copy selected history between repostories

Hi All,

This is something I'm not sure how to do with git, or even if it is possible in git; and because google has not been my friend this time, I though I would ask here.

I have 2 repositories: repoA and repoB.

repoB is massive. lots of history and files. But it has a set of files inside a particular dir with 40-60 files that I just realized are better in repoA.

Is there any way to import that directory into repoA with all it's history, but NOT the history from the other commits?
Another way of putting is this: can I make git forget the history of all other commits but those from this directory?

View this message in context: http://www.nabble.com/copy-selected-history-between-repostories-tp18533605p18533605.html
Sent from the git mailing list archive at Nabble.com.

I couldn't figure out git-filter-branch, so I did the following, which worked for me:

First, export commits as patches in mbox format:

	git log -p --first-parent --reverse --pretty=email $* > mbox

Second, run the mbox file through a filter which selectively renames files (since almost always I want the files in a different directory in the new repository):

myfilter scripts=bin perllib=lib < mbox > mbox2

Finally, import into new repo:

git-am mbox2

It's a bit slow on the import step but it meets my needs at this time.

Nick.

Filter script follows, for what it's worth ...

#!/usr/bin/perl
# @(#) git-filter-rename.pl
#
# Read an mbox file containing patches on standard input,
# modify the filenames within according to a list of substitutions
# supplied on the command line, and write the modified mbox file
# to standard output.

my @subs;

foreach (@ARGV) {
	if (/(.*)=(.*)/) {
		my $len = length($1);
		push(@subs, [ $1, $2, $len ]);
	} else {
		print STDERR "Unknown arg: $_\n";
	}
}

while (<STDIN>) {
	chomp;
	my $line = $_;

	if ($line =~ /^(---|\+\+\+) ([ab]\/)(.+)$/) {
		my $line_1 = $1;
		my $line_2 = $2;
		my $line_3 = $3;

		subFilename($line_3);

		print "$line_1 $line_2$line_3\n";
		next;
	}

	if ($line =~ /^diff --git ([ab]\/)(\S+) ([ab]\/)(\S+)/) {
		my ($line_1, $line_2, $line_3, $line_4) = ($1, $2, $3, $4);
		subFilename($line_2);
		subFilename($line_4);
		print "diff --git $line_1$line_2 $line_3$line_4\n";
		next;
	}

	print $line, "\n";
}

exit(0);

# ------------------------------------------------------------------------ # Substitute a filename in-place (modifies argument)
# ------------------------------------------------------------------------

sub subFilename {

	foreach my $lr (@subs) {
		my ($lhs, $rhs, $len) = @$lr;
		if (substr($_[0], 0, $len) eq $lhs) {
			print STDERR "Match on $lhs, $_[0]\n";
			substr($_[0], 0, $len) = $rhs;
			last;
		}
	}
}

Note that this will lose any interesting history topology you had, like branching and merging (and it may have trouble applying all patches in that case, too). But if your history was a straight line, it should produce equivalent results.

-Peff

Hi,

Note that if your repository only contains text files, "git fast-export | my-filter | git fast-import --force" would have worked faster, probably.

But I am glad you figured out how to do what you wanted to do.

Ciao,
Dscho

Try cloning the repository, then running git-filter-branch with the --subdirectory-filter option. It works quite nicely.

Have fun,

Avery