *needle;
 };
 
-static int find_name(size_t k, void *arg)
+static int refname_needle_lesseq(size_t k, void *_args)
 {
-	struct find_arg *f_arg = arg;
-	return strcmp(f_arg->names[k], f_arg->want) >= 0;
+	struct refname_needle_lesseq_args *args = _args;
+	return strcmp(args->needle, args->haystack[k]) <= 0;
 }
 
 static int modification_has_ref(struct modification *mod, const char *name)
@@ -29,21 +29,21 @@ static int modification_has_ref(struct modification *mod, const char *name)
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = name,
+		struct refname_needle_lesseq_args args = {
+			.haystack = mod->add,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &args);
 		if (idx < mod->add_len && !strcmp(mod->add[idx], name))
 			return 0;
 	}
 
 	if (mod->del_len > 0) {
-		struct find_arg arg = {
-			.names = mod->del,
-			.want = name,
+		struct refname_needle_lesseq_args args = {
+			.haystack = mod->del,
+			.needle = name,
 		};
-		size_t idx = binsearch(mod->del_len, find_name, &arg);
+		size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &args);
 		if (idx < mod->del_len && !strcmp(mod->del[idx], name))
 			return 1;
 	}
@@ -71,11 +71,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 	int err = 0;
 
 	if (mod->add_len > 0) {
-		struct find_arg arg = {
-			.names = mod->add,
-			.want = prefix,
+		struct refname_needle_lesseq_args args = {
+			.haystack = mod->add,
+			.needle = prefix,
 		};
-		size_t idx = binsearch(mod->add_len, find_name, &arg);
+		size_t idx = binsearch(mod->add_len, refname_needle_lesseq, &args);
 		if (idx < mod->add_len &&
 		    !strncmp(prefix, mod->add[idx], strlen(prefix)))
 			goto done;
@@ -90,11 +90,11 @@ static int modification_has_ref_with_prefix(struct modification *mod,
 			goto done;
 
 		if (mod->del_len > 0) {
-			struct find_arg arg = {
-				.names = mod->del,
-				.want = ref.refname,
+			struct refname_needle_lesseq_args args = {
+				.haystack = mod->del,
+				.needle = ref.refname,
 			};
-			size_t idx = binsearch(mod->del_len, find_name, &arg);
+			size_t idx = binsearch(mod->del_len, refname_needle_lesseq, &args);
 			if (idx < mod->del_len &&
 			    !strcmp(ref.refname, mod->del[idx]))
 				continue;
Signature · 1 line
2.44.GIT
#41 lore raw

PATCH v4 4/7 reftable/block: refactor binary search over restart points

In reply to Patrick Steinhardt

When seeking a record in our block reader we perform a binary search over the block's restart points so that we don't have to do a linear scan over the whole block. The logic to do so is quite intricate though, which makes it hard to understand.

Improve documentation and rename some of the functions and variables so that the code becomes easier to understand overall. This refactoring should not result in any change in behaviour.

---
 reftable/block.c | 100 ++++++++++++++++++++++++++++++++++-------------
 1 file changed, 73 insertions(+), 27 deletions(-)
reftable/block.c +73 -27
index 422885bddb..57e3dbf658 100644
@@ -273,34 +273,36 @@ void block_reader_start(struct block_reader *br, struct block_iter *it)
 	it->next_off = br->header_off + 4;
 }
 
-struct restart_find_args {
+struct restart_needle_less_args {
 	int error;
-	struct strbuf key;
-	struct block_reader *r;
+	struct strbuf needle;
+	struct block_reader *reader;
 };
 
-static int restart_key_less(size_t idx, void *args)
+static int restart_needle_less(size_t idx, void *_args)
 {
-	struct restart_find_args *a = args;
-	uint32_t off = block_reader_restart_offset(a->r, idx);
+	struct restart_needle_less_args *args = _args;
+	uint32_t off = block_reader_restart_offset(args->reader, idx);
 	struct string_view in = {
-		.buf = a->r->block.data + off,
-		.len = a->r->block_len - off,
+		.buf = args->reader->block.data + off,
+		.len = args->reader->block_len - off,
 	};
-
-	/* the restart key is verbatim in the block, so this could avoid the
-	   alloc for decoding the key */
-	struct strbuf rkey = STRBUF_INIT;
+	struct strbuf kth_restart_key = STRBUF_INIT;
 	uint8_t unused_extra;
-	int n = reftable_decode_key(&rkey, &unused_extra, in);
-	int result;
+	int result, n;
+
+	/*
+	 * TODO: The restart key is verbatim in the block, so we can in theory
+	 * avoid decoding the key and thus save some allocations.
+	 */
+	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
 	if (n < 0) {
-		a->error = 1;
+		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&a->key, &rkey);
-	strbuf_release(&rkey);
+	result = strbuf_cmp(&args->needle, &kth_restart_key);
+	strbuf_release(&kth_restart_key);
 	return result < 0;
 }
 
@@ -376,9 +378,9 @@ void block_iter_close(struct block_iter *it)
 int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		      struct strbuf *want)
 {
-	struct restart_find_args args = {
-		.key = *want,
-		.r = br,
+	struct restart_needle_less_args args = {
+		.needle = *want,
+		.reader = br,
 	};
 	struct block_iter next = BLOCK_ITER_INIT;
 	struct reftable_record rec;
@@ -390,7 +392,38 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 		goto done;
 	}
 
-	i = binsearch(br->restart_count, &restart_key_less, &args);
+	/*
+	 * Perform a binary search over the block's restart points, which
+	 * avoids doing a linear scan over the whole block. Like this, we
+	 * identify the section of the block that should contain our key.
+	 *
+	 * Note that we explicitly search for the first restart point _greater_
+	 * than the sought-after record, not _greater or equal_ to it. In case
+	 * the sought-after record is located directly at the restart point we
+	 * would otherwise start doing the linear search at the preceding
+	 * restart point. While that works alright, we would end up scanning
+	 * too many record.
+	 */
+	i = binsearch(br->restart_count, &restart_needle_less, &args);
+
+	/*
+	 * Now there are multiple cases:
+	 *
+	 *   - `i == 0`: The wanted record is smaller than the record found at
+	 *     the first restart point. As the first restart point is the first
+	 *     record in the block, our wanted record cannot be located in this
+	 *     block at all. We still need to position the iterator so that the
+	 *     next call to `block_iter_next()` will yield an end-of-iterator
+	 *     signal.
+	 *
+	 *   - `i == restart_count`: The wanted record was not found at any of
+	 *     the restart points. As there is no restart point at the end of
+	 *     the section the record may thus be contained in the last block.
+	 *
+	 *   - `i > 0`: The wanted record must be contained in the section
+	 *     before the found restart point. We thus do a linear search
+	 *     starting from the preceding restart point.
+	 */
 	if (i > 0)
 		it->next_off = block_reader_restart_offset(br, i - 1);
 	else
@@ -399,21 +432,34 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 
 	reftable_record_init(&rec, block_reader_type(br));
 
-	/* We're looking for the last entry less/equal than the wanted key, so
-	   we have to go one entry too far and then back up.
-	*/
+	/*
+	 * We're looking for the last entry less than the wanted key so that
+	 * the next call to `block_reader_next()` would yield the wanted
+	 * record. We thus don't want to position our reader at the sought
+	 * after record, but one before. To do so, we have to go one entry too
+	 * far and then back up.
+	 */
 	while (1) {
 		block_iter_copy_from(&next, it);
 		err = block_iter_next(&next, &rec);
 		if (err < 0)
 			goto done;
-
-		reftable_record_key(&rec, &it->last_key);
-		if (err > 0 || strbuf_cmp(&it->last_key, want) >= 0) {
+		if (err > 0) {
 			err = 0;
 			goto done;
 		}
 
+		/*
+		 * Check whether the current key is greater or equal to the
+		 * sought-after key. In case it is greater we know that the
+		 * record does not exist in the block and can thus abort early.
+		 * In case it is equal to the sought-after key we have found
+		 * the desired record.
+		 */
+		reftable_record_key(&rec, &it->last_key);
+		if (strbuf_cmp(&it->last_key, want) >= 0)
+			goto done;
+
 		block_iter_copy_from(it, &next);
 	}
 
Signature · 1 line
2.44.GIT
#42 lore raw

PATCH v4 5/7 reftable/block: fix error handling when searching restart points

In reply to Patrick Steinhardt

When doing the binary search over restart points in a block we need to decode the record keys. This decoding step can result in an error when the block is corrupted, which we indicate to the caller of the binary search by setting args.error = 1. But the only caller that exists mishandles this because it in fact performs the error check before calling binsearch().

Fix this bug by checking for errors at the right point in time. Furthermore, refactor binsearch() so that it aborts the search in case the callback function returns a negative value so that we don't needlessly continue to search the block.

---
 reftable/basics.c | 5 ++++-
 reftable/basics.h | 5 +++--
 reftable/block.c  | 9 ++++-----
 3 files changed, 11 insertions(+), 8 deletions(-)
reftable/basics.c +4 -1
index 2c5f34b39e..fea711db7e 100644
@@ -39,8 +39,11 @@ size_t binsearch(size_t sz, int (*f)(size_t k, void *args), void *args)
 	 */
 	while (hi - lo > 1) {
 		size_t mid = lo + (hi - lo) / 2;
+		int ret = f(mid, args);
+		if (ret < 0)
+			return sz;
 
-		if (f(mid, args))
+		if (ret > 0)
 			hi = mid;
 		else
 			lo = mid;
reftable/basics.h +3 -2
index 2672520e76..523ecd5307 100644
@@ -22,8 +22,9 @@ uint32_t get_be24(uint8_t *in);
 void put_be16(uint8_t *out, uint16_t i);
 
 /*
- * find smallest index i in [0, sz) at which f(i) is true, assuming
- * that f is ascending. Return sz if f(i) is false for all indices.
+ * find smallest index i in [0, sz) at which `f(i) > 0`, assuming that f is
+ * ascending. Return sz if `f(i) == 0` for all indices. The search is aborted
+ * and `sz` is returned in case `f(i) < 0`.
  *
  * Contrary to bsearch(3), this returns something useful if the argument is not
  * found.
reftable/block.c +4 -5
index 57e3dbf658..ca217636ae 100644
@@ -387,11 +387,6 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	int err = 0;
 	size_t i;
 
-	if (args.error) {
-		err = REFTABLE_FORMAT_ERROR;
-		goto done;
-	}
-
 	/*
 	 * Perform a binary search over the block's restart points, which
 	 * avoids doing a linear scan over the whole block. Like this, we
@@ -405,6 +400,10 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
 	 * too many record.
 	 */
 	i = binsearch(br->restart_count, &restart_needle_less, &args);
+	if (args.error) {
+		err = REFTABLE_FORMAT_ERROR;
+		goto done;
+	}
 
 	/*
 	 * Now there are multiple cases:
Signature · 1 line
2.44.GIT
#43 lore raw

PATCH v4 6/7 reftable/record: extract function to decode key lengths

In reply to Patrick Steinhardt

We're about to refactor the binary search over restart points so that it does not need to fully decode the record keys anymore. To do so we will need to decode the record key lengths, which is non-trivial logic.

Extract the logic to decode these lengths from refatble_decode_key() so that we can reuse it.

---
 reftable/record.c | 34 +++++++++++++++++++++++++---------
 reftable/record.h |  6 ++++++
 2 files changed, 31 insertions(+), 9 deletions(-)
reftable/record.c +25 -9
index 23b497adab..5506f3e913 100644
@@ -159,26 +159,42 @@ int reftable_encode_key(int *restart, struct string_view dest,
 	return start.len - dest.len;
 }
 
-int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
-			struct string_view in)
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra)
 {
-	int start_len = in.len;
-	uint64_t prefix_len = 0;
-	uint64_t suffix_len = 0;
+	size_t start_len = in.len;
 	int n;
 
-	n = get_var_int(&prefix_len, &in);
+	n = get_var_int(prefix_len, &in);
 	if (n < 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	n = get_var_int(&suffix_len, &in);
+	n = get_var_int(suffix_len, &in);
 	if (n <= 0)
 		return -1;
 	string_view_consume(&in, n);
 
-	*extra = (uint8_t)(suffix_len & 0x7);
-	suffix_len >>= 3;
+	*extra = (uint8_t)(*suffix_len & 0x7);
+	*suffix_len >>= 3;
+
+	return start_len - in.len;
+}
+
+int reftable_decode_key(struct strbuf *last_key, uint8_t *extra,
+			struct string_view in)
+{
+	int start_len = in.len;
+	uint64_t prefix_len = 0;
+	uint64_t suffix_len = 0;
+	int n;
+
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, extra);
+	if (n < 0)
+		return -1;
+	string_view_consume(&in, n);
 
 	if (in.len < suffix_len ||
 	    prefix_len > last_key->len)
reftable/record.h +6 -0
index 826ee1c55c..d778133e6e 100644
@@ -86,6 +86,12 @@ int reftable_encode_key(int *is_restart, struct string_view dest,
 			struct strbuf prev_key, struct strbuf key,
 			uint8_t extra);
 
+/* Decode a record's key lengths. */
+int reftable_decode_keylen(struct string_view in,
+			   uint64_t *prefix_len,
+			   uint64_t *suffix_len,
+			   uint8_t *extra);
+
 /*
  * Decode into `last_key` and `extra` from `in`. `last_key` is expected to
  * contain the decoded key of the preceding record, if any.
Signature · 1 line
2.44.GIT
#44 lore raw

PATCH v4 7/7 reftable/block: avoid decoding keys when searching restart points

In reply to Patrick Steinhardt

When searching over restart points in a block we decode the key of each of the records, which results in a memory allocation. This is quite pointless though given that records it restart points will never use prefix compression and thus store their keys verbatim in the block.

Refactor the code so that we can avoid decoding the keys, which saves us some allocations.

---
 reftable/block.c | 29 +++++++++++++++++++----------
 1 file changed, 19 insertions(+), 10 deletions(-)
reftable/block.c +19 -10
index ca217636ae..298e8c56b9 100644
@@ -287,23 +287,32 @@ static int restart_needle_less(size_t idx, void *_args)
 		.buf = args->reader->block.data + off,
 		.len = args->reader->block_len - off,
 	};
-	struct strbuf kth_restart_key = STRBUF_INIT;
-	uint8_t unused_extra;
-	int result, n;
+	uint64_t prefix_len, suffix_len;
+	uint8_t extra;
+	int n;
 
 	/*
-	 * TODO: The restart key is verbatim in the block, so we can in theory
-	 * avoid decoding the key and thus save some allocations.
+	 * Records at restart points are stored without prefix compression, so
+	 * there is no need to fully decode the record key here. This removes
+	 * the need for allocating memory.
 	 */
-	n = reftable_decode_key(&kth_restart_key, &unused_extra, in);
-	if (n < 0) {
+	n = reftable_decode_keylen(in, &prefix_len, &suffix_len, &extra);
+	if (n < 0 || prefix_len) {
 		args->error = 1;
 		return -1;
 	}
 
-	result = strbuf_cmp(&args->needle, &kth_restart_key);
-	strbuf_release(&kth_restart_key);
-	return result < 0;
+	string_view_consume(&in, n);
+	if (suffix_len > in.len) {
+		args->error = 1;
+		return -1;
+	}
+
+	n = memcmp(args->needle.buf, in.buf,
+		   args->needle.len < suffix_len ? args->needle.len : suffix_len);
+	if (n)
+		return n < 0;
+	return args->needle.len < suffix_len;
 }
 
 void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
Signature · 1 line
2.44.GIT