3 Git - Perl interface to the Git version control system
12 use warnings $ENV{GIT_PERL_FATAL_WARNINGS} ? qw(FATAL all) : ();
19 our ($VERSION, @ISA, @EXPORT, @EXPORT_OK);
21 # Totally unstable API.
29 my $version = Git::command_oneline('version');
31 git_cmd_try { Git::command_noisy('update-server-info') }
32 '%s failed w/ code %d';
34 my $repo = Git->repository (Directory => '/srv/git/cogito.git');
37 my @revs = $repo->command('rev-list', '--since=last monday', '--all');
39 my ($fh, $c) = $repo->command_output_pipe('rev-list', '--since=last monday', '--all');
40 my $lastrev = <$fh>; chomp $lastrev;
41 $repo->command_close_pipe($fh, $c);
43 my $lastrev = $repo->command_oneline( [ 'rev-list', '--all' ],
46 my $sha1 = $repo->hash_and_insert_object('file.txt');
47 my $tempfile = tempfile();
48 my $size = $repo->cat_blob($sha1, $tempfile);
57 @EXPORT = qw(git_cmd_try);
59 # Methods which can be called as standalone functions as well:
60 @EXPORT_OK = qw(command command_oneline command_noisy
61 command_output_pipe command_input_pipe command_close_pipe
62 command_bidi_pipe command_close_bidi_pipe
63 version exec_path html_path hash_object git_cmd_try
65 get_tz_offset get_record
66 credential credential_read credential_write
67 temp_acquire temp_is_locked temp_release temp_reset temp_path
73 This module provides Perl scripts easy way to interface the Git version control
74 system. The modules have an easy and well-tested way to call arbitrary Git
75 commands; in the future, the interface will also provide specialized methods
76 for doing easily operations which are not totally trivial to do over
77 the generic command interface.
79 While some commands can be executed outside of any context (e.g. 'version'
80 or 'init'), most operations require a repository context, which in practice
81 means getting an instance of the Git object using the repository() constructor.
82 (In the future, we will also get a new_repository() constructor.) All commands
83 called as methods of the object are then executed in the context of the
86 Part of the "repository state" is also information about path to the attached
87 working copy (unless you work with a bare repository). You can also navigate
88 inside of the working copy using the C<wc_chdir()> method. (Note that
89 the repository object is self-contained and will not change working directory
92 TODO: In the future, we might also do
94 my $remoterepo = $repo->remote_repository (Name => 'cogito', Branch => 'master');
95 $remoterepo ||= Git->remote_repository ('http://git.or.cz/cogito.git/');
96 my @refs = $remoterepo->refs();
98 Currently, the module merely wraps calls to external Git tools. In the future,
99 it will provide a much faster way to interact with Git by linking directly
100 to libgit. This should be completely opaque to the user, though (performance
101 increase notwithstanding).
106 use Carp qw(carp croak); # but croak is bad - throw instead
107 use Git::LoadCPAN::Error qw(:try);
108 use Cwd qw(abs_path cwd);
109 use IPC::Open2 qw(open2);
110 use Fcntl qw(SEEK_SET SEEK_CUR);
111 use Time::Local qw(timegm);
119 =item repository ( OPTIONS )
121 =item repository ( DIRECTORY )
125 Construct a new repository object.
126 C<OPTIONS> are passed in a hash like fashion, using key and value pairs.
127 Possible options are:
129 B<Repository> - Path to the Git repository.
131 B<WorkingCopy> - Path to the associated working copy; not strictly required
132 as many commands will happily crunch on a bare repository.
134 B<WorkingSubdir> - Subdirectory in the working copy to work inside.
135 Just left undefined if you do not want to limit the scope of operations.
137 B<Directory> - Path to the Git working directory in its usual setup.
138 The C<.git> directory is searched in the directory and all the parent
139 directories; if found, C<WorkingCopy> is set to the directory containing
140 it and C<Repository> to the C<.git> directory itself. If no C<.git>
141 directory was found, the C<Directory> is assumed to be a bare repository,
142 C<Repository> is set to point at it and C<WorkingCopy> is left undefined.
143 If the C<$GIT_DIR> environment variable is set, things behave as expected
146 You should not use both C<Directory> and either of C<Repository> and
147 C<WorkingCopy> - the results of that are undefined.
149 Alternatively, a directory path may be passed as a single scalar argument
150 to the constructor; it is equivalent to setting only the C<Directory> option
153 Calling the constructor with no options whatsoever is equivalent to
154 calling it with C<< Directory => '.' >>. In general, if you are building
155 a standard porcelain command, simply doing C<< Git->repository() >> should
156 do the right thing and setup the object to reflect exactly where the user
167 if (defined $args[0]) {
168 if ($#args % 2 != 1) {
170 $#args == 0 or throw Error::Simple("bad usage");
171 %opts = ( Directory => $args[0] );
177 if (not defined $opts{Repository} and not defined $opts{WorkingCopy}
178 and not defined $opts{Directory}) {
179 $opts{Directory} = '.';
182 if (defined $opts{Directory}) {
183 -d $opts{Directory} or throw Error::Simple("Directory not found: $opts{Directory} $!");
185 my $search = Git->repository(WorkingCopy => $opts{Directory});
188 $dir = $search->command_oneline(['rev-parse', '--git-dir'],
190 } catch Git::Error::Command with {
195 File::Spec->file_name_is_absolute($dir) or $dir = $opts{Directory} . '/' . $dir;
196 $opts{Repository} = abs_path($dir);
198 # If --git-dir went ok, this shouldn't die either.
199 my $prefix = $search->command_oneline('rev-parse', '--show-prefix');
200 $dir = abs_path($opts{Directory}) . '/';
202 if (substr($dir, -length($prefix)) ne $prefix) {
203 throw Error::Simple("rev-parse confused me - $dir does not have trailing $prefix");
205 substr($dir, -length($prefix)) = '';
207 $opts{WorkingCopy} = $dir;
208 $opts{WorkingSubdir} = $prefix;
211 # A bare repository? Let's see...
212 $dir = $opts{Directory};
214 unless (-d "$dir/refs" and -d "$dir/objects" and -e "$dir/HEAD") {
215 # Mimic git-rev-parse --git-dir error message:
216 throw Error::Simple("fatal: Not a git repository: $dir");
218 my $search = Git->repository(Repository => $dir);
220 $search->command('symbolic-ref', 'HEAD');
221 } catch Git::Error::Command with {
222 # Mimic git-rev-parse --git-dir error message:
223 throw Error::Simple("fatal: Not a git repository: $dir");
226 $opts{Repository} = abs_path($dir);
229 delete $opts{Directory};
232 $self = { opts => \%opts };
242 =item command ( COMMAND [, ARGUMENTS... ] )
244 =item command ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
246 Execute the given Git C<COMMAND> (specify it without the 'git-'
247 prefix), optionally with the specified extra C<ARGUMENTS>.
249 The second more elaborate form can be used if you want to further adjust
250 the command execution. Currently, only one option is supported:
252 B<STDERR> - How to deal with the command's error output. By default (C<undef>)
253 it is delivered to the caller's C<STDERR>. A false value (0 or '') will cause
254 it to be thrown away. If you want to process it, you can get it in a filehandle
255 you specify, but you must be extremely careful; if the error output is not
256 very short and you want to read it in the same process as where you called
257 C<command()>, you are set up for a nice deadlock!
259 The method can be called without any instance or on a specified Git repository
260 (in that case the command will be run in the repository context).
262 In scalar context, it returns all the command output in a single string
265 In array context, it returns an array containing lines printed to the
266 command's stdout (without trailing newlines).
268 In both cases, the command's stdin and stderr are the same as the caller's.
273 my ($fh, $ctx) = command_output_pipe(@_);
275 if (not defined wantarray) {
276 # Nothing to pepper the possible exception with.
277 _cmd_close($ctx, $fh);
279 } elsif (not wantarray) {
283 _cmd_close($ctx, $fh);
284 } catch Git::Error::Command with {
285 # Pepper with the output:
287 $E->{'-outputref'} = \$text;
294 defined and chomp for @lines;
296 _cmd_close($ctx, $fh);
297 } catch Git::Error::Command with {
299 $E->{'-outputref'} = \@lines;
307 =item command_oneline ( COMMAND [, ARGUMENTS... ] )
309 =item command_oneline ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
311 Execute the given C<COMMAND> in the same way as command()
312 does but always return a scalar string containing the first line
313 of the command's standard output.
317 sub command_oneline {
318 my ($fh, $ctx) = command_output_pipe(@_);
321 defined $line and chomp $line;
323 _cmd_close($ctx, $fh);
324 } catch Git::Error::Command with {
325 # Pepper with the output:
327 $E->{'-outputref'} = \$line;
334 =item command_output_pipe ( COMMAND [, ARGUMENTS... ] )
336 =item command_output_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
338 Execute the given C<COMMAND> in the same way as command()
339 does but return a pipe filehandle from which the command output can be
342 The function can return C<($pipe, $ctx)> in array context.
343 See C<command_close_pipe()> for details.
347 sub command_output_pipe {
348 _command_common_pipe('-|', @_);
352 =item command_input_pipe ( COMMAND [, ARGUMENTS... ] )
354 =item command_input_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
356 Execute the given C<COMMAND> in the same way as command_output_pipe()
357 does but return an input pipe filehandle instead; the command output
360 The function can return C<($pipe, $ctx)> in array context.
361 See C<command_close_pipe()> for details.
365 sub command_input_pipe {
366 _command_common_pipe('|-', @_);
370 =item command_close_pipe ( PIPE [, CTX ] )
372 Close the C<PIPE> as returned from C<command_*_pipe()>, checking
373 whether the command finished successfully. The optional C<CTX> argument
374 is required if you want to see the command name in the error message,
375 and it is the second value returned by C<command_*_pipe()> when
376 called in array context. The call idiom is:
378 my ($fh, $ctx) = $r->command_output_pipe('status');
379 while (<$fh>) { ... }
380 $r->command_close_pipe($fh, $ctx);
382 Note that you should not rely on whatever actually is in C<CTX>;
383 currently it is simply the command name but in future the context might
384 have more complicated structure.
388 sub command_close_pipe {
389 my ($self, $fh, $ctx) = _maybe_self(@_);
390 $ctx ||= '<unknown>';
391 _cmd_close($ctx, $fh);
394 =item command_bidi_pipe ( COMMAND [, ARGUMENTS... ] )
396 Execute the given C<COMMAND> in the same way as command_output_pipe()
397 does but return both an input pipe filehandle and an output pipe filehandle.
399 The function will return C<($pid, $pipe_in, $pipe_out, $ctx)>.
400 See C<command_close_bidi_pipe()> for details.
404 sub command_bidi_pipe {
405 my ($pid, $in, $out);
406 my ($self) = _maybe_self(@_);
408 my $cwd_save = undef;
412 _setup_git_cmd_env($self);
414 $pid = open2($in, $out, 'git', @_);
415 chdir($cwd_save) if $cwd_save;
416 return ($pid, $in, $out, join(' ', @_));
419 =item command_close_bidi_pipe ( PID, PIPE_IN, PIPE_OUT [, CTX] )
421 Close the C<PIPE_IN> and C<PIPE_OUT> as returned from C<command_bidi_pipe()>,
422 checking whether the command finished successfully. The optional C<CTX>
423 argument is required if you want to see the command name in the error message,
424 and it is the fourth value returned by C<command_bidi_pipe()>. The call idiom
427 my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
428 print $out "000000000\n";
429 while (<$in>) { ... }
430 $r->command_close_bidi_pipe($pid, $in, $out, $ctx);
432 Note that you should not rely on whatever actually is in C<CTX>;
433 currently it is simply the command name but in future the context might
434 have more complicated structure.
436 C<PIPE_IN> and C<PIPE_OUT> may be C<undef> if they have been closed prior to
437 calling this function. This may be useful in a query-response type of
438 commands where caller first writes a query and later reads response, eg:
440 my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
441 print $out "000000000\n";
443 while (<$in>) { ... }
444 $r->command_close_bidi_pipe($pid, $in, undef, $ctx);
446 This idiom may prevent potential dead locks caused by data sent to the output
447 pipe not being flushed and thus not reaching the executed command.
451 sub command_close_bidi_pipe {
453 my ($self, $pid, $in, $out, $ctx) = _maybe_self(@_);
454 _cmd_close($ctx, (grep { defined } ($in, $out)));
457 throw Git::Error::Command($ctx, $? >>8);
462 =item command_noisy ( COMMAND [, ARGUMENTS... ] )
464 Execute the given C<COMMAND> in the same way as command() does but do not
465 capture the command output - the standard output is not redirected and goes
466 to the standard output of the caller application.
468 While the method is called command_noisy(), you might want to as well use
469 it for the most silent Git commands which you know will never pollute your
470 stdout but you want to avoid the overhead of the pipe setup when calling them.
472 The function returns only after the command has finished running.
477 my ($self, $cmd, @args) = _maybe_self(@_);
478 _check_valid_cmd($cmd);
481 if (not defined $pid) {
482 throw Error::Simple("fork failed: $!");
483 } elsif ($pid == 0) {
484 _cmd_exec($self, $cmd, @args);
486 if (waitpid($pid, 0) > 0 and $?>>8 != 0) {
487 throw Git::Error::Command(join(' ', $cmd, @args), $? >> 8);
494 Return the Git version in use.
499 my $verstr = command_oneline('--version');
500 $verstr =~ s/^git version //;
507 Return path to the Git sub-command executables (the same as
508 C<git --exec-path>). Useful mostly only internally.
512 sub exec_path { command_oneline('--exec-path') }
517 Return path to the Git html documentation (the same as
518 C<git --html-path>). Useful mostly only internally.
522 sub html_path { command_oneline('--html-path') }
525 =item get_tz_offset ( TIME )
527 Return the time zone offset from GMT in the form +/-HHMM where HH is
528 the number of hours from GMT and MM is the number of minutes. This is
529 the equivalent of what strftime("%z", ...) would provide on a GNU
532 If TIME is not supplied, the current local time is used.
537 # some systems don't handle or mishandle %z, so be creative.
538 my $t = shift || time;
539 my @t = localtime($t);
542 my $sign = qw( + + - )[ $gm <=> $t ];
543 return sprintf("%s%02d%02d", $sign, (gmtime(abs($t - $gm)))[2,1]);
546 =item get_record ( FILEHANDLE, INPUT_RECORD_SEPARATOR )
548 Read one record from FILEHANDLE delimited by INPUT_RECORD_SEPARATOR,
549 removing any trailing INPUT_RECORD_SEPARATOR.
557 chomp $rec if defined $rec;
561 =item prompt ( PROMPT , ISPASSWORD )
563 Query user C<PROMPT> and return answer from user.
565 Honours GIT_ASKPASS and SSH_ASKPASS environment variables for querying
566 the user. If no *_ASKPASS variable is set or an error occurred,
567 the terminal is tried as a fallback.
568 If C<ISPASSWORD> is set and true, the terminal disables echo.
573 my ($prompt, $isPassword) = @_;
575 if (exists $ENV{'GIT_ASKPASS'}) {
576 $ret = _prompt($ENV{'GIT_ASKPASS'}, $prompt);
578 if (!defined $ret && exists $ENV{'SSH_ASKPASS'}) {
579 $ret = _prompt($ENV{'SSH_ASKPASS'}, $prompt);
582 print STDERR $prompt;
584 if (defined $isPassword && $isPassword) {
585 require Term::ReadKey;
586 Term::ReadKey::ReadMode('noecho');
588 while (defined(my $key = Term::ReadKey::ReadKey(0))) {
589 last if $key =~ /[\012\015]/; # \n\r
592 Term::ReadKey::ReadMode('restore');
596 chomp($ret = <STDIN>);
603 my ($askpass, $prompt) = @_;
604 return unless length $askpass;
607 open my $fh, "-|", $askpass, $prompt or return;
609 $ret =~ s/[\015\012]//g; # strip \r\n, chomp does not work on all systems (i.e. windows) as expected
616 Return path to the git repository. Must be called on a repository instance.
620 sub repo_path { $_[0]->{opts}->{Repository} }
625 Return path to the working copy. Must be called on a repository instance.
629 sub wc_path { $_[0]->{opts}->{WorkingCopy} }
634 Return path to the subdirectory inside of a working copy. Must be called
635 on a repository instance.
639 sub wc_subdir { $_[0]->{opts}->{WorkingSubdir} ||= '' }
642 =item wc_chdir ( SUBDIR )
644 Change the working copy subdirectory to work within. The C<SUBDIR> is
645 relative to the working copy root directory (not the current subdirectory).
646 Must be called on a repository instance attached to a working copy
647 and the directory must exist.
652 my ($self, $subdir) = @_;
654 or throw Error::Simple("bare repository");
656 -d $self->wc_path().'/'.$subdir
657 or throw Error::Simple("subdir not found: $subdir $!");
658 # Of course we will not "hold" the subdirectory so anyone
659 # can delete it now and we will never know. But at least we tried.
661 $self->{opts}->{WorkingSubdir} = $subdir;
665 =item config ( VARIABLE )
667 Retrieve the configuration C<VARIABLE> in the same manner as C<config>
668 does. In scalar context requires the variable to be set only one time
669 (exception is thrown otherwise), in array context returns allows the
670 variable to be set multiple times and returns all the values.
675 return _config_common({}, @_);
679 =item config_bool ( VARIABLE )
681 Retrieve the bool configuration C<VARIABLE>. The return value
682 is usable as a boolean in perl (and C<undef> if it's not defined,
688 my $val = scalar _config_common({'kind' => '--bool'}, @_);
690 # Do not rewrite this as return (defined $val && $val eq 'true')
691 # as some callers do care what kind of falsehood they receive.
695 return $val eq 'true';
700 =item config_path ( VARIABLE )
702 Retrieve the path configuration C<VARIABLE>. The return value
703 is an expanded path or C<undef> if it's not defined.
708 return _config_common({'kind' => '--path'}, @_);
712 =item config_int ( VARIABLE )
714 Retrieve the integer configuration C<VARIABLE>. The return value
715 is simple decimal number. An optional value suffix of 'k', 'm',
716 or 'g' in the config file will cause the value to be multiplied
717 by 1024, 1048576 (1024^2), or 1073741824 (1024^3) prior to output.
718 It would return C<undef> if configuration variable is not defined.
723 return scalar _config_common({'kind' => '--int'}, @_);
726 =item config_regexp ( RE )
728 Retrieve the list of configuration key names matching the regular
729 expression C<RE>. The return value is a list of strings matching
735 my ($self, $regex) = _maybe_self(@_);
737 my @cmd = ('config', '--name-only', '--get-regexp', $regex);
738 unshift @cmd, $self if $self;
739 my @matches = command(@cmd);
741 } catch Git::Error::Command with {
743 if ($E->value() == 1) {
752 # Common subroutine to implement bulk of what the config* family of methods
753 # do. This currently wraps command('config') so it is not so fast.
755 my ($opts) = shift @_;
756 my ($self, $var) = _maybe_self(@_);
759 my @cmd = ('config', $opts->{'kind'} ? $opts->{'kind'} : ());
760 unshift @cmd, $self if $self;
762 return command(@cmd, '--get-all', $var);
764 return command_oneline(@cmd, '--get', $var);
766 } catch Git::Error::Command with {
768 if ($E->value() == 1) {
777 =item get_colorbool ( NAME )
779 Finds if color should be used for NAMEd operation from the configuration,
780 and returns boolean (true for "use color", false for "do not use color").
785 my ($self, $var) = @_;
786 my $stdout_to_tty = (-t STDOUT) ? "true" : "false";
787 my $use_color = $self->command_oneline('config', '--get-colorbool',
788 $var, $stdout_to_tty);
789 return ($use_color eq 'true');
792 =item get_color ( SLOT, COLOR )
794 Finds color for SLOT from the configuration, while defaulting to COLOR,
795 and returns the ANSI color escape sequence:
797 print $repo->get_color("color.interactive.prompt", "underline blue white");
799 print $repo->get_color("", "normal");
804 my ($self, $slot, $default) = @_;
805 my $color = $self->command_oneline('config', '--get-color', $slot, $default);
806 if (!defined $color) {
812 =item remote_refs ( REPOSITORY [, GROUPS [, REFGLOBS ] ] )
814 This function returns a hashref of refs stored in a given remote repository.
815 The hash is in the format C<refname =\> hash>. For tags, the C<refname> entry
816 contains the tag object while a C<refname^{}> entry gives the tagged objects.
818 C<REPOSITORY> has the same meaning as the appropriate C<git-ls-remote>
819 argument; either a URL or a remote name (if called on a repository instance).
820 C<GROUPS> is an optional arrayref that can contain 'tags' to return all the
821 tags and/or 'heads' to return all the heads. C<REFGLOB> is an optional array
822 of strings containing a shell-like glob to further limit the refs returned in
823 the hash; the meaning is again the same as the appropriate C<git-ls-remote>
826 This function may or may not be called on a repository instance. In the former
827 case, remote names as defined in the repository are recognized as repository
833 my ($self, $repo, $groups, $refglobs) = _maybe_self(@_);
835 if (ref $groups eq 'ARRAY') {
838 push (@args, '--heads');
839 } elsif ($_ eq 'tags') {
840 push (@args, '--tags');
842 # Ignore unknown groups for future
848 if (ref $refglobs eq 'ARRAY') {
849 push (@args, @$refglobs);
852 my @self = $self ? ($self) : (); # Ultra trickery
853 my ($fh, $ctx) = Git::command_output_pipe(@self, 'ls-remote', @args);
857 my ($hash, $ref) = split(/\t/, $_, 2);
860 Git::command_close_pipe(@self, $fh, $ctx);
865 =item ident ( TYPE | IDENTSTR )
867 =item ident_person ( TYPE | IDENTSTR | IDENTARRAY )
869 This suite of functions retrieves and parses ident information, as stored
870 in the commit and tag objects or produced by C<var GIT_type_IDENT> (thus
871 C<TYPE> can be either I<author> or I<committer>; case is insignificant).
873 The C<ident> method retrieves the ident information from C<git var>
874 and either returns it as a scalar string or as an array with the fields parsed.
875 Alternatively, it can take a prepared ident string (e.g. from the commit
876 object) and just parse it.
878 C<ident_person> returns the person part of the ident - name and email;
879 it can take the same arguments as C<ident> or the array returned by C<ident>.
881 The synopsis is like:
883 my ($name, $email, $time_tz) = ident('author');
884 "$name <$email>" eq ident_person('author');
885 "$name <$email>" eq ident_person($name);
886 $time_tz =~ /^\d+ [+-]\d{4}$/;
891 my ($self, $type) = _maybe_self(@_);
893 if (lc $type eq lc 'committer' or lc $type eq lc 'author') {
894 my @cmd = ('var', 'GIT_'.uc($type).'_IDENT');
895 unshift @cmd, $self if $self;
896 $identstr = command_oneline(@cmd);
901 return $identstr =~ /^(.*) <(.*)> (\d+ [+-]\d{4})$/;
908 my ($self, @ident) = _maybe_self(@_);
909 $#ident == 0 and @ident = $self ? $self->ident($ident[0]) : ident($ident[0]);
910 return "$ident[0] <$ident[1]>";
913 =item hash_object ( TYPE, FILENAME )
915 Compute the SHA1 object id of the given C<FILENAME> considering it is
916 of the C<TYPE> object type (C<blob>, C<commit>, C<tree>).
918 The method can be called without any instance or on a specified Git repository,
919 it makes zero difference.
921 The function returns the SHA1 hash.
925 # TODO: Support for passing FILEHANDLE instead of FILENAME
927 my ($self, $type, $file) = _maybe_self(@_);
928 command_oneline('hash-object', '-t', $type, $file);
932 =item hash_and_insert_object ( FILENAME )
934 Compute the SHA1 object id of the given C<FILENAME> and add the object to the
937 The function returns the SHA1 hash.
941 # TODO: Support for passing FILEHANDLE instead of FILENAME
942 sub hash_and_insert_object {
943 my ($self, $filename) = @_;
945 carp "Bad filename \"$filename\"" if $filename =~ /[\r\n]/;
947 $self->_open_hash_and_insert_object_if_needed();
948 my ($in, $out) = ($self->{hash_object_in}, $self->{hash_object_out});
950 unless (print $out $filename, "\n") {
951 $self->_close_hash_and_insert_object();
952 throw Error::Simple("out pipe went bad");
955 chomp(my $hash = <$in>);
956 unless (defined($hash)) {
957 $self->_close_hash_and_insert_object();
958 throw Error::Simple("in pipe went bad");
964 sub _open_hash_and_insert_object_if_needed {
967 return if defined($self->{hash_object_pid});
969 ($self->{hash_object_pid}, $self->{hash_object_in},
970 $self->{hash_object_out}, $self->{hash_object_ctx}) =
971 $self->command_bidi_pipe(qw(hash-object -w --stdin-paths --no-filters));
974 sub _close_hash_and_insert_object {
977 return unless defined($self->{hash_object_pid});
979 my @vars = map { 'hash_object_' . $_ } qw(pid in out ctx);
981 command_close_bidi_pipe(@$self{@vars});
982 delete @$self{@vars};
985 =item cat_blob ( SHA1, FILEHANDLE )
987 Prints the contents of the blob identified by C<SHA1> to C<FILEHANDLE> and
988 returns the number of bytes printed.
993 my ($self, $sha1, $fh) = @_;
995 $self->_open_cat_blob_if_needed();
996 my ($in, $out) = ($self->{cat_blob_in}, $self->{cat_blob_out});
998 unless (print $out $sha1, "\n") {
999 $self->_close_cat_blob();
1000 throw Error::Simple("out pipe went bad");
1003 my $description = <$in>;
1004 if ($description =~ / missing$/) {
1005 carp "$sha1 doesn't exist in the repository";
1009 if ($description !~ /^[0-9a-fA-F]{40}(?:[0-9a-fA-F]{24})? \S+ (\d+)$/) {
1010 carp "Unexpected result returned from git cat-file";
1017 my $bytesLeft = $size;
1020 last unless $bytesLeft;
1022 my $bytesToRead = $bytesLeft < 1024 ? $bytesLeft : 1024;
1023 my $read = read($in, $blob, $bytesToRead);
1024 unless (defined($read)) {
1025 $self->_close_cat_blob();
1026 throw Error::Simple("in pipe went bad");
1028 unless (print $fh $blob) {
1029 $self->_close_cat_blob();
1030 throw Error::Simple("couldn't write to passed in filehandle");
1032 $bytesLeft -= $read;
1035 # Skip past the trailing newline.
1037 my $read = read($in, $newline, 1);
1038 unless (defined($read)) {
1039 $self->_close_cat_blob();
1040 throw Error::Simple("in pipe went bad");
1042 unless ($read == 1 && $newline eq "\n") {
1043 $self->_close_cat_blob();
1044 throw Error::Simple("didn't find newline after blob");
1050 sub _open_cat_blob_if_needed {
1053 return if defined($self->{cat_blob_pid});
1055 ($self->{cat_blob_pid}, $self->{cat_blob_in},
1056 $self->{cat_blob_out}, $self->{cat_blob_ctx}) =
1057 $self->command_bidi_pipe(qw(cat-file --batch));
1060 sub _close_cat_blob {
1063 return unless defined($self->{cat_blob_pid});
1065 my @vars = map { 'cat_blob_' . $_ } qw(pid in out ctx);
1067 command_close_bidi_pipe(@$self{@vars});
1068 delete @$self{@vars};
1072 =item credential_read( FILEHANDLE )
1074 Reads credential key-value pairs from C<FILEHANDLE>. Reading stops at EOF or
1075 when an empty line is encountered. Each line must be of the form C<key=value>
1076 with a non-empty key. Function returns hash with all read values. Any white
1077 space (other than new-line character) is preserved.
1081 sub credential_read {
1082 my ($self, $reader) = _maybe_self(@_);
1088 } elsif (!/^([^=]+)=(.*)$/) {
1089 throw Error::Simple("unable to parse git credential data:\n$_");
1091 $credential{$1} = $2;
1096 =item credential_write( FILEHANDLE, CREDENTIAL_HASHREF )
1098 Writes credential key-value pairs from hash referenced by
1099 C<CREDENTIAL_HASHREF> to C<FILEHANDLE>. Keys and values cannot contain
1100 new-lines or NUL bytes characters, and key cannot contain equal signs nor be
1101 empty (if they do Error::Simple is thrown). Any white space is preserved. If
1102 value for a key is C<undef>, it will be skipped.
1104 If C<'url'> key exists it will be written first. (All the other key-value
1105 pairs are written in sorted order but you should not depend on that). Once
1106 all lines are written, an empty line is printed.
1110 sub credential_write {
1111 my ($self, $writer, $credential) = _maybe_self(@_);
1114 # Check if $credential is valid prior to writing anything
1115 while (($key, $value) = each %$credential) {
1116 if (!defined $key || !length $key) {
1117 throw Error::Simple("credential key empty or undefined");
1118 } elsif ($key =~ /[=\n\0]/) {
1119 throw Error::Simple("credential key contains invalid characters: $key");
1120 } elsif (defined $value && $value =~ /[\n\0]/) {
1121 throw Error::Simple("credential value for key=$key contains invalid characters: $value");
1126 # url overwrites other fields, so it must come first
1127 return -1 if $a eq 'url';
1128 return 1 if $b eq 'url';
1130 } keys %$credential) {
1131 if (defined $credential->{$key}) {
1132 print $writer $key, '=', $credential->{$key}, "\n";
1138 sub _credential_run {
1139 my ($self, $credential, $op) = _maybe_self(@_);
1140 my ($pid, $reader, $writer, $ctx) = command_bidi_pipe('credential', $op);
1142 credential_write $writer, $credential;
1145 if ($op eq "fill") {
1146 %$credential = credential_read $reader;
1149 throw Error::Simple("unexpected output from git credential $op response:\n$_\n");
1152 command_close_bidi_pipe($pid, $reader, undef, $ctx);
1155 =item credential( CREDENTIAL_HASHREF [, OPERATION ] )
1157 =item credential( CREDENTIAL_HASHREF, CODE )
1159 Executes C<git credential> for a given set of credentials and specified
1160 operation. In both forms C<CREDENTIAL_HASHREF> needs to be a reference to
1161 a hash which stores credentials. Under certain conditions the hash can
1164 In the first form, C<OPERATION> can be C<'fill'>, C<'approve'> or C<'reject'>,
1165 and function will execute corresponding C<git credential> sub-command. If
1166 it's omitted C<'fill'> is assumed. In case of C<'fill'> the values stored in
1167 C<CREDENTIAL_HASHREF> will be changed to the ones returned by the C<git
1168 credential fill> command. The usual usage would look something like:
1171 'protocol' => 'https',
1172 'host' => 'example.com',
1175 Git::credential \%cred;
1176 if (try_to_authenticate($cred{'username'}, $cred{'password'})) {
1177 Git::credential \%cred, 'approve';
1178 ... do more stuff ...
1180 Git::credential \%cred, 'reject';
1183 In the second form, C<CODE> needs to be a reference to a subroutine. The
1184 function will execute C<git credential fill> to fill the provided credential
1185 hash, then call C<CODE> with C<CREDENTIAL_HASHREF> as the sole argument. If
1186 C<CODE>'s return value is defined, the function will execute C<git credential
1187 approve> (if return value yields true) or C<git credential reject> (if return
1188 value is false). If the return value is undef, nothing at all is executed;
1189 this is useful, for example, if the credential could neither be verified nor
1190 rejected due to an unrelated network error. The return value is the same as
1191 what C<CODE> returns. With this form, the usage might look as follows:
1193 if (Git::credential {
1194 'protocol' => 'https',
1195 'host' => 'example.com',
1199 return !!try_to_authenticate($cred->{'username'},
1200 $cred->{'password'});
1202 ... do more stuff ...
1208 my ($self, $credential, $op_or_code) = (_maybe_self(@_), 'fill');
1210 if ('CODE' eq ref $op_or_code) {
1211 _credential_run $credential, 'fill';
1212 my $ret = $op_or_code->($credential);
1214 _credential_run $credential, $ret ? 'approve' : 'reject';
1218 _credential_run $credential, $op_or_code;
1222 { # %TEMP_* Lexical Context
1224 my (%TEMP_FILEMAP, %TEMP_FILES);
1226 =item temp_acquire ( NAME )
1228 Attempts to retrieve the temporary file mapped to the string C<NAME>. If an
1229 associated temp file has not been created this session or was closed, it is
1230 created, cached, and set for autoflush and binmode.
1232 Internally locks the file mapped to C<NAME>. This lock must be released with
1233 C<temp_release()> when the temp file is no longer needed. Subsequent attempts
1234 to retrieve temporary files mapped to the same C<NAME> while still locked will
1235 cause an error. This locking mechanism provides a weak guarantee and is not
1236 threadsafe. It does provide some error checking to help prevent temp file refs
1237 writing over one another.
1239 In general, the L<File::Handle> returned should not be closed by consumers as
1240 it defeats the purpose of this caching mechanism. If you need to close the temp
1241 file handle, then you should use L<File::Temp> or another temp file faculty
1242 directly. If a handle is closed and then requested again, then a warning will
1248 my $temp_fd = _temp_cache(@_);
1250 $TEMP_FILES{$temp_fd}{locked} = 1;
1254 =item temp_is_locked ( NAME )
1256 Returns true if the internal lock created by a previous C<temp_acquire()>
1257 call with C<NAME> is still in effect.
1259 When temp_acquire is called on a C<NAME>, it internally locks the temporary
1260 file mapped to C<NAME>. That lock will not be released until C<temp_release()>
1261 is called with either the original C<NAME> or the L<File::Handle> that was
1262 returned from the original call to temp_acquire.
1264 Subsequent attempts to call C<temp_acquire()> with the same C<NAME> will fail
1265 unless there has been an intervening C<temp_release()> call for that C<NAME>
1266 (or its corresponding L<File::Handle> that was returned by the original
1267 C<temp_acquire()> call).
1269 If true is returned by C<temp_is_locked()> for a C<NAME>, an attempt to
1270 C<temp_acquire()> the same C<NAME> will cause an error unless
1271 C<temp_release> is first called on that C<NAME> (or its corresponding
1272 L<File::Handle> that was returned by the original C<temp_acquire()> call).
1276 sub temp_is_locked {
1277 my ($self, $name) = _maybe_self(@_);
1278 my $temp_fd = \$TEMP_FILEMAP{$name};
1280 defined $$temp_fd && $$temp_fd->opened && $TEMP_FILES{$$temp_fd}{locked};
1283 =item temp_release ( NAME )
1285 =item temp_release ( FILEHANDLE )
1287 Releases a lock acquired through C<temp_acquire()>. Can be called either with
1288 the C<NAME> mapping used when acquiring the temp file or with the C<FILEHANDLE>
1289 referencing a locked temp file.
1291 Warns if an attempt is made to release a file that is not locked.
1293 The temp file will be truncated before being released. This can help to reduce
1294 disk I/O where the system is smart enough to detect the truncation while data
1295 is in the output buffers. Beware that after the temp file is released and
1296 truncated, any operations on that file may fail miserably until it is
1297 re-acquired. All contents are lost between each release and acquire mapped to
1303 my ($self, $temp_fd, $trunc) = _maybe_self(@_);
1305 if (exists $TEMP_FILEMAP{$temp_fd}) {
1306 $temp_fd = $TEMP_FILES{$temp_fd};
1308 unless ($TEMP_FILES{$temp_fd}{locked}) {
1309 carp "Attempt to release temp file '",
1310 $temp_fd, "' that has not been locked";
1312 temp_reset($temp_fd) if $trunc and $temp_fd->opened;
1314 $TEMP_FILES{$temp_fd}{locked} = 0;
1319 my ($self, $name) = _maybe_self(@_);
1321 my $temp_fd = \$TEMP_FILEMAP{$name};
1322 if (defined $$temp_fd and $$temp_fd->opened) {
1323 if ($TEMP_FILES{$$temp_fd}{locked}) {
1324 throw Error::Simple("Temp file with moniker '" .
1325 $name . "' already in use");
1328 if (defined $$temp_fd) {
1329 # then we're here because of a closed handle.
1330 carp "Temp file '", $name,
1331 "' was closed. Opening replacement.";
1336 if (defined $self) {
1337 $tmpdir = $self->repo_path();
1341 $n =~ s/\W/_/g; # no strange chars
1343 ($$temp_fd, $fname) = File::Temp::tempfile(
1344 "Git_${n}_XXXXXX", UNLINK => 1, DIR => $tmpdir,
1345 ) or throw Error::Simple("couldn't open new temp file");
1347 $$temp_fd->autoflush;
1349 $TEMP_FILES{$$temp_fd}{fname} = $fname;
1354 =item temp_reset ( FILEHANDLE )
1356 Truncates and resets the position of the C<FILEHANDLE>.
1361 my ($self, $temp_fd) = _maybe_self(@_);
1363 truncate $temp_fd, 0
1364 or throw Error::Simple("couldn't truncate file");
1365 sysseek($temp_fd, 0, SEEK_SET) and seek($temp_fd, 0, SEEK_SET)
1366 or throw Error::Simple("couldn't seek to beginning of file");
1367 sysseek($temp_fd, 0, SEEK_CUR) == 0 and tell($temp_fd) == 0
1368 or throw Error::Simple("expected file position to be reset");
1371 =item temp_path ( NAME )
1373 =item temp_path ( FILEHANDLE )
1375 Returns the filename associated with the given tempfile.
1380 my ($self, $temp_fd) = _maybe_self(@_);
1382 if (exists $TEMP_FILEMAP{$temp_fd}) {
1383 $temp_fd = $TEMP_FILEMAP{$temp_fd};
1385 $TEMP_FILES{$temp_fd}{fname};
1389 unlink values %TEMP_FILEMAP if %TEMP_FILEMAP;
1392 } # %TEMP_* Lexical Context
1394 =item prefix_lines ( PREFIX, STRING [, STRING... ])
1396 Prefixes lines in C<STRING> with C<PREFIX>.
1402 my $string = join("\n", @_);
1403 $string =~ s/^/$prefix/mg;
1407 =item unquote_path ( PATH )
1409 Unquote a quoted path containing c-escapes as returned by ls-files etc.
1410 when not using -z or when parsing the output of diff -u.
1429 my ($retval, $remainder);
1430 if (!/^\042(.*)\042$/) {
1433 ($_, $retval) = ($1, "");
1434 while (/^([^\\]*)\\(.*)$/) {
1438 if (/^([0-3][0-7][0-7])(.*)$/) {
1439 $retval .= chr(oct($1));
1443 if (/^([\\\042abtnvfr])(.*)$/) {
1444 $retval .= $cquote_map{$1};
1449 throw Error::Simple("invalid quoted path $_[0]");
1458 =item get_comment_line_char ( )
1460 Gets the core.commentchar configuration value.
1461 The value falls-back to '#' if core.commentchar is set to 'auto'.
1465 sub get_comment_line_char {
1466 my $comment_line_char = config("core.commentchar") || '#';
1467 $comment_line_char = '#' if ($comment_line_char eq 'auto');
1468 $comment_line_char = '#' if (length($comment_line_char) != 1);
1469 return $comment_line_char;
1472 =item comment_lines ( STRING [, STRING... ])
1474 Comments lines following core.commentchar configuration.
1479 my $comment_line_char = get_comment_line_char;
1480 return prefix_lines("$comment_line_char ", @_);
1485 =head1 ERROR HANDLING
1487 All functions are supposed to throw Perl exceptions in case of errors.
1488 See the L<Error> module on how to catch those. Most exceptions are mere
1489 L<Error::Simple> instances.
1491 However, the C<command()>, C<command_oneline()> and C<command_noisy()>
1492 functions suite can throw C<Git::Error::Command> exceptions as well: those are
1493 thrown when the external command returns an error code and contain the error
1494 code as well as access to the captured command's output. The exception class
1495 provides the usual C<stringify> and C<value> (command's exit code) methods and
1496 in addition also a C<cmd_output> method that returns either an array or a
1497 string with the captured command output (depending on the original function
1498 call context; C<command_noisy()> returns C<undef>) and $<cmdline> which
1499 returns the command and its arguments (but without proper quoting).
1501 Note that the C<command_*_pipe()> functions cannot throw this exception since
1502 it has no idea whether the command failed or not. You will only find out
1503 at the time you C<close> the pipe; if you want to have that automated,
1504 use C<command_close_pipe()>, which can throw the exception.
1509 package Git::Error::Command;
1511 @Git::Error::Command::ISA = qw(Error);
1515 my $cmdline = '' . shift;
1516 my $value = 0 + shift;
1517 my $outputref = shift;
1520 local $Error::Depth = $Error::Depth + 1;
1522 push(@args, '-cmdline', $cmdline);
1523 push(@args, '-value', $value);
1524 push(@args, '-outputref', $outputref);
1526 $self->SUPER::new(-text => 'command returned error', @args);
1531 my $text = $self->SUPER::stringify;
1532 $self->cmdline() . ': ' . $text . ': ' . $self->value() . "\n";
1537 $self->{'-cmdline'};
1542 my $ref = $self->{'-outputref'};
1543 defined $ref or undef;
1544 if (ref $ref eq 'ARRAY') {
1554 =item git_cmd_try { CODE } ERRMSG
1556 This magical statement will automatically catch any C<Git::Error::Command>
1557 exceptions thrown by C<CODE> and make your program die with C<ERRMSG>
1558 on its lips; the message will have %s substituted for the command line
1559 and %d for the exit status. This statement is useful mostly for producing
1560 more user-friendly error messages.
1562 In case of no exception caught the statement returns C<CODE>'s return value.
1564 Note that this is the only auto-exported function.
1568 sub git_cmd_try(&$) {
1569 my ($code, $errmsg) = @_;
1572 my $array = wantarray;
1577 $result[0] = &$code;
1579 } catch Git::Error::Command with {
1582 $err =~ s/\%s/$E->cmdline()/ge;
1583 $err =~ s/\%d/$E->value()/ge;
1584 # We can't croak here since Error.pm would mangle
1585 # that to Error::Simple.
1587 $err and croak $err;
1588 return $array ? @result : $result[0];
1596 Copyright 2006 by Petr Baudis E<lt>pasky@suse.czE<gt>.
1598 This module is free software; it may be used, copied, modified
1599 and distributed under the terms of the GNU General Public Licence,
1600 either version 2, or (at your option) any later version.
1605 # Take raw method argument list and return ($obj, @args) in case
1606 # the method was called upon an instance and (undef, @args) if
1607 # it was called directly.
1609 UNIVERSAL::isa($_[0], 'Git') ? @_ : (undef, @_);
1612 # Check if the command id is something reasonable.
1613 sub _check_valid_cmd {
1615 $cmd =~ /^[a-z0-9A-Z_-]+$/ or throw Error::Simple("bad command: $cmd");
1618 # Common backend for the pipe creators.
1619 sub _command_common_pipe {
1620 my $direction = shift;
1621 my ($self, @p) = _maybe_self(@_);
1622 my (%opts, $cmd, @args);
1624 ($cmd, @args) = @{shift @p};
1625 %opts = ref $p[0] ? %{$p[0]} : @p;
1629 _check_valid_cmd($cmd);
1632 if ($^O eq 'MSWin32') {
1634 #defined $opts{STDERR} and
1635 # warn 'ignoring STDERR option - running w/ ActiveState';
1636 $direction eq '-|' or
1637 die 'input pipe for ActiveState not implemented';
1638 # the strange construction with *ACPIPE is just to
1639 # explain the tie below that we want to bind to
1640 # a handle class, not scalar. It is not known if
1641 # it is something specific to ActiveState Perl or
1642 # just a Perl quirk.
1643 tie (*ACPIPE, 'Git::activestate_pipe', $cmd, @args);
1647 my $pid = open($fh, $direction);
1648 if (not defined $pid) {
1649 throw Error::Simple("open failed: $!");
1650 } elsif ($pid == 0) {
1651 if ($opts{STDERR}) {
1652 open (STDERR, '>&', $opts{STDERR})
1653 or die "dup failed: $!";
1654 } elsif (defined $opts{STDERR}) {
1655 open (STDERR, '>', '/dev/null')
1656 or die "opening /dev/null failed: $!";
1658 _cmd_exec($self, $cmd, @args);
1661 return wantarray ? ($fh, join(' ', $cmd, @args)) : $fh;
1664 # When already in the subprocess, set up the appropriate state
1665 # for the given repository and execute the git command.
1667 my ($self, @args) = @_;
1668 _setup_git_cmd_env($self);
1669 _execv_git_cmd(@args);
1670 die qq[exec "@args" failed: $!];
1673 # set up the appropriate state for git command
1674 sub _setup_git_cmd_env {
1677 $self->repo_path() and $ENV{'GIT_DIR'} = $self->repo_path();
1678 $self->repo_path() and $self->wc_path()
1679 and $ENV{'GIT_WORK_TREE'} = $self->wc_path();
1680 $self->wc_path() and chdir($self->wc_path());
1681 $self->wc_subdir() and chdir($self->wc_subdir());
1685 # Execute the given Git command ($_[0]) with arguments ($_[1..])
1686 # by searching for it at proper places.
1687 sub _execv_git_cmd { exec('git', @_); }
1689 # Close pipe to a subprocess.
1692 foreach my $fh (@_) {
1696 # It's just close, no point in fatalities
1697 carp "error closing pipe: $!";
1699 # The caller should pepper this.
1700 throw Git::Error::Command($ctx, $? >> 8);
1702 # else we might e.g. closed a live stream; the command
1703 # dying of SIGPIPE would drive us here.
1710 $self->_close_hash_and_insert_object();
1711 $self->_close_cat_blob();
1715 # Pipe implementation for ActiveState Perl.
1717 package Git::activestate_pipe;
1720 my ($class, @params) = @_;
1721 # FIXME: This is probably horrible idea and the thing will explode
1722 # at the moment you give it arguments that require some quoting,
1723 # but I have no ActiveState clue... --pasky
1724 # Let's just hope ActiveState Perl does at least the quoting
1726 my @data = qx{git @params};
1727 bless { i => 0, data => \@data }, $class;
1732 if ($self->{i} >= scalar @{$self->{data}}) {
1737 $self->{i} = $#{$self->{'data'}} + 1;
1738 return splice(@{$self->{'data'}}, $i);
1740 $self->{i} = $i + 1;
1741 return $self->{'data'}->[ $i ];
1746 delete $self->{data};
1752 return ($self->{i} >= scalar @{$self->{data}});
1756 1; # Famous last words