3 Git - Perl interface to the Git version control system
16 our ($VERSION, @ISA, @EXPORT, @EXPORT_OK);
18 # Totally unstable API.
26 my $version = Git::command_oneline('version');
28 git_cmd_try { Git::command_noisy('update-server-info') }
29 '%s failed w/ code %d';
31 my $repo = Git->repository (Directory => '/srv/git/cogito.git');
34 my @revs = $repo->command('rev-list', '--since=last monday', '--all');
36 my ($fh, $c) = $repo->command_output_pipe('rev-list', '--since=last monday', '--all');
37 my $lastrev = <$fh>; chomp $lastrev;
38 $repo->command_close_pipe($fh, $c);
40 my $lastrev = $repo->command_oneline( [ 'rev-list', '--all' ],
43 my $sha1 = $repo->hash_and_insert_object('file.txt');
44 my $tempfile = tempfile();
45 my $size = $repo->cat_blob($sha1, $tempfile);
54 @EXPORT = qw(git_cmd_try);
56 # Methods which can be called as standalone functions as well:
57 @EXPORT_OK = qw(command command_oneline command_noisy
58 command_output_pipe command_input_pipe command_close_pipe
59 command_bidi_pipe command_close_bidi_pipe
60 version exec_path html_path hash_object git_cmd_try
62 get_tz_offset get_record
63 credential credential_read credential_write
64 temp_acquire temp_is_locked temp_release temp_reset temp_path
70 This module provides Perl scripts easy way to interface the Git version control
71 system. The modules have an easy and well-tested way to call arbitrary Git
72 commands; in the future, the interface will also provide specialized methods
73 for doing easily operations which are not totally trivial to do over
74 the generic command interface.
76 While some commands can be executed outside of any context (e.g. 'version'
77 or 'init'), most operations require a repository context, which in practice
78 means getting an instance of the Git object using the repository() constructor.
79 (In the future, we will also get a new_repository() constructor.) All commands
80 called as methods of the object are then executed in the context of the
83 Part of the "repository state" is also information about path to the attached
84 working copy (unless you work with a bare repository). You can also navigate
85 inside of the working copy using the C<wc_chdir()> method. (Note that
86 the repository object is self-contained and will not change working directory
89 TODO: In the future, we might also do
91 my $remoterepo = $repo->remote_repository (Name => 'cogito', Branch => 'master');
92 $remoterepo ||= Git->remote_repository ('http://git.or.cz/cogito.git/');
93 my @refs = $remoterepo->refs();
95 Currently, the module merely wraps calls to external Git tools. In the future,
96 it will provide a much faster way to interact with Git by linking directly
97 to libgit. This should be completely opaque to the user, though (performance
98 increase notwithstanding).
103 use Carp qw(carp croak); # but croak is bad - throw instead
105 use Cwd qw(abs_path cwd);
106 use IPC::Open2 qw(open2);
107 use Fcntl qw(SEEK_SET SEEK_CUR);
108 use Time::Local qw(timegm);
116 =item repository ( OPTIONS )
118 =item repository ( DIRECTORY )
122 Construct a new repository object.
123 C<OPTIONS> are passed in a hash like fashion, using key and value pairs.
124 Possible options are:
126 B<Repository> - Path to the Git repository.
128 B<WorkingCopy> - Path to the associated working copy; not strictly required
129 as many commands will happily crunch on a bare repository.
131 B<WorkingSubdir> - Subdirectory in the working copy to work inside.
132 Just left undefined if you do not want to limit the scope of operations.
134 B<Directory> - Path to the Git working directory in its usual setup.
135 The C<.git> directory is searched in the directory and all the parent
136 directories; if found, C<WorkingCopy> is set to the directory containing
137 it and C<Repository> to the C<.git> directory itself. If no C<.git>
138 directory was found, the C<Directory> is assumed to be a bare repository,
139 C<Repository> is set to point at it and C<WorkingCopy> is left undefined.
140 If the C<$GIT_DIR> environment variable is set, things behave as expected
143 You should not use both C<Directory> and either of C<Repository> and
144 C<WorkingCopy> - the results of that are undefined.
146 Alternatively, a directory path may be passed as a single scalar argument
147 to the constructor; it is equivalent to setting only the C<Directory> option
150 Calling the constructor with no options whatsoever is equivalent to
151 calling it with C<< Directory => '.' >>. In general, if you are building
152 a standard porcelain command, simply doing C<< Git->repository() >> should
153 do the right thing and setup the object to reflect exactly where the user
164 if (defined $args[0]) {
165 if ($#args % 2 != 1) {
167 $#args == 0 or throw Error::Simple("bad usage");
168 %opts = ( Directory => $args[0] );
174 if (not defined $opts{Repository} and not defined $opts{WorkingCopy}
175 and not defined $opts{Directory}) {
176 $opts{Directory} = '.';
179 if (defined $opts{Directory}) {
180 -d $opts{Directory} or throw Error::Simple("Directory not found: $opts{Directory} $!");
182 my $search = Git->repository(WorkingCopy => $opts{Directory});
185 $dir = $search->command_oneline(['rev-parse', '--git-dir'],
187 } catch Git::Error::Command with {
193 File::Spec->file_name_is_absolute($dir) or $dir = $opts{Directory} . '/' . $dir;
194 $opts{Repository} = abs_path($dir);
196 # If --git-dir went ok, this shouldn't die either.
197 my $prefix = $search->command_oneline('rev-parse', '--show-prefix');
198 $dir = abs_path($opts{Directory}) . '/';
200 if (substr($dir, -length($prefix)) ne $prefix) {
201 throw Error::Simple("rev-parse confused me - $dir does not have trailing $prefix");
203 substr($dir, -length($prefix)) = '';
205 $opts{WorkingCopy} = $dir;
206 $opts{WorkingSubdir} = $prefix;
209 # A bare repository? Let's see...
210 $dir = $opts{Directory};
212 unless (-d "$dir/refs" and -d "$dir/objects" and -e "$dir/HEAD") {
213 # Mimic git-rev-parse --git-dir error message:
214 throw Error::Simple("fatal: Not a git repository: $dir");
216 my $search = Git->repository(Repository => $dir);
218 $search->command('symbolic-ref', 'HEAD');
219 } catch Git::Error::Command with {
220 # Mimic git-rev-parse --git-dir error message:
221 throw Error::Simple("fatal: Not a git repository: $dir");
224 $opts{Repository} = abs_path($dir);
227 delete $opts{Directory};
230 $self = { opts => \%opts };
240 =item command ( COMMAND [, ARGUMENTS... ] )
242 =item command ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
244 Execute the given Git C<COMMAND> (specify it without the 'git-'
245 prefix), optionally with the specified extra C<ARGUMENTS>.
247 The second more elaborate form can be used if you want to further adjust
248 the command execution. Currently, only one option is supported:
250 B<STDERR> - How to deal with the command's error output. By default (C<undef>)
251 it is delivered to the caller's C<STDERR>. A false value (0 or '') will cause
252 it to be thrown away. If you want to process it, you can get it in a filehandle
253 you specify, but you must be extremely careful; if the error output is not
254 very short and you want to read it in the same process as where you called
255 C<command()>, you are set up for a nice deadlock!
257 The method can be called without any instance or on a specified Git repository
258 (in that case the command will be run in the repository context).
260 In scalar context, it returns all the command output in a single string
263 In array context, it returns an array containing lines printed to the
264 command's stdout (without trailing newlines).
266 In both cases, the command's stdin and stderr are the same as the caller's.
271 my ($fh, $ctx) = command_output_pipe(@_);
273 if (not defined wantarray) {
274 # Nothing to pepper the possible exception with.
275 _cmd_close($ctx, $fh);
277 } elsif (not wantarray) {
281 _cmd_close($ctx, $fh);
282 } catch Git::Error::Command with {
283 # Pepper with the output:
285 $E->{'-outputref'} = \$text;
292 defined and chomp for @lines;
294 _cmd_close($ctx, $fh);
295 } catch Git::Error::Command with {
297 $E->{'-outputref'} = \@lines;
305 =item command_oneline ( COMMAND [, ARGUMENTS... ] )
307 =item command_oneline ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
309 Execute the given C<COMMAND> in the same way as command()
310 does but always return a scalar string containing the first line
311 of the command's standard output.
315 sub command_oneline {
316 my ($fh, $ctx) = command_output_pipe(@_);
319 defined $line and chomp $line;
321 _cmd_close($ctx, $fh);
322 } catch Git::Error::Command with {
323 # Pepper with the output:
325 $E->{'-outputref'} = \$line;
332 =item command_output_pipe ( COMMAND [, ARGUMENTS... ] )
334 =item command_output_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
336 Execute the given C<COMMAND> in the same way as command()
337 does but return a pipe filehandle from which the command output can be
340 The function can return C<($pipe, $ctx)> in array context.
341 See C<command_close_pipe()> for details.
345 sub command_output_pipe {
346 _command_common_pipe('-|', @_);
350 =item command_input_pipe ( COMMAND [, ARGUMENTS... ] )
352 =item command_input_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
354 Execute the given C<COMMAND> in the same way as command_output_pipe()
355 does but return an input pipe filehandle instead; the command output
358 The function can return C<($pipe, $ctx)> in array context.
359 See C<command_close_pipe()> for details.
363 sub command_input_pipe {
364 _command_common_pipe('|-', @_);
368 =item command_close_pipe ( PIPE [, CTX ] )
370 Close the C<PIPE> as returned from C<command_*_pipe()>, checking
371 whether the command finished successfully. The optional C<CTX> argument
372 is required if you want to see the command name in the error message,
373 and it is the second value returned by C<command_*_pipe()> when
374 called in array context. The call idiom is:
376 my ($fh, $ctx) = $r->command_output_pipe('status');
377 while (<$fh>) { ... }
378 $r->command_close_pipe($fh, $ctx);
380 Note that you should not rely on whatever actually is in C<CTX>;
381 currently it is simply the command name but in future the context might
382 have more complicated structure.
386 sub command_close_pipe {
387 my ($self, $fh, $ctx) = _maybe_self(@_);
388 $ctx ||= '<unknown>';
389 _cmd_close($ctx, $fh);
392 =item command_bidi_pipe ( COMMAND [, ARGUMENTS... ] )
394 Execute the given C<COMMAND> in the same way as command_output_pipe()
395 does but return both an input pipe filehandle and an output pipe filehandle.
397 The function will return C<($pid, $pipe_in, $pipe_out, $ctx)>.
398 See C<command_close_bidi_pipe()> for details.
402 sub command_bidi_pipe {
403 my ($pid, $in, $out);
404 my ($self) = _maybe_self(@_);
406 my $cwd_save = undef;
410 _setup_git_cmd_env($self);
412 $pid = open2($in, $out, 'git', @_);
413 chdir($cwd_save) if $cwd_save;
414 return ($pid, $in, $out, join(' ', @_));
417 =item command_close_bidi_pipe ( PID, PIPE_IN, PIPE_OUT [, CTX] )
419 Close the C<PIPE_IN> and C<PIPE_OUT> as returned from C<command_bidi_pipe()>,
420 checking whether the command finished successfully. The optional C<CTX>
421 argument is required if you want to see the command name in the error message,
422 and it is the fourth value returned by C<command_bidi_pipe()>. The call idiom
425 my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
426 print $out "000000000\n";
427 while (<$in>) { ... }
428 $r->command_close_bidi_pipe($pid, $in, $out, $ctx);
430 Note that you should not rely on whatever actually is in C<CTX>;
431 currently it is simply the command name but in future the context might
432 have more complicated structure.
434 C<PIPE_IN> and C<PIPE_OUT> may be C<undef> if they have been closed prior to
435 calling this function. This may be useful in a query-response type of
436 commands where caller first writes a query and later reads response, eg:
438 my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
439 print $out "000000000\n";
441 while (<$in>) { ... }
442 $r->command_close_bidi_pipe($pid, $in, undef, $ctx);
444 This idiom may prevent potential dead locks caused by data sent to the output
445 pipe not being flushed and thus not reaching the executed command.
449 sub command_close_bidi_pipe {
451 my ($self, $pid, $in, $out, $ctx) = _maybe_self(@_);
452 _cmd_close($ctx, (grep { defined } ($in, $out)));
455 throw Git::Error::Command($ctx, $? >>8);
460 =item command_noisy ( COMMAND [, ARGUMENTS... ] )
462 Execute the given C<COMMAND> in the same way as command() does but do not
463 capture the command output - the standard output is not redirected and goes
464 to the standard output of the caller application.
466 While the method is called command_noisy(), you might want to as well use
467 it for the most silent Git commands which you know will never pollute your
468 stdout but you want to avoid the overhead of the pipe setup when calling them.
470 The function returns only after the command has finished running.
475 my ($self, $cmd, @args) = _maybe_self(@_);
476 _check_valid_cmd($cmd);
479 if (not defined $pid) {
480 throw Error::Simple("fork failed: $!");
481 } elsif ($pid == 0) {
482 _cmd_exec($self, $cmd, @args);
484 if (waitpid($pid, 0) > 0 and $?>>8 != 0) {
485 throw Git::Error::Command(join(' ', $cmd, @args), $? >> 8);
492 Return the Git version in use.
497 my $verstr = command_oneline('--version');
498 $verstr =~ s/^git version //;
505 Return path to the Git sub-command executables (the same as
506 C<git --exec-path>). Useful mostly only internally.
510 sub exec_path { command_oneline('--exec-path') }
515 Return path to the Git html documentation (the same as
516 C<git --html-path>). Useful mostly only internally.
520 sub html_path { command_oneline('--html-path') }
523 =item get_tz_offset ( TIME )
525 Return the time zone offset from GMT in the form +/-HHMM where HH is
526 the number of hours from GMT and MM is the number of minutes. This is
527 the equivalent of what strftime("%z", ...) would provide on a GNU
530 If TIME is not supplied, the current local time is used.
535 # some systems don't handle or mishandle %z, so be creative.
536 my $t = shift || time;
537 my @t = localtime($t);
540 my $sign = qw( + + - )[ $gm <=> $t ];
541 return sprintf("%s%02d%02d", $sign, (gmtime(abs($t - $gm)))[2,1]);
544 =item get_record ( FILEHANDLE, INPUT_RECORD_SEPARATOR )
546 Read one record from FILEHANDLE delimited by INPUT_RECORD_SEPARATOR,
547 removing any trailing INPUT_RECORD_SEPARATOR.
555 chomp $rec if defined $rs;
559 =item prompt ( PROMPT , ISPASSWORD )
561 Query user C<PROMPT> and return answer from user.
563 Honours GIT_ASKPASS and SSH_ASKPASS environment variables for querying
564 the user. If no *_ASKPASS variable is set or an error occoured,
565 the terminal is tried as a fallback.
566 If C<ISPASSWORD> is set and true, the terminal disables echo.
571 my ($prompt, $isPassword) = @_;
573 if (exists $ENV{'GIT_ASKPASS'}) {
574 $ret = _prompt($ENV{'GIT_ASKPASS'}, $prompt);
576 if (!defined $ret && exists $ENV{'SSH_ASKPASS'}) {
577 $ret = _prompt($ENV{'SSH_ASKPASS'}, $prompt);
580 print STDERR $prompt;
582 if (defined $isPassword && $isPassword) {
583 require Term::ReadKey;
584 Term::ReadKey::ReadMode('noecho');
586 while (defined(my $key = Term::ReadKey::ReadKey(0))) {
587 last if $key =~ /[\012\015]/; # \n\r
590 Term::ReadKey::ReadMode('restore');
594 chomp($ret = <STDIN>);
601 my ($askpass, $prompt) = @_;
602 return unless length $askpass;
605 open my $fh, "-|", $askpass, $prompt or return;
607 $ret =~ s/[\015\012]//g; # strip \r\n, chomp does not work on all systems (i.e. windows) as expected
614 Return path to the git repository. Must be called on a repository instance.
618 sub repo_path { $_[0]->{opts}->{Repository} }
623 Return path to the working copy. Must be called on a repository instance.
627 sub wc_path { $_[0]->{opts}->{WorkingCopy} }
632 Return path to the subdirectory inside of a working copy. Must be called
633 on a repository instance.
637 sub wc_subdir { $_[0]->{opts}->{WorkingSubdir} ||= '' }
640 =item wc_chdir ( SUBDIR )
642 Change the working copy subdirectory to work within. The C<SUBDIR> is
643 relative to the working copy root directory (not the current subdirectory).
644 Must be called on a repository instance attached to a working copy
645 and the directory must exist.
650 my ($self, $subdir) = @_;
652 or throw Error::Simple("bare repository");
654 -d $self->wc_path().'/'.$subdir
655 or throw Error::Simple("subdir not found: $subdir $!");
656 # Of course we will not "hold" the subdirectory so anyone
657 # can delete it now and we will never know. But at least we tried.
659 $self->{opts}->{WorkingSubdir} = $subdir;
663 =item config ( VARIABLE )
665 Retrieve the configuration C<VARIABLE> in the same manner as C<config>
666 does. In scalar context requires the variable to be set only one time
667 (exception is thrown otherwise), in array context returns allows the
668 variable to be set multiple times and returns all the values.
673 return _config_common({}, @_);
677 =item config_bool ( VARIABLE )
679 Retrieve the bool configuration C<VARIABLE>. The return value
680 is usable as a boolean in perl (and C<undef> if it's not defined,
686 my $val = scalar _config_common({'kind' => '--bool'}, @_);
688 # Do not rewrite this as return (defined $val && $val eq 'true')
689 # as some callers do care what kind of falsehood they receive.
693 return $val eq 'true';
698 =item config_path ( VARIABLE )
700 Retrieve the path configuration C<VARIABLE>. The return value
701 is an expanded path or C<undef> if it's not defined.
706 return _config_common({'kind' => '--path'}, @_);
710 =item config_int ( VARIABLE )
712 Retrieve the integer configuration C<VARIABLE>. The return value
713 is simple decimal number. An optional value suffix of 'k', 'm',
714 or 'g' in the config file will cause the value to be multiplied
715 by 1024, 1048576 (1024^2), or 1073741824 (1024^3) prior to output.
716 It would return C<undef> if configuration variable is not defined.
721 return scalar _config_common({'kind' => '--int'}, @_);
724 # Common subroutine to implement bulk of what the config* family of methods
725 # do. This currently wraps command('config') so it is not so fast.
727 my ($opts) = shift @_;
728 my ($self, $var) = _maybe_self(@_);
731 my @cmd = ('config', $opts->{'kind'} ? $opts->{'kind'} : ());
732 unshift @cmd, $self if $self;
734 return command(@cmd, '--get-all', $var);
736 return command_oneline(@cmd, '--get', $var);
738 } catch Git::Error::Command with {
740 if ($E->value() == 1) {
749 =item get_colorbool ( NAME )
751 Finds if color should be used for NAMEd operation from the configuration,
752 and returns boolean (true for "use color", false for "do not use color").
757 my ($self, $var) = @_;
758 my $stdout_to_tty = (-t STDOUT) ? "true" : "false";
759 my $use_color = $self->command_oneline('config', '--get-colorbool',
760 $var, $stdout_to_tty);
761 return ($use_color eq 'true');
764 =item get_color ( SLOT, COLOR )
766 Finds color for SLOT from the configuration, while defaulting to COLOR,
767 and returns the ANSI color escape sequence:
769 print $repo->get_color("color.interactive.prompt", "underline blue white");
771 print $repo->get_color("", "normal");
776 my ($self, $slot, $default) = @_;
777 my $color = $self->command_oneline('config', '--get-color', $slot, $default);
778 if (!defined $color) {
784 =item remote_refs ( REPOSITORY [, GROUPS [, REFGLOBS ] ] )
786 This function returns a hashref of refs stored in a given remote repository.
787 The hash is in the format C<refname =\> hash>. For tags, the C<refname> entry
788 contains the tag object while a C<refname^{}> entry gives the tagged objects.
790 C<REPOSITORY> has the same meaning as the appropriate C<git-ls-remote>
791 argument; either a URL or a remote name (if called on a repository instance).
792 C<GROUPS> is an optional arrayref that can contain 'tags' to return all the
793 tags and/or 'heads' to return all the heads. C<REFGLOB> is an optional array
794 of strings containing a shell-like glob to further limit the refs returned in
795 the hash; the meaning is again the same as the appropriate C<git-ls-remote>
798 This function may or may not be called on a repository instance. In the former
799 case, remote names as defined in the repository are recognized as repository
805 my ($self, $repo, $groups, $refglobs) = _maybe_self(@_);
807 if (ref $groups eq 'ARRAY') {
810 push (@args, '--heads');
811 } elsif ($_ eq 'tags') {
812 push (@args, '--tags');
814 # Ignore unknown groups for future
820 if (ref $refglobs eq 'ARRAY') {
821 push (@args, @$refglobs);
824 my @self = $self ? ($self) : (); # Ultra trickery
825 my ($fh, $ctx) = Git::command_output_pipe(@self, 'ls-remote', @args);
829 my ($hash, $ref) = split(/\t/, $_, 2);
832 Git::command_close_pipe(@self, $fh, $ctx);
837 =item ident ( TYPE | IDENTSTR )
839 =item ident_person ( TYPE | IDENTSTR | IDENTARRAY )
841 This suite of functions retrieves and parses ident information, as stored
842 in the commit and tag objects or produced by C<var GIT_type_IDENT> (thus
843 C<TYPE> can be either I<author> or I<committer>; case is insignificant).
845 The C<ident> method retrieves the ident information from C<git var>
846 and either returns it as a scalar string or as an array with the fields parsed.
847 Alternatively, it can take a prepared ident string (e.g. from the commit
848 object) and just parse it.
850 C<ident_person> returns the person part of the ident - name and email;
851 it can take the same arguments as C<ident> or the array returned by C<ident>.
853 The synopsis is like:
855 my ($name, $email, $time_tz) = ident('author');
856 "$name <$email>" eq ident_person('author');
857 "$name <$email>" eq ident_person($name);
858 $time_tz =~ /^\d+ [+-]\d{4}$/;
863 my ($self, $type) = _maybe_self(@_);
865 if (lc $type eq lc 'committer' or lc $type eq lc 'author') {
866 my @cmd = ('var', 'GIT_'.uc($type).'_IDENT');
867 unshift @cmd, $self if $self;
868 $identstr = command_oneline(@cmd);
873 return $identstr =~ /^(.*) <(.*)> (\d+ [+-]\d{4})$/;
880 my ($self, @ident) = _maybe_self(@_);
881 $#ident == 0 and @ident = $self ? $self->ident($ident[0]) : ident($ident[0]);
882 return "$ident[0] <$ident[1]>";
885 =item parse_mailboxes
887 Return an array of mailboxes extracted from a string.
891 # Very close to Mail::Address's parser, but we still have minor
892 # differences in some cases (see t9000 for examples).
893 sub parse_mailboxes {
894 my $re_comment = qr/\((?:[^)]*)\)/;
895 my $re_quote = qr/"(?:[^\"\\]|\\.)*"/;
896 my $re_word = qr/(?:[^]["\s()<>:;@\\,.]|\\.)+/;
898 # divide the string in tokens of the above form
899 my $re_token = qr/(?:$re_quote|$re_word|$re_comment|\S)/;
900 my @tokens = map { $_ =~ /\s*($re_token)\s*/g } @_;
901 my $end_of_addr_seen = 0;
903 # add a delimiter to simplify treatment for the last mailbox
906 my (@addr_list, @phrase, @address, @comment, @buffer) = ();
907 foreach my $token (@tokens) {
908 if ($token =~ /^[,;]$/) {
909 # if buffer still contains undeterminated strings
910 # append it at the end of @address or @phrase
911 if ($end_of_addr_seen) {
912 push @phrase, @buffer;
914 push @address, @buffer;
917 my $str_phrase = join ' ', @phrase;
918 my $str_address = join '', @address;
919 my $str_comment = join ' ', @comment;
921 # quote are necessary if phrase contains
923 if ($str_phrase =~ /[][()<>:;@\\,.\000-\037\177]/) {
924 $str_phrase =~ s/(^|[^\\])"/$1/g;
925 $str_phrase = qq["$str_phrase"];
928 # add "<>" around the address if necessary
929 if ($str_address ne "" && $str_phrase ne "") {
930 $str_address = qq[<$str_address>];
933 my $str_mailbox = "$str_phrase $str_address $str_comment";
934 $str_mailbox =~ s/^\s*|\s*$//g;
935 push @addr_list, $str_mailbox if ($str_mailbox);
937 @phrase = @address = @comment = @buffer = ();
938 $end_of_addr_seen = 0;
939 } elsif ($token =~ /^\(/) {
940 push @comment, $token;
941 } elsif ($token eq "<") {
942 push @phrase, (splice @address), (splice @buffer);
943 } elsif ($token eq ">") {
944 $end_of_addr_seen = 1;
945 push @address, (splice @buffer);
946 } elsif ($token eq "@" && !$end_of_addr_seen) {
947 push @address, (splice @buffer), "@";
949 push @buffer, $token;
956 =item hash_object ( TYPE, FILENAME )
958 Compute the SHA1 object id of the given C<FILENAME> considering it is
959 of the C<TYPE> object type (C<blob>, C<commit>, C<tree>).
961 The method can be called without any instance or on a specified Git repository,
962 it makes zero difference.
964 The function returns the SHA1 hash.
968 # TODO: Support for passing FILEHANDLE instead of FILENAME
970 my ($self, $type, $file) = _maybe_self(@_);
971 command_oneline('hash-object', '-t', $type, $file);
975 =item hash_and_insert_object ( FILENAME )
977 Compute the SHA1 object id of the given C<FILENAME> and add the object to the
980 The function returns the SHA1 hash.
984 # TODO: Support for passing FILEHANDLE instead of FILENAME
985 sub hash_and_insert_object {
986 my ($self, $filename) = @_;
988 carp "Bad filename \"$filename\"" if $filename =~ /[\r\n]/;
990 $self->_open_hash_and_insert_object_if_needed();
991 my ($in, $out) = ($self->{hash_object_in}, $self->{hash_object_out});
993 unless (print $out $filename, "\n") {
994 $self->_close_hash_and_insert_object();
995 throw Error::Simple("out pipe went bad");
998 chomp(my $hash = <$in>);
999 unless (defined($hash)) {
1000 $self->_close_hash_and_insert_object();
1001 throw Error::Simple("in pipe went bad");
1007 sub _open_hash_and_insert_object_if_needed {
1010 return if defined($self->{hash_object_pid});
1012 ($self->{hash_object_pid}, $self->{hash_object_in},
1013 $self->{hash_object_out}, $self->{hash_object_ctx}) =
1014 $self->command_bidi_pipe(qw(hash-object -w --stdin-paths --no-filters));
1017 sub _close_hash_and_insert_object {
1020 return unless defined($self->{hash_object_pid});
1022 my @vars = map { 'hash_object_' . $_ } qw(pid in out ctx);
1024 command_close_bidi_pipe(@$self{@vars});
1025 delete @$self{@vars};
1028 =item cat_blob ( SHA1, FILEHANDLE )
1030 Prints the contents of the blob identified by C<SHA1> to C<FILEHANDLE> and
1031 returns the number of bytes printed.
1036 my ($self, $sha1, $fh) = @_;
1038 $self->_open_cat_blob_if_needed();
1039 my ($in, $out) = ($self->{cat_blob_in}, $self->{cat_blob_out});
1041 unless (print $out $sha1, "\n") {
1042 $self->_close_cat_blob();
1043 throw Error::Simple("out pipe went bad");
1046 my $description = <$in>;
1047 if ($description =~ / missing$/) {
1048 carp "$sha1 doesn't exist in the repository";
1052 if ($description !~ /^[0-9a-fA-F]{40} \S+ (\d+)$/) {
1053 carp "Unexpected result returned from git cat-file";
1060 my $bytesLeft = $size;
1063 last unless $bytesLeft;
1065 my $bytesToRead = $bytesLeft < 1024 ? $bytesLeft : 1024;
1066 my $read = read($in, $blob, $bytesToRead);
1067 unless (defined($read)) {
1068 $self->_close_cat_blob();
1069 throw Error::Simple("in pipe went bad");
1071 unless (print $fh $blob) {
1072 $self->_close_cat_blob();
1073 throw Error::Simple("couldn't write to passed in filehandle");
1075 $bytesLeft -= $read;
1078 # Skip past the trailing newline.
1080 my $read = read($in, $newline, 1);
1081 unless (defined($read)) {
1082 $self->_close_cat_blob();
1083 throw Error::Simple("in pipe went bad");
1085 unless ($read == 1 && $newline eq "\n") {
1086 $self->_close_cat_blob();
1087 throw Error::Simple("didn't find newline after blob");
1093 sub _open_cat_blob_if_needed {
1096 return if defined($self->{cat_blob_pid});
1098 ($self->{cat_blob_pid}, $self->{cat_blob_in},
1099 $self->{cat_blob_out}, $self->{cat_blob_ctx}) =
1100 $self->command_bidi_pipe(qw(cat-file --batch));
1103 sub _close_cat_blob {
1106 return unless defined($self->{cat_blob_pid});
1108 my @vars = map { 'cat_blob_' . $_ } qw(pid in out ctx);
1110 command_close_bidi_pipe(@$self{@vars});
1111 delete @$self{@vars};
1115 =item credential_read( FILEHANDLE )
1117 Reads credential key-value pairs from C<FILEHANDLE>. Reading stops at EOF or
1118 when an empty line is encountered. Each line must be of the form C<key=value>
1119 with a non-empty key. Function returns hash with all read values. Any white
1120 space (other than new-line character) is preserved.
1124 sub credential_read {
1125 my ($self, $reader) = _maybe_self(@_);
1131 } elsif (!/^([^=]+)=(.*)$/) {
1132 throw Error::Simple("unable to parse git credential data:\n$_");
1134 $credential{$1} = $2;
1139 =item credential_write( FILEHANDLE, CREDENTIAL_HASHREF )
1141 Writes credential key-value pairs from hash referenced by
1142 C<CREDENTIAL_HASHREF> to C<FILEHANDLE>. Keys and values cannot contain
1143 new-lines or NUL bytes characters, and key cannot contain equal signs nor be
1144 empty (if they do Error::Simple is thrown). Any white space is preserved. If
1145 value for a key is C<undef>, it will be skipped.
1147 If C<'url'> key exists it will be written first. (All the other key-value
1148 pairs are written in sorted order but you should not depend on that). Once
1149 all lines are written, an empty line is printed.
1153 sub credential_write {
1154 my ($self, $writer, $credential) = _maybe_self(@_);
1157 # Check if $credential is valid prior to writing anything
1158 while (($key, $value) = each %$credential) {
1159 if (!defined $key || !length $key) {
1160 throw Error::Simple("credential key empty or undefined");
1161 } elsif ($key =~ /[=\n\0]/) {
1162 throw Error::Simple("credential key contains invalid characters: $key");
1163 } elsif (defined $value && $value =~ /[\n\0]/) {
1164 throw Error::Simple("credential value for key=$key contains invalid characters: $value");
1169 # url overwrites other fields, so it must come first
1170 return -1 if $a eq 'url';
1171 return 1 if $b eq 'url';
1173 } keys %$credential) {
1174 if (defined $credential->{$key}) {
1175 print $writer $key, '=', $credential->{$key}, "\n";
1181 sub _credential_run {
1182 my ($self, $credential, $op) = _maybe_self(@_);
1183 my ($pid, $reader, $writer, $ctx) = command_bidi_pipe('credential', $op);
1185 credential_write $writer, $credential;
1188 if ($op eq "fill") {
1189 %$credential = credential_read $reader;
1192 throw Error::Simple("unexpected output from git credential $op response:\n$_\n");
1195 command_close_bidi_pipe($pid, $reader, undef, $ctx);
1198 =item credential( CREDENTIAL_HASHREF [, OPERATION ] )
1200 =item credential( CREDENTIAL_HASHREF, CODE )
1202 Executes C<git credential> for a given set of credentials and specified
1203 operation. In both forms C<CREDENTIAL_HASHREF> needs to be a reference to
1204 a hash which stores credentials. Under certain conditions the hash can
1207 In the first form, C<OPERATION> can be C<'fill'>, C<'approve'> or C<'reject'>,
1208 and function will execute corresponding C<git credential> sub-command. If
1209 it's omitted C<'fill'> is assumed. In case of C<'fill'> the values stored in
1210 C<CREDENTIAL_HASHREF> will be changed to the ones returned by the C<git
1211 credential fill> command. The usual usage would look something like:
1214 'protocol' => 'https',
1215 'host' => 'example.com',
1218 Git::credential \%cred;
1219 if (try_to_authenticate($cred{'username'}, $cred{'password'})) {
1220 Git::credential \%cred, 'approve';
1221 ... do more stuff ...
1223 Git::credential \%cred, 'reject';
1226 In the second form, C<CODE> needs to be a reference to a subroutine. The
1227 function will execute C<git credential fill> to fill the provided credential
1228 hash, then call C<CODE> with C<CREDENTIAL_HASHREF> as the sole argument. If
1229 C<CODE>'s return value is defined, the function will execute C<git credential
1230 approve> (if return value yields true) or C<git credential reject> (if return
1231 value is false). If the return value is undef, nothing at all is executed;
1232 this is useful, for example, if the credential could neither be verified nor
1233 rejected due to an unrelated network error. The return value is the same as
1234 what C<CODE> returns. With this form, the usage might look as follows:
1236 if (Git::credential {
1237 'protocol' => 'https',
1238 'host' => 'example.com',
1242 return !!try_to_authenticate($cred->{'username'},
1243 $cred->{'password'});
1245 ... do more stuff ...
1251 my ($self, $credential, $op_or_code) = (_maybe_self(@_), 'fill');
1253 if ('CODE' eq ref $op_or_code) {
1254 _credential_run $credential, 'fill';
1255 my $ret = $op_or_code->($credential);
1257 _credential_run $credential, $ret ? 'approve' : 'reject';
1261 _credential_run $credential, $op_or_code;
1265 { # %TEMP_* Lexical Context
1267 my (%TEMP_FILEMAP, %TEMP_FILES);
1269 =item temp_acquire ( NAME )
1271 Attempts to retrieve the temporary file mapped to the string C<NAME>. If an
1272 associated temp file has not been created this session or was closed, it is
1273 created, cached, and set for autoflush and binmode.
1275 Internally locks the file mapped to C<NAME>. This lock must be released with
1276 C<temp_release()> when the temp file is no longer needed. Subsequent attempts
1277 to retrieve temporary files mapped to the same C<NAME> while still locked will
1278 cause an error. This locking mechanism provides a weak guarantee and is not
1279 threadsafe. It does provide some error checking to help prevent temp file refs
1280 writing over one another.
1282 In general, the L<File::Handle> returned should not be closed by consumers as
1283 it defeats the purpose of this caching mechanism. If you need to close the temp
1284 file handle, then you should use L<File::Temp> or another temp file faculty
1285 directly. If a handle is closed and then requested again, then a warning will
1291 my $temp_fd = _temp_cache(@_);
1293 $TEMP_FILES{$temp_fd}{locked} = 1;
1297 =item temp_is_locked ( NAME )
1299 Returns true if the internal lock created by a previous C<temp_acquire()>
1300 call with C<NAME> is still in effect.
1302 When temp_acquire is called on a C<NAME>, it internally locks the temporary
1303 file mapped to C<NAME>. That lock will not be released until C<temp_release()>
1304 is called with either the original C<NAME> or the L<File::Handle> that was
1305 returned from the original call to temp_acquire.
1307 Subsequent attempts to call C<temp_acquire()> with the same C<NAME> will fail
1308 unless there has been an intervening C<temp_release()> call for that C<NAME>
1309 (or its corresponding L<File::Handle> that was returned by the original
1310 C<temp_acquire()> call).
1312 If true is returned by C<temp_is_locked()> for a C<NAME>, an attempt to
1313 C<temp_acquire()> the same C<NAME> will cause an error unless
1314 C<temp_release> is first called on that C<NAME> (or its corresponding
1315 L<File::Handle> that was returned by the original C<temp_acquire()> call).
1319 sub temp_is_locked {
1320 my ($self, $name) = _maybe_self(@_);
1321 my $temp_fd = \$TEMP_FILEMAP{$name};
1323 defined $$temp_fd && $$temp_fd->opened && $TEMP_FILES{$$temp_fd}{locked};
1326 =item temp_release ( NAME )
1328 =item temp_release ( FILEHANDLE )
1330 Releases a lock acquired through C<temp_acquire()>. Can be called either with
1331 the C<NAME> mapping used when acquiring the temp file or with the C<FILEHANDLE>
1332 referencing a locked temp file.
1334 Warns if an attempt is made to release a file that is not locked.
1336 The temp file will be truncated before being released. This can help to reduce
1337 disk I/O where the system is smart enough to detect the truncation while data
1338 is in the output buffers. Beware that after the temp file is released and
1339 truncated, any operations on that file may fail miserably until it is
1340 re-acquired. All contents are lost between each release and acquire mapped to
1346 my ($self, $temp_fd, $trunc) = _maybe_self(@_);
1348 if (exists $TEMP_FILEMAP{$temp_fd}) {
1349 $temp_fd = $TEMP_FILES{$temp_fd};
1351 unless ($TEMP_FILES{$temp_fd}{locked}) {
1352 carp "Attempt to release temp file '",
1353 $temp_fd, "' that has not been locked";
1355 temp_reset($temp_fd) if $trunc and $temp_fd->opened;
1357 $TEMP_FILES{$temp_fd}{locked} = 0;
1362 my ($self, $name) = _maybe_self(@_);
1366 my $temp_fd = \$TEMP_FILEMAP{$name};
1367 if (defined $$temp_fd and $$temp_fd->opened) {
1368 if ($TEMP_FILES{$$temp_fd}{locked}) {
1369 throw Error::Simple("Temp file with moniker '" .
1370 $name . "' already in use");
1373 if (defined $$temp_fd) {
1374 # then we're here because of a closed handle.
1375 carp "Temp file '", $name,
1376 "' was closed. Opening replacement.";
1381 if (defined $self) {
1382 $tmpdir = $self->repo_path();
1386 $n =~ s/\W/_/g; # no strange chars
1388 ($$temp_fd, $fname) = File::Temp::tempfile(
1389 "Git_${n}_XXXXXX", UNLINK => 1, DIR => $tmpdir,
1390 ) or throw Error::Simple("couldn't open new temp file");
1392 $$temp_fd->autoflush;
1394 $TEMP_FILES{$$temp_fd}{fname} = $fname;
1399 sub _verify_require {
1400 eval { require File::Temp; require File::Spec; };
1401 $@ and throw Error::Simple($@);
1404 =item temp_reset ( FILEHANDLE )
1406 Truncates and resets the position of the C<FILEHANDLE>.
1411 my ($self, $temp_fd) = _maybe_self(@_);
1413 truncate $temp_fd, 0
1414 or throw Error::Simple("couldn't truncate file");
1415 sysseek($temp_fd, 0, SEEK_SET) and seek($temp_fd, 0, SEEK_SET)
1416 or throw Error::Simple("couldn't seek to beginning of file");
1417 sysseek($temp_fd, 0, SEEK_CUR) == 0 and tell($temp_fd) == 0
1418 or throw Error::Simple("expected file position to be reset");
1421 =item temp_path ( NAME )
1423 =item temp_path ( FILEHANDLE )
1425 Returns the filename associated with the given tempfile.
1430 my ($self, $temp_fd) = _maybe_self(@_);
1432 if (exists $TEMP_FILEMAP{$temp_fd}) {
1433 $temp_fd = $TEMP_FILEMAP{$temp_fd};
1435 $TEMP_FILES{$temp_fd}{fname};
1439 unlink values %TEMP_FILEMAP if %TEMP_FILEMAP;
1442 } # %TEMP_* Lexical Context
1444 =item prefix_lines ( PREFIX, STRING [, STRING... ])
1446 Prefixes lines in C<STRING> with C<PREFIX>.
1452 my $string = join("\n", @_);
1453 $string =~ s/^/$prefix/mg;
1457 =item unquote_path ( PATH )
1459 Unquote a quoted path containing c-escapes as returned by ls-files etc.
1460 when not using -z or when parsing the output of diff -u.
1479 my ($retval, $remainder);
1480 if (!/^\042(.*)\042$/) {
1483 ($_, $retval) = ($1, "");
1484 while (/^([^\\]*)\\(.*)$/) {
1488 if (/^([0-3][0-7][0-7])(.*)$/) {
1489 $retval .= chr(oct($1));
1493 if (/^([\\\042abtnvfr])(.*)$/) {
1494 $retval .= $cquote_map{$1};
1499 throw Error::Simple("invalid quoted path $_[0]");
1508 =item get_comment_line_char ( )
1510 Gets the core.commentchar configuration value.
1511 The value falls-back to '#' if core.commentchar is set to 'auto'.
1515 sub get_comment_line_char {
1516 my $comment_line_char = config("core.commentchar") || '#';
1517 $comment_line_char = '#' if ($comment_line_char eq 'auto');
1518 $comment_line_char = '#' if (length($comment_line_char) != 1);
1519 return $comment_line_char;
1522 =item comment_lines ( STRING [, STRING... ])
1524 Comments lines following core.commentchar configuration.
1529 my $comment_line_char = get_comment_line_char;
1530 return prefix_lines("$comment_line_char ", @_);
1535 =head1 ERROR HANDLING
1537 All functions are supposed to throw Perl exceptions in case of errors.
1538 See the L<Error> module on how to catch those. Most exceptions are mere
1539 L<Error::Simple> instances.
1541 However, the C<command()>, C<command_oneline()> and C<command_noisy()>
1542 functions suite can throw C<Git::Error::Command> exceptions as well: those are
1543 thrown when the external command returns an error code and contain the error
1544 code as well as access to the captured command's output. The exception class
1545 provides the usual C<stringify> and C<value> (command's exit code) methods and
1546 in addition also a C<cmd_output> method that returns either an array or a
1547 string with the captured command output (depending on the original function
1548 call context; C<command_noisy()> returns C<undef>) and $<cmdline> which
1549 returns the command and its arguments (but without proper quoting).
1551 Note that the C<command_*_pipe()> functions cannot throw this exception since
1552 it has no idea whether the command failed or not. You will only find out
1553 at the time you C<close> the pipe; if you want to have that automated,
1554 use C<command_close_pipe()>, which can throw the exception.
1559 package Git::Error::Command;
1561 @Git::Error::Command::ISA = qw(Error);
1565 my $cmdline = '' . shift;
1566 my $value = 0 + shift;
1567 my $outputref = shift;
1570 local $Error::Depth = $Error::Depth + 1;
1572 push(@args, '-cmdline', $cmdline);
1573 push(@args, '-value', $value);
1574 push(@args, '-outputref', $outputref);
1576 $self->SUPER::new(-text => 'command returned error', @args);
1581 my $text = $self->SUPER::stringify;
1582 $self->cmdline() . ': ' . $text . ': ' . $self->value() . "\n";
1587 $self->{'-cmdline'};
1592 my $ref = $self->{'-outputref'};
1593 defined $ref or undef;
1594 if (ref $ref eq 'ARRAY') {
1604 =item git_cmd_try { CODE } ERRMSG
1606 This magical statement will automatically catch any C<Git::Error::Command>
1607 exceptions thrown by C<CODE> and make your program die with C<ERRMSG>
1608 on its lips; the message will have %s substituted for the command line
1609 and %d for the exit status. This statement is useful mostly for producing
1610 more user-friendly error messages.
1612 In case of no exception caught the statement returns C<CODE>'s return value.
1614 Note that this is the only auto-exported function.
1618 sub git_cmd_try(&$) {
1619 my ($code, $errmsg) = @_;
1622 my $array = wantarray;
1627 $result[0] = &$code;
1629 } catch Git::Error::Command with {
1632 $err =~ s/\%s/$E->cmdline()/ge;
1633 $err =~ s/\%d/$E->value()/ge;
1634 # We can't croak here since Error.pm would mangle
1635 # that to Error::Simple.
1637 $err and croak $err;
1638 return $array ? @result : $result[0];
1646 Copyright 2006 by Petr Baudis E<lt>pasky@suse.czE<gt>.
1648 This module is free software; it may be used, copied, modified
1649 and distributed under the terms of the GNU General Public Licence,
1650 either version 2, or (at your option) any later version.
1655 # Take raw method argument list and return ($obj, @args) in case
1656 # the method was called upon an instance and (undef, @args) if
1657 # it was called directly.
1659 UNIVERSAL::isa($_[0], 'Git') ? @_ : (undef, @_);
1662 # Check if the command id is something reasonable.
1663 sub _check_valid_cmd {
1665 $cmd =~ /^[a-z0-9A-Z_-]+$/ or throw Error::Simple("bad command: $cmd");
1668 # Common backend for the pipe creators.
1669 sub _command_common_pipe {
1670 my $direction = shift;
1671 my ($self, @p) = _maybe_self(@_);
1672 my (%opts, $cmd, @args);
1674 ($cmd, @args) = @{shift @p};
1675 %opts = ref $p[0] ? %{$p[0]} : @p;
1679 _check_valid_cmd($cmd);
1682 if ($^O eq 'MSWin32') {
1684 #defined $opts{STDERR} and
1685 # warn 'ignoring STDERR option - running w/ ActiveState';
1686 $direction eq '-|' or
1687 die 'input pipe for ActiveState not implemented';
1688 # the strange construction with *ACPIPE is just to
1689 # explain the tie below that we want to bind to
1690 # a handle class, not scalar. It is not known if
1691 # it is something specific to ActiveState Perl or
1692 # just a Perl quirk.
1693 tie (*ACPIPE, 'Git::activestate_pipe', $cmd, @args);
1697 my $pid = open($fh, $direction);
1698 if (not defined $pid) {
1699 throw Error::Simple("open failed: $!");
1700 } elsif ($pid == 0) {
1701 if ($opts{STDERR}) {
1702 open (STDERR, '>&', $opts{STDERR})
1703 or die "dup failed: $!";
1704 } elsif (defined $opts{STDERR}) {
1705 open (STDERR, '>', '/dev/null')
1706 or die "opening /dev/null failed: $!";
1708 _cmd_exec($self, $cmd, @args);
1711 return wantarray ? ($fh, join(' ', $cmd, @args)) : $fh;
1714 # When already in the subprocess, set up the appropriate state
1715 # for the given repository and execute the git command.
1717 my ($self, @args) = @_;
1718 _setup_git_cmd_env($self);
1719 _execv_git_cmd(@args);
1720 die qq[exec "@args" failed: $!];
1723 # set up the appropriate state for git command
1724 sub _setup_git_cmd_env {
1727 $self->repo_path() and $ENV{'GIT_DIR'} = $self->repo_path();
1728 $self->repo_path() and $self->wc_path()
1729 and $ENV{'GIT_WORK_TREE'} = $self->wc_path();
1730 $self->wc_path() and chdir($self->wc_path());
1731 $self->wc_subdir() and chdir($self->wc_subdir());
1735 # Execute the given Git command ($_[0]) with arguments ($_[1..])
1736 # by searching for it at proper places.
1737 sub _execv_git_cmd { exec('git', @_); }
1739 # Close pipe to a subprocess.
1742 foreach my $fh (@_) {
1746 # It's just close, no point in fatalities
1747 carp "error closing pipe: $!";
1749 # The caller should pepper this.
1750 throw Git::Error::Command($ctx, $? >> 8);
1752 # else we might e.g. closed a live stream; the command
1753 # dying of SIGPIPE would drive us here.
1760 $self->_close_hash_and_insert_object();
1761 $self->_close_cat_blob();
1765 # Pipe implementation for ActiveState Perl.
1767 package Git::activestate_pipe;
1771 my ($class, @params) = @_;
1772 # FIXME: This is probably horrible idea and the thing will explode
1773 # at the moment you give it arguments that require some quoting,
1774 # but I have no ActiveState clue... --pasky
1775 # Let's just hope ActiveState Perl does at least the quoting
1777 my @data = qx{git @params};
1778 bless { i => 0, data => \@data }, $class;
1783 if ($self->{i} >= scalar @{$self->{data}}) {
1788 $self->{i} = $#{$self->{'data'}} + 1;
1789 return splice(@{$self->{'data'}}, $i);
1791 $self->{i} = $i + 1;
1792 return $self->{'data'}->[ $i ];
1797 delete $self->{data};
1803 return ($self->{i} >= scalar @{$self->{data}});
1807 1; # Famous last words