1 =head1 NAME
3 Git - Perl interface to the Git version control system
5 =cut
8 package Git;
10 use strict;
13 BEGIN {
15 our ($VERSION, @ISA, @EXPORT, @EXPORT_OK);
17 # Totally unstable API.
18 $VERSION = '0.01';
21 =head1 SYNOPSIS
23 use Git;
25 my $version = Git::command_oneline('version');
27 git_cmd_try { Git::command_noisy('update-server-info') }
28 '%s failed w/ code %d';
30 my $repo = Git->repository (Directory => '/srv/git/cogito.git');
33 my @revs = $repo->command('rev-list', '--since=last monday', '--all');
35 my ($fh, $c) = $repo->command_output_pipe('rev-list', '--since=last monday', '--all');
36 my $lastrev = <$fh>; chomp $lastrev;
37 $repo->command_close_pipe($fh, $c);
39 my $lastrev = $repo->command_oneline( [ 'rev-list', '--all' ],
40 STDERR => 0 );
42 =cut
45 require Exporter;
47 @ISA = qw(Exporter);
49 @EXPORT = qw(git_cmd_try);
51 # Methods which can be called as standalone functions as well:
52 @EXPORT_OK = qw(command command_oneline command_noisy
53 command_output_pipe command_input_pipe command_close_pipe
54 version exec_path hash_object git_cmd_try);
57 =head1 DESCRIPTION
59 This module provides Perl scripts easy way to interface the Git version control
60 system. The modules have an easy and well-tested way to call arbitrary Git
61 commands; in the future, the interface will also provide specialized methods
62 for doing easily operations which are not totally trivial to do over
63 the generic command interface.
65 While some commands can be executed outside of any context (e.g. 'version'
66 or 'init-db'), most operations require a repository context, which in practice
67 means getting an instance of the Git object using the repository() constructor.
68 (In the future, we will also get a new_repository() constructor.) All commands
69 called as methods of the object are then executed in the context of the
70 repository.
72 Part of the "repository state" is also information about path to the attached
73 working copy (unless you work with a bare repository). You can also navigate
74 inside of the working copy using the C<wc_chdir()> method. (Note that
75 the repository object is self-contained and will not change working directory
76 of your process.)
78 TODO: In the future, we might also do
80 my $remoterepo = $repo->remote_repository (Name => 'cogito', Branch => 'master');
81 $remoterepo ||= Git->remote_repository ('http://git.or.cz/cogito.git/');
82 my @refs = $remoterepo->refs();
84 Currently, the module merely wraps calls to external Git tools. In the future,
85 it will provide a much faster way to interact with Git by linking directly
86 to libgit. This should be completely opaque to the user, though (performance
87 increate nonwithstanding).
89 =cut
92 use Carp qw(carp croak); # but croak is bad - throw instead
93 use Error qw(:try);
94 use Cwd qw(abs_path);
96 }
99 =head1 CONSTRUCTORS
101 =over 4
103 =item repository ( OPTIONS )
105 =item repository ( DIRECTORY )
107 =item repository ()
109 Construct a new repository object.
110 C<OPTIONS> are passed in a hash like fashion, using key and value pairs.
111 Possible options are:
113 B<Repository> - Path to the Git repository.
115 B<WorkingCopy> - Path to the associated working copy; not strictly required
116 as many commands will happily crunch on a bare repository.
118 B<WorkingSubdir> - Subdirectory in the working copy to work inside.
119 Just left undefined if you do not want to limit the scope of operations.
121 B<Directory> - Path to the Git working directory in its usual setup.
122 The C<.git> directory is searched in the directory and all the parent
123 directories; if found, C<WorkingCopy> is set to the directory containing
124 it and C<Repository> to the C<.git> directory itself. If no C<.git>
125 directory was found, the C<Directory> is assumed to be a bare repository,
126 C<Repository> is set to point at it and C<WorkingCopy> is left undefined.
127 If the C<$GIT_DIR> environment variable is set, things behave as expected
128 as well.
130 You should not use both C<Directory> and either of C<Repository> and
131 C<WorkingCopy> - the results of that are undefined.
133 Alternatively, a directory path may be passed as a single scalar argument
134 to the constructor; it is equivalent to setting only the C<Directory> option
135 field.
137 Calling the constructor with no options whatsoever is equivalent to
138 calling it with C<< Directory => '.' >>. In general, if you are building
139 a standard porcelain command, simply doing C<< Git->repository() >> should
140 do the right thing and setup the object to reflect exactly where the user
141 is right now.
143 =cut
145 sub repository {
146 my $class = shift;
147 my @args = @_;
148 my %opts = ();
149 my $self;
151 if (defined $args[0]) {
152 if ($#args % 2 != 1) {
153 # Not a hash.
154 $#args == 0 or throw Error::Simple("bad usage");
155 %opts = ( Directory => $args[0] );
156 } else {
157 %opts = @args;
158 }
159 }
161 if (not defined $opts{Repository} and not defined $opts{WorkingCopy}) {
162 $opts{Directory} ||= '.';
163 }
165 if ($opts{Directory}) {
166 -d $opts{Directory} or throw Error::Simple("Directory not found: $!");
168 my $search = Git->repository(WorkingCopy => $opts{Directory});
169 my $dir;
170 try {
171 $dir = $search->command_oneline(['rev-parse', '--git-dir'],
172 STDERR => 0);
173 } catch Git::Error::Command with {
174 $dir = undef;
175 };
177 if ($dir) {
178 $dir =~ m#^/# or $dir = $opts{Directory} . '/' . $dir;
179 $opts{Repository} = $dir;
181 # If --git-dir went ok, this shouldn't die either.
182 my $prefix = $search->command_oneline('rev-parse', '--show-prefix');
183 $dir = abs_path($opts{Directory}) . '/';
184 if ($prefix) {
185 if (substr($dir, -length($prefix)) ne $prefix) {
186 throw Error::Simple("rev-parse confused me - $dir does not have trailing $prefix");
187 }
188 substr($dir, -length($prefix)) = '';
189 }
190 $opts{WorkingCopy} = $dir;
191 $opts{WorkingSubdir} = $prefix;
193 } else {
194 # A bare repository? Let's see...
195 $dir = $opts{Directory};
197 unless (-d "$dir/refs" and -d "$dir/objects" and -e "$dir/HEAD") {
198 # Mimick git-rev-parse --git-dir error message:
199 throw Error::Simple('fatal: Not a git repository');
200 }
201 my $search = Git->repository(Repository => $dir);
202 try {
203 $search->command('symbolic-ref', 'HEAD');
204 } catch Git::Error::Command with {
205 # Mimick git-rev-parse --git-dir error message:
206 throw Error::Simple('fatal: Not a git repository');
207 }
209 $opts{Repository} = abs_path($dir);
210 }
212 delete $opts{Directory};
213 }
215 $self = { opts => \%opts };
216 bless $self, $class;
217 }
220 =back
222 =head1 METHODS
224 =over 4
226 =item command ( COMMAND [, ARGUMENTS... ] )
228 =item command ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
230 Execute the given Git C<COMMAND> (specify it without the 'git-'
231 prefix), optionally with the specified extra C<ARGUMENTS>.
233 The second more elaborate form can be used if you want to further adjust
234 the command execution. Currently, only one option is supported:
236 B<STDERR> - How to deal with the command's error output. By default (C<undef>)
237 it is delivered to the caller's C<STDERR>. A false value (0 or '') will cause
238 it to be thrown away. If you want to process it, you can get it in a filehandle
239 you specify, but you must be extremely careful; if the error output is not
240 very short and you want to read it in the same process as where you called
241 C<command()>, you are set up for a nice deadlock!
243 The method can be called without any instance or on a specified Git repository
244 (in that case the command will be run in the repository context).
246 In scalar context, it returns all the command output in a single string
247 (verbatim).
249 In array context, it returns an array containing lines printed to the
250 command's stdout (without trailing newlines).
252 In both cases, the command's stdin and stderr are the same as the caller's.
254 =cut
256 sub command {
257 my ($fh, $ctx) = command_output_pipe(@_);
259 if (not defined wantarray) {
260 # Nothing to pepper the possible exception with.
261 _cmd_close($fh, $ctx);
263 } elsif (not wantarray) {
264 local $/;
265 my $text = <$fh>;
266 try {
267 _cmd_close($fh, $ctx);
268 } catch Git::Error::Command with {
269 # Pepper with the output:
270 my $E = shift;
271 $E->{'-outputref'} = \$text;
272 throw $E;
273 };
274 return $text;
276 } else {
277 my @lines = <$fh>;
278 chomp @lines;
279 try {
280 _cmd_close($fh, $ctx);
281 } catch Git::Error::Command with {
282 my $E = shift;
283 $E->{'-outputref'} = \@lines;
284 throw $E;
285 };
286 return @lines;
287 }
288 }
291 =item command_oneline ( COMMAND [, ARGUMENTS... ] )
293 =item command_oneline ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
295 Execute the given C<COMMAND> in the same way as command()
296 does but always return a scalar string containing the first line
297 of the command's standard output.
299 =cut
301 sub command_oneline {
302 my ($fh, $ctx) = command_output_pipe(@_);
304 my $line = <$fh>;
305 defined $line and chomp $line;
306 try {
307 _cmd_close($fh, $ctx);
308 } catch Git::Error::Command with {
309 # Pepper with the output:
310 my $E = shift;
311 $E->{'-outputref'} = \$line;
312 throw $E;
313 };
314 return $line;
315 }
318 =item command_output_pipe ( COMMAND [, ARGUMENTS... ] )
320 =item command_output_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
322 Execute the given C<COMMAND> in the same way as command()
323 does but return a pipe filehandle from which the command output can be
324 read.
326 The function can return C<($pipe, $ctx)> in array context.
327 See C<command_close_pipe()> for details.
329 =cut
331 sub command_output_pipe {
332 _command_common_pipe('-|', @_);
333 }
336 =item command_input_pipe ( COMMAND [, ARGUMENTS... ] )
338 =item command_input_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
340 Execute the given C<COMMAND> in the same way as command_output_pipe()
341 does but return an input pipe filehandle instead; the command output
342 is not captured.
344 The function can return C<($pipe, $ctx)> in array context.
345 See C<command_close_pipe()> for details.
347 =cut
349 sub command_input_pipe {
350 _command_common_pipe('|-', @_);
351 }
354 =item command_close_pipe ( PIPE [, CTX ] )
356 Close the C<PIPE> as returned from C<command_*_pipe()>, checking
357 whether the command finished successfuly. The optional C<CTX> argument
358 is required if you want to see the command name in the error message,
359 and it is the second value returned by C<command_*_pipe()> when
360 called in array context. The call idiom is:
362 my ($fh, $ctx) = $r->command_output_pipe('status');
363 while (<$fh>) { ... }
364 $r->command_close_pipe($fh, $ctx);
366 Note that you should not rely on whatever actually is in C<CTX>;
367 currently it is simply the command name but in future the context might
368 have more complicated structure.
370 =cut
372 sub command_close_pipe {
373 my ($self, $fh, $ctx) = _maybe_self(@_);
374 $ctx ||= '<unknown>';
375 _cmd_close($fh, $ctx);
376 }
379 =item command_noisy ( COMMAND [, ARGUMENTS... ] )
381 Execute the given C<COMMAND> in the same way as command() does but do not
382 capture the command output - the standard output is not redirected and goes
383 to the standard output of the caller application.
385 While the method is called command_noisy(), you might want to as well use
386 it for the most silent Git commands which you know will never pollute your
387 stdout but you want to avoid the overhead of the pipe setup when calling them.
389 The function returns only after the command has finished running.
391 =cut
393 sub command_noisy {
394 my ($self, $cmd, @args) = _maybe_self(@_);
395 _check_valid_cmd($cmd);
397 my $pid = fork;
398 if (not defined $pid) {
399 throw Error::Simple("fork failed: $!");
400 } elsif ($pid == 0) {
401 _cmd_exec($self, $cmd, @args);
402 }
403 if (waitpid($pid, 0) > 0 and $?>>8 != 0) {
404 throw Git::Error::Command(join(' ', $cmd, @args), $? >> 8);
405 }
406 }
409 =item version ()
411 Return the Git version in use.
413 =cut
415 sub version {
416 my $verstr = command_oneline('--version');
417 $verstr =~ s/^git version //;
418 $verstr;
419 }
422 =item exec_path ()
424 Return path to the Git sub-command executables (the same as
425 C<git --exec-path>). Useful mostly only internally.
427 =cut
429 sub exec_path { command_oneline('--exec-path') }
432 =item repo_path ()
434 Return path to the git repository. Must be called on a repository instance.
436 =cut
438 sub repo_path { $_[0]->{opts}->{Repository} }
441 =item wc_path ()
443 Return path to the working copy. Must be called on a repository instance.
445 =cut
447 sub wc_path { $_[0]->{opts}->{WorkingCopy} }
450 =item wc_subdir ()
452 Return path to the subdirectory inside of a working copy. Must be called
453 on a repository instance.
455 =cut
457 sub wc_subdir { $_[0]->{opts}->{WorkingSubdir} ||= '' }
460 =item wc_chdir ( SUBDIR )
462 Change the working copy subdirectory to work within. The C<SUBDIR> is
463 relative to the working copy root directory (not the current subdirectory).
464 Must be called on a repository instance attached to a working copy
465 and the directory must exist.
467 =cut
469 sub wc_chdir {
470 my ($self, $subdir) = @_;
471 $self->wc_path()
472 or throw Error::Simple("bare repository");
474 -d $self->wc_path().'/'.$subdir
475 or throw Error::Simple("subdir not found: $!");
476 # Of course we will not "hold" the subdirectory so anyone
477 # can delete it now and we will never know. But at least we tried.
479 $self->{opts}->{WorkingSubdir} = $subdir;
480 }
483 =item config ( VARIABLE )
485 Retrieve the configuration C<VARIABLE> in the same manner as C<repo-config>
486 does. In scalar context requires the variable to be set only one time
487 (exception is thrown otherwise), in array context returns allows the
488 variable to be set multiple times and returns all the values.
490 Must be called on a repository instance.
492 This currently wraps command('repo-config') so it is not so fast.
494 =cut
496 sub config {
497 my ($self, $var) = @_;
498 $self->repo_path()
499 or throw Error::Simple("not a repository");
501 try {
502 if (wantarray) {
503 return $self->command('repo-config', '--get-all', $var);
504 } else {
505 return $self->command_oneline('repo-config', '--get', $var);
506 }
507 } catch Git::Error::Command with {
508 my $E = shift;
509 if ($E->value() == 1) {
510 # Key not found.
511 return undef;
512 } else {
513 throw $E;
514 }
515 };
516 }
519 =item ident ( TYPE | IDENTSTR )
521 =item ident_person ( TYPE | IDENTSTR | IDENTARRAY )
523 This suite of functions retrieves and parses ident information, as stored
524 in the commit and tag objects or produced by C<var GIT_type_IDENT> (thus
525 C<TYPE> can be either I<author> or I<committer>; case is insignificant).
527 The C<ident> method retrieves the ident information from C<git-var>
528 and either returns it as a scalar string or as an array with the fields parsed.
529 Alternatively, it can take a prepared ident string (e.g. from the commit
530 object) and just parse it.
532 C<ident_person> returns the person part of the ident - name and email;
533 it can take the same arguments as C<ident> or the array returned by C<ident>.
535 The synopsis is like:
537 my ($name, $email, $time_tz) = ident('author');
538 "$name <$email>" eq ident_person('author');
539 "$name <$email>" eq ident_person($name);
540 $time_tz =~ /^\d+ [+-]\d{4}$/;
542 Both methods must be called on a repository instance.
544 =cut
546 sub ident {
547 my ($self, $type) = @_;
548 my $identstr;
549 if (lc $type eq lc 'committer' or lc $type eq lc 'author') {
550 $identstr = $self->command_oneline('var', 'GIT_'.uc($type).'_IDENT');
551 } else {
552 $identstr = $type;
553 }
554 if (wantarray) {
555 return $identstr =~ /^(.*) <(.*)> (\d+ [+-]\d{4})$/;
556 } else {
557 return $identstr;
558 }
559 }
561 sub ident_person {
562 my ($self, @ident) = @_;
563 $#ident == 0 and @ident = $self->ident($ident[0]);
564 return "$ident[0] <$ident[1]>";
565 }
568 =item hash_object ( TYPE, FILENAME )
570 Compute the SHA1 object id of the given C<FILENAME> (or data waiting in
571 C<FILEHANDLE>) considering it is of the C<TYPE> object type (C<blob>,
572 C<commit>, C<tree>).
574 The method can be called without any instance or on a specified Git repository,
575 it makes zero difference.
577 The function returns the SHA1 hash.
579 =cut
581 # TODO: Support for passing FILEHANDLE instead of FILENAME
582 sub hash_object {
583 my ($self, $type, $file) = _maybe_self(@_);
584 command_oneline('hash-object', '-t', $type, $file);
585 }
589 =back
591 =head1 ERROR HANDLING
593 All functions are supposed to throw Perl exceptions in case of errors.
594 See the L<Error> module on how to catch those. Most exceptions are mere
595 L<Error::Simple> instances.
597 However, the C<command()>, C<command_oneline()> and C<command_noisy()>
598 functions suite can throw C<Git::Error::Command> exceptions as well: those are
599 thrown when the external command returns an error code and contain the error
600 code as well as access to the captured command's output. The exception class
601 provides the usual C<stringify> and C<value> (command's exit code) methods and
602 in addition also a C<cmd_output> method that returns either an array or a
603 string with the captured command output (depending on the original function
604 call context; C<command_noisy()> returns C<undef>) and $<cmdline> which
605 returns the command and its arguments (but without proper quoting).
607 Note that the C<command_*_pipe()> functions cannot throw this exception since
608 it has no idea whether the command failed or not. You will only find out
609 at the time you C<close> the pipe; if you want to have that automated,
610 use C<command_close_pipe()>, which can throw the exception.
612 =cut
614 {
615 package Git::Error::Command;
617 @Git::Error::Command::ISA = qw(Error);
619 sub new {
620 my $self = shift;
621 my $cmdline = '' . shift;
622 my $value = 0 + shift;
623 my $outputref = shift;
624 my(@args) = ();
626 local $Error::Depth = $Error::Depth + 1;
628 push(@args, '-cmdline', $cmdline);
629 push(@args, '-value', $value);
630 push(@args, '-outputref', $outputref);
632 $self->SUPER::new(-text => 'command returned error', @args);
633 }
635 sub stringify {
636 my $self = shift;
637 my $text = $self->SUPER::stringify;
638 $self->cmdline() . ': ' . $text . ': ' . $self->value() . "\n";
639 }
641 sub cmdline {
642 my $self = shift;
643 $self->{'-cmdline'};
644 }
646 sub cmd_output {
647 my $self = shift;
648 my $ref = $self->{'-outputref'};
649 defined $ref or undef;
650 if (ref $ref eq 'ARRAY') {
651 return @$ref;
652 } else { # SCALAR
653 return $$ref;
654 }
655 }
656 }
658 =over 4
660 =item git_cmd_try { CODE } ERRMSG
662 This magical statement will automatically catch any C<Git::Error::Command>
663 exceptions thrown by C<CODE> and make your program die with C<ERRMSG>
664 on its lips; the message will have %s substituted for the command line
665 and %d for the exit status. This statement is useful mostly for producing
666 more user-friendly error messages.
668 In case of no exception caught the statement returns C<CODE>'s return value.
670 Note that this is the only auto-exported function.
672 =cut
674 sub git_cmd_try(&$) {
675 my ($code, $errmsg) = @_;
676 my @result;
677 my $err;
678 my $array = wantarray;
679 try {
680 if ($array) {
681 @result = &$code;
682 } else {
683 $result[0] = &$code;
684 }
685 } catch Git::Error::Command with {
686 my $E = shift;
687 $err = $errmsg;
688 $err =~ s/\%s/$E->cmdline()/ge;
689 $err =~ s/\%d/$E->value()/ge;
690 # We can't croak here since Error.pm would mangle
691 # that to Error::Simple.
692 };
693 $err and croak $err;
694 return $array ? @result : $result[0];
695 }
698 =back
700 =head1 COPYRIGHT
702 Copyright 2006 by Petr Baudis E<lt>pasky@suse.czE<gt>.
704 This module is free software; it may be used, copied, modified
705 and distributed under the terms of the GNU General Public Licence,
706 either version 2, or (at your option) any later version.
708 =cut
711 # Take raw method argument list and return ($obj, @args) in case
712 # the method was called upon an instance and (undef, @args) if
713 # it was called directly.
714 sub _maybe_self {
715 # This breaks inheritance. Oh well.
716 ref $_[0] eq 'Git' ? @_ : (undef, @_);
717 }
719 # Check if the command id is something reasonable.
720 sub _check_valid_cmd {
721 my ($cmd) = @_;
722 $cmd =~ /^[a-z0-9A-Z_-]+$/ or throw Error::Simple("bad command: $cmd");
723 }
725 # Common backend for the pipe creators.
726 sub _command_common_pipe {
727 my $direction = shift;
728 my ($self, @p) = _maybe_self(@_);
729 my (%opts, $cmd, @args);
730 if (ref $p[0]) {
731 ($cmd, @args) = @{shift @p};
732 %opts = ref $p[0] ? %{$p[0]} : @p;
733 } else {
734 ($cmd, @args) = @p;
735 }
736 _check_valid_cmd($cmd);
738 my $fh;
739 if ($^O eq '##INSERT_ACTIVESTATE_STRING_HERE##') {
740 # ActiveState Perl
741 #defined $opts{STDERR} and
742 # warn 'ignoring STDERR option - running w/ ActiveState';
743 $direction eq '-|' or
744 die 'input pipe for ActiveState not implemented';
745 tie ($fh, 'Git::activestate_pipe', $cmd, @args);
747 } else {
748 my $pid = open($fh, $direction);
749 if (not defined $pid) {
750 throw Error::Simple("open failed: $!");
751 } elsif ($pid == 0) {
752 if (defined $opts{STDERR}) {
753 close STDERR;
754 }
755 if ($opts{STDERR}) {
756 open (STDERR, '>&', $opts{STDERR})
757 or die "dup failed: $!";
758 }
759 _cmd_exec($self, $cmd, @args);
760 }
761 }
762 return wantarray ? ($fh, join(' ', $cmd, @args)) : $fh;
763 }
765 # When already in the subprocess, set up the appropriate state
766 # for the given repository and execute the git command.
767 sub _cmd_exec {
768 my ($self, @args) = @_;
769 if ($self) {
770 $self->repo_path() and $ENV{'GIT_DIR'} = $self->repo_path();
771 $self->wc_path() and chdir($self->wc_path());
772 $self->wc_subdir() and chdir($self->wc_subdir());
773 }
774 _execv_git_cmd(@args);
775 die "exec failed: $!";
776 }
778 # Execute the given Git command ($_[0]) with arguments ($_[1..])
779 # by searching for it at proper places.
780 sub _execv_git_cmd { exec('git', @_); }
782 # Close pipe to a subprocess.
783 sub _cmd_close {
784 my ($fh, $ctx) = @_;
785 if (not close $fh) {
786 if ($!) {
787 # It's just close, no point in fatalities
788 carp "error closing pipe: $!";
789 } elsif ($? >> 8) {
790 # The caller should pepper this.
791 throw Git::Error::Command($ctx, $? >> 8);
792 }
793 # else we might e.g. closed a live stream; the command
794 # dying of SIGPIPE would drive us here.
795 }
796 }
799 sub DESTROY { }
802 # Pipe implementation for ActiveState Perl.
804 package Git::activestate_pipe;
805 use strict;
807 sub TIEHANDLE {
808 my ($class, @params) = @_;
809 # FIXME: This is probably horrible idea and the thing will explode
810 # at the moment you give it arguments that require some quoting,
811 # but I have no ActiveState clue... --pasky
812 my $cmdline = join " ", @params;
813 my @data = qx{$cmdline};
814 bless { i => 0, data => \@data }, $class;
815 }
817 sub READLINE {
818 my $self = shift;
819 if ($self->{i} >= scalar @{$self->{data}}) {
820 return undef;
821 }
822 return $self->{'data'}->[ $self->{i}++ ];
823 }
825 sub CLOSE {
826 my $self = shift;
827 delete $self->{data};
828 delete $self->{i};
829 }
831 sub EOF {
832 my $self = shift;
833 return ($self->{i} >= scalar @{$self->{data}});
834 }
837 1; # Famous last words