Git.pm: call Error::Simple() properly
[git/git.git] / perl / Git.pm
1 =head1 NAME
2
3 Git - Perl interface to the Git version control system
4
5 =cut
6
7
8 package Git;
9
10 use strict;
11
12
13 BEGIN {
14
15 our ($VERSION, @ISA, @EXPORT, @EXPORT_OK);
16
17 # Totally unstable API.
18 $VERSION = '0.01';
19
20
21 =head1 SYNOPSIS
22
23 use Git;
24
25 my $version = Git::command_oneline('version');
26
27 git_cmd_try { Git::command_noisy('update-server-info') }
28 '%s failed w/ code %d';
29
30 my $repo = Git->repository (Directory => '/srv/git/cogito.git');
31
32
33 my @revs = $repo->command('rev-list', '--since=last monday', '--all');
34
35 my ($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);
38
39 my $lastrev = $repo->command_oneline( [ 'rev-list', '--all' ],
40 STDERR => 0 );
41
42 my $sha1 = $repo->hash_and_insert_object('file.txt');
43 my $tempfile = tempfile();
44 my $size = $repo->cat_blob($sha1, $tempfile);
45
46 =cut
47
48
49 require Exporter;
50
51 @ISA = qw(Exporter);
52
53 @EXPORT = qw(git_cmd_try);
54
55 # Methods which can be called as standalone functions as well:
56 @EXPORT_OK = qw(command command_oneline command_noisy
57 command_output_pipe command_input_pipe command_close_pipe
58 command_bidi_pipe command_close_bidi_pipe
59 version exec_path hash_object git_cmd_try
60 remote_refs
61 temp_acquire temp_release temp_reset temp_path);
62
63
64 =head1 DESCRIPTION
65
66 This module provides Perl scripts easy way to interface the Git version control
67 system. The modules have an easy and well-tested way to call arbitrary Git
68 commands; in the future, the interface will also provide specialized methods
69 for doing easily operations which are not totally trivial to do over
70 the generic command interface.
71
72 While some commands can be executed outside of any context (e.g. 'version'
73 or 'init'), most operations require a repository context, which in practice
74 means getting an instance of the Git object using the repository() constructor.
75 (In the future, we will also get a new_repository() constructor.) All commands
76 called as methods of the object are then executed in the context of the
77 repository.
78
79 Part of the "repository state" is also information about path to the attached
80 working copy (unless you work with a bare repository). You can also navigate
81 inside of the working copy using the C<wc_chdir()> method. (Note that
82 the repository object is self-contained and will not change working directory
83 of your process.)
84
85 TODO: In the future, we might also do
86
87 my $remoterepo = $repo->remote_repository (Name => 'cogito', Branch => 'master');
88 $remoterepo ||= Git->remote_repository ('http://git.or.cz/cogito.git/');
89 my @refs = $remoterepo->refs();
90
91 Currently, the module merely wraps calls to external Git tools. In the future,
92 it will provide a much faster way to interact with Git by linking directly
93 to libgit. This should be completely opaque to the user, though (performance
94 increase notwithstanding).
95
96 =cut
97
98
99 use Carp qw(carp croak); # but croak is bad - throw instead
100 use Error qw(:try);
101 use Cwd qw(abs_path);
102 use IPC::Open2 qw(open2);
103 use Fcntl qw(SEEK_SET SEEK_CUR);
104 }
105
106
107 =head1 CONSTRUCTORS
108
109 =over 4
110
111 =item repository ( OPTIONS )
112
113 =item repository ( DIRECTORY )
114
115 =item repository ()
116
117 Construct a new repository object.
118 C<OPTIONS> are passed in a hash like fashion, using key and value pairs.
119 Possible options are:
120
121 B<Repository> - Path to the Git repository.
122
123 B<WorkingCopy> - Path to the associated working copy; not strictly required
124 as many commands will happily crunch on a bare repository.
125
126 B<WorkingSubdir> - Subdirectory in the working copy to work inside.
127 Just left undefined if you do not want to limit the scope of operations.
128
129 B<Directory> - Path to the Git working directory in its usual setup.
130 The C<.git> directory is searched in the directory and all the parent
131 directories; if found, C<WorkingCopy> is set to the directory containing
132 it and C<Repository> to the C<.git> directory itself. If no C<.git>
133 directory was found, the C<Directory> is assumed to be a bare repository,
134 C<Repository> is set to point at it and C<WorkingCopy> is left undefined.
135 If the C<$GIT_DIR> environment variable is set, things behave as expected
136 as well.
137
138 You should not use both C<Directory> and either of C<Repository> and
139 C<WorkingCopy> - the results of that are undefined.
140
141 Alternatively, a directory path may be passed as a single scalar argument
142 to the constructor; it is equivalent to setting only the C<Directory> option
143 field.
144
145 Calling the constructor with no options whatsoever is equivalent to
146 calling it with C<< Directory => '.' >>. In general, if you are building
147 a standard porcelain command, simply doing C<< Git->repository() >> should
148 do the right thing and setup the object to reflect exactly where the user
149 is right now.
150
151 =cut
152
153 sub repository {
154 my $class = shift;
155 my @args = @_;
156 my %opts = ();
157 my $self;
158
159 if (defined $args[0]) {
160 if ($#args % 2 != 1) {
161 # Not a hash.
162 $#args == 0 or throw Error::Simple("bad usage");
163 %opts = ( Directory => $args[0] );
164 } else {
165 %opts = @args;
166 }
167 }
168
169 if (not defined $opts{Repository} and not defined $opts{WorkingCopy}) {
170 $opts{Directory} ||= '.';
171 }
172
173 if ($opts{Directory}) {
174 -d $opts{Directory} or throw Error::Simple("Directory not found: $!");
175
176 my $search = Git->repository(WorkingCopy => $opts{Directory});
177 my $dir;
178 try {
179 $dir = $search->command_oneline(['rev-parse', '--git-dir'],
180 STDERR => 0);
181 } catch Git::Error::Command with {
182 $dir = undef;
183 };
184
185 if ($dir) {
186 $dir =~ m#^/# or $dir = $opts{Directory} . '/' . $dir;
187 $opts{Repository} = $dir;
188
189 # If --git-dir went ok, this shouldn't die either.
190 my $prefix = $search->command_oneline('rev-parse', '--show-prefix');
191 $dir = abs_path($opts{Directory}) . '/';
192 if ($prefix) {
193 if (substr($dir, -length($prefix)) ne $prefix) {
194 throw Error::Simple("rev-parse confused me - $dir does not have trailing $prefix");
195 }
196 substr($dir, -length($prefix)) = '';
197 }
198 $opts{WorkingCopy} = $dir;
199 $opts{WorkingSubdir} = $prefix;
200
201 } else {
202 # A bare repository? Let's see...
203 $dir = $opts{Directory};
204
205 unless (-d "$dir/refs" and -d "$dir/objects" and -e "$dir/HEAD") {
206 # Mimick git-rev-parse --git-dir error message:
207 throw Error::Simple('fatal: Not a git repository');
208 }
209 my $search = Git->repository(Repository => $dir);
210 try {
211 $search->command('symbolic-ref', 'HEAD');
212 } catch Git::Error::Command with {
213 # Mimick git-rev-parse --git-dir error message:
214 throw Error::Simple('fatal: Not a git repository');
215 }
216
217 $opts{Repository} = abs_path($dir);
218 }
219
220 delete $opts{Directory};
221 }
222
223 $self = { opts => \%opts };
224 bless $self, $class;
225 }
226
227 =back
228
229 =head1 METHODS
230
231 =over 4
232
233 =item command ( COMMAND [, ARGUMENTS... ] )
234
235 =item command ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
236
237 Execute the given Git C<COMMAND> (specify it without the 'git-'
238 prefix), optionally with the specified extra C<ARGUMENTS>.
239
240 The second more elaborate form can be used if you want to further adjust
241 the command execution. Currently, only one option is supported:
242
243 B<STDERR> - How to deal with the command's error output. By default (C<undef>)
244 it is delivered to the caller's C<STDERR>. A false value (0 or '') will cause
245 it to be thrown away. If you want to process it, you can get it in a filehandle
246 you specify, but you must be extremely careful; if the error output is not
247 very short and you want to read it in the same process as where you called
248 C<command()>, you are set up for a nice deadlock!
249
250 The method can be called without any instance or on a specified Git repository
251 (in that case the command will be run in the repository context).
252
253 In scalar context, it returns all the command output in a single string
254 (verbatim).
255
256 In array context, it returns an array containing lines printed to the
257 command's stdout (without trailing newlines).
258
259 In both cases, the command's stdin and stderr are the same as the caller's.
260
261 =cut
262
263 sub command {
264 my ($fh, $ctx) = command_output_pipe(@_);
265
266 if (not defined wantarray) {
267 # Nothing to pepper the possible exception with.
268 _cmd_close($fh, $ctx);
269
270 } elsif (not wantarray) {
271 local $/;
272 my $text = <$fh>;
273 try {
274 _cmd_close($fh, $ctx);
275 } catch Git::Error::Command with {
276 # Pepper with the output:
277 my $E = shift;
278 $E->{'-outputref'} = \$text;
279 throw $E;
280 };
281 return $text;
282
283 } else {
284 my @lines = <$fh>;
285 defined and chomp for @lines;
286 try {
287 _cmd_close($fh, $ctx);
288 } catch Git::Error::Command with {
289 my $E = shift;
290 $E->{'-outputref'} = \@lines;
291 throw $E;
292 };
293 return @lines;
294 }
295 }
296
297
298 =item command_oneline ( COMMAND [, ARGUMENTS... ] )
299
300 =item command_oneline ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
301
302 Execute the given C<COMMAND> in the same way as command()
303 does but always return a scalar string containing the first line
304 of the command's standard output.
305
306 =cut
307
308 sub command_oneline {
309 my ($fh, $ctx) = command_output_pipe(@_);
310
311 my $line = <$fh>;
312 defined $line and chomp $line;
313 try {
314 _cmd_close($fh, $ctx);
315 } catch Git::Error::Command with {
316 # Pepper with the output:
317 my $E = shift;
318 $E->{'-outputref'} = \$line;
319 throw $E;
320 };
321 return $line;
322 }
323
324
325 =item command_output_pipe ( COMMAND [, ARGUMENTS... ] )
326
327 =item command_output_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
328
329 Execute the given C<COMMAND> in the same way as command()
330 does but return a pipe filehandle from which the command output can be
331 read.
332
333 The function can return C<($pipe, $ctx)> in array context.
334 See C<command_close_pipe()> for details.
335
336 =cut
337
338 sub command_output_pipe {
339 _command_common_pipe('-|', @_);
340 }
341
342
343 =item command_input_pipe ( COMMAND [, ARGUMENTS... ] )
344
345 =item command_input_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
346
347 Execute the given C<COMMAND> in the same way as command_output_pipe()
348 does but return an input pipe filehandle instead; the command output
349 is not captured.
350
351 The function can return C<($pipe, $ctx)> in array context.
352 See C<command_close_pipe()> for details.
353
354 =cut
355
356 sub command_input_pipe {
357 _command_common_pipe('|-', @_);
358 }
359
360
361 =item command_close_pipe ( PIPE [, CTX ] )
362
363 Close the C<PIPE> as returned from C<command_*_pipe()>, checking
364 whether the command finished successfully. The optional C<CTX> argument
365 is required if you want to see the command name in the error message,
366 and it is the second value returned by C<command_*_pipe()> when
367 called in array context. The call idiom is:
368
369 my ($fh, $ctx) = $r->command_output_pipe('status');
370 while (<$fh>) { ... }
371 $r->command_close_pipe($fh, $ctx);
372
373 Note that you should not rely on whatever actually is in C<CTX>;
374 currently it is simply the command name but in future the context might
375 have more complicated structure.
376
377 =cut
378
379 sub command_close_pipe {
380 my ($self, $fh, $ctx) = _maybe_self(@_);
381 $ctx ||= '<unknown>';
382 _cmd_close($fh, $ctx);
383 }
384
385 =item command_bidi_pipe ( COMMAND [, ARGUMENTS... ] )
386
387 Execute the given C<COMMAND> in the same way as command_output_pipe()
388 does but return both an input pipe filehandle and an output pipe filehandle.
389
390 The function will return return C<($pid, $pipe_in, $pipe_out, $ctx)>.
391 See C<command_close_bidi_pipe()> for details.
392
393 =cut
394
395 sub command_bidi_pipe {
396 my ($pid, $in, $out);
397 $pid = open2($in, $out, 'git', @_);
398 return ($pid, $in, $out, join(' ', @_));
399 }
400
401 =item command_close_bidi_pipe ( PID, PIPE_IN, PIPE_OUT [, CTX] )
402
403 Close the C<PIPE_IN> and C<PIPE_OUT> as returned from C<command_bidi_pipe()>,
404 checking whether the command finished successfully. The optional C<CTX>
405 argument is required if you want to see the command name in the error message,
406 and it is the fourth value returned by C<command_bidi_pipe()>. The call idiom
407 is:
408
409 my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
410 print "000000000\n" $out;
411 while (<$in>) { ... }
412 $r->command_close_bidi_pipe($pid, $in, $out, $ctx);
413
414 Note that you should not rely on whatever actually is in C<CTX>;
415 currently it is simply the command name but in future the context might
416 have more complicated structure.
417
418 =cut
419
420 sub command_close_bidi_pipe {
421 local $?;
422 my ($pid, $in, $out, $ctx) = @_;
423 foreach my $fh ($in, $out) {
424 unless (close $fh) {
425 if ($!) {
426 carp "error closing pipe: $!";
427 } elsif ($? >> 8) {
428 throw Git::Error::Command($ctx, $? >>8);
429 }
430 }
431 }
432
433 waitpid $pid, 0;
434
435 if ($? >> 8) {
436 throw Git::Error::Command($ctx, $? >>8);
437 }
438 }
439
440
441 =item command_noisy ( COMMAND [, ARGUMENTS... ] )
442
443 Execute the given C<COMMAND> in the same way as command() does but do not
444 capture the command output - the standard output is not redirected and goes
445 to the standard output of the caller application.
446
447 While the method is called command_noisy(), you might want to as well use
448 it for the most silent Git commands which you know will never pollute your
449 stdout but you want to avoid the overhead of the pipe setup when calling them.
450
451 The function returns only after the command has finished running.
452
453 =cut
454
455 sub command_noisy {
456 my ($self, $cmd, @args) = _maybe_self(@_);
457 _check_valid_cmd($cmd);
458
459 my $pid = fork;
460 if (not defined $pid) {
461 throw Error::Simple("fork failed: $!");
462 } elsif ($pid == 0) {
463 _cmd_exec($self, $cmd, @args);
464 }
465 if (waitpid($pid, 0) > 0 and $?>>8 != 0) {
466 throw Git::Error::Command(join(' ', $cmd, @args), $? >> 8);
467 }
468 }
469
470
471 =item version ()
472
473 Return the Git version in use.
474
475 =cut
476
477 sub version {
478 my $verstr = command_oneline('--version');
479 $verstr =~ s/^git version //;
480 $verstr;
481 }
482
483
484 =item exec_path ()
485
486 Return path to the Git sub-command executables (the same as
487 C<git --exec-path>). Useful mostly only internally.
488
489 =cut
490
491 sub exec_path { command_oneline('--exec-path') }
492
493
494 =item repo_path ()
495
496 Return path to the git repository. Must be called on a repository instance.
497
498 =cut
499
500 sub repo_path { $_[0]->{opts}->{Repository} }
501
502
503 =item wc_path ()
504
505 Return path to the working copy. Must be called on a repository instance.
506
507 =cut
508
509 sub wc_path { $_[0]->{opts}->{WorkingCopy} }
510
511
512 =item wc_subdir ()
513
514 Return path to the subdirectory inside of a working copy. Must be called
515 on a repository instance.
516
517 =cut
518
519 sub wc_subdir { $_[0]->{opts}->{WorkingSubdir} ||= '' }
520
521
522 =item wc_chdir ( SUBDIR )
523
524 Change the working copy subdirectory to work within. The C<SUBDIR> is
525 relative to the working copy root directory (not the current subdirectory).
526 Must be called on a repository instance attached to a working copy
527 and the directory must exist.
528
529 =cut
530
531 sub wc_chdir {
532 my ($self, $subdir) = @_;
533 $self->wc_path()
534 or throw Error::Simple("bare repository");
535
536 -d $self->wc_path().'/'.$subdir
537 or throw Error::Simple("subdir not found: $!");
538 # Of course we will not "hold" the subdirectory so anyone
539 # can delete it now and we will never know. But at least we tried.
540
541 $self->{opts}->{WorkingSubdir} = $subdir;
542 }
543
544
545 =item config ( VARIABLE )
546
547 Retrieve the configuration C<VARIABLE> in the same manner as C<config>
548 does. In scalar context requires the variable to be set only one time
549 (exception is thrown otherwise), in array context returns allows the
550 variable to be set multiple times and returns all the values.
551
552 This currently wraps command('config') so it is not so fast.
553
554 =cut
555
556 sub config {
557 my ($self, $var) = _maybe_self(@_);
558
559 try {
560 my @cmd = ('config');
561 unshift @cmd, $self if $self;
562 if (wantarray) {
563 return command(@cmd, '--get-all', $var);
564 } else {
565 return command_oneline(@cmd, '--get', $var);
566 }
567 } catch Git::Error::Command with {
568 my $E = shift;
569 if ($E->value() == 1) {
570 # Key not found.
571 return;
572 } else {
573 throw $E;
574 }
575 };
576 }
577
578
579 =item config_bool ( VARIABLE )
580
581 Retrieve the bool configuration C<VARIABLE>. The return value
582 is usable as a boolean in perl (and C<undef> if it's not defined,
583 of course).
584
585 This currently wraps command('config') so it is not so fast.
586
587 =cut
588
589 sub config_bool {
590 my ($self, $var) = _maybe_self(@_);
591
592 try {
593 my @cmd = ('config', '--bool', '--get', $var);
594 unshift @cmd, $self if $self;
595 my $val = command_oneline(@cmd);
596 return undef unless defined $val;
597 return $val eq 'true';
598 } catch Git::Error::Command with {
599 my $E = shift;
600 if ($E->value() == 1) {
601 # Key not found.
602 return undef;
603 } else {
604 throw $E;
605 }
606 };
607 }
608
609 =item config_int ( VARIABLE )
610
611 Retrieve the integer configuration C<VARIABLE>. The return value
612 is simple decimal number. An optional value suffix of 'k', 'm',
613 or 'g' in the config file will cause the value to be multiplied
614 by 1024, 1048576 (1024^2), or 1073741824 (1024^3) prior to output.
615 It would return C<undef> if configuration variable is not defined,
616
617 This currently wraps command('config') so it is not so fast.
618
619 =cut
620
621 sub config_int {
622 my ($self, $var) = _maybe_self(@_);
623
624 try {
625 my @cmd = ('config', '--int', '--get', $var);
626 unshift @cmd, $self if $self;
627 return command_oneline(@cmd);
628 } catch Git::Error::Command with {
629 my $E = shift;
630 if ($E->value() == 1) {
631 # Key not found.
632 return undef;
633 } else {
634 throw $E;
635 }
636 };
637 }
638
639 =item get_colorbool ( NAME )
640
641 Finds if color should be used for NAMEd operation from the configuration,
642 and returns boolean (true for "use color", false for "do not use color").
643
644 =cut
645
646 sub get_colorbool {
647 my ($self, $var) = @_;
648 my $stdout_to_tty = (-t STDOUT) ? "true" : "false";
649 my $use_color = $self->command_oneline('config', '--get-colorbool',
650 $var, $stdout_to_tty);
651 return ($use_color eq 'true');
652 }
653
654 =item get_color ( SLOT, COLOR )
655
656 Finds color for SLOT from the configuration, while defaulting to COLOR,
657 and returns the ANSI color escape sequence:
658
659 print $repo->get_color("color.interactive.prompt", "underline blue white");
660 print "some text";
661 print $repo->get_color("", "normal");
662
663 =cut
664
665 sub get_color {
666 my ($self, $slot, $default) = @_;
667 my $color = $self->command_oneline('config', '--get-color', $slot, $default);
668 if (!defined $color) {
669 $color = "";
670 }
671 return $color;
672 }
673
674 =item remote_refs ( REPOSITORY [, GROUPS [, REFGLOBS ] ] )
675
676 This function returns a hashref of refs stored in a given remote repository.
677 The hash is in the format C<refname =\> hash>. For tags, the C<refname> entry
678 contains the tag object while a C<refname^{}> entry gives the tagged objects.
679
680 C<REPOSITORY> has the same meaning as the appropriate C<git-ls-remote>
681 argument; either an URL or a remote name (if called on a repository instance).
682 C<GROUPS> is an optional arrayref that can contain 'tags' to return all the
683 tags and/or 'heads' to return all the heads. C<REFGLOB> is an optional array
684 of strings containing a shell-like glob to further limit the refs returned in
685 the hash; the meaning is again the same as the appropriate C<git-ls-remote>
686 argument.
687
688 This function may or may not be called on a repository instance. In the former
689 case, remote names as defined in the repository are recognized as repository
690 specifiers.
691
692 =cut
693
694 sub remote_refs {
695 my ($self, $repo, $groups, $refglobs) = _maybe_self(@_);
696 my @args;
697 if (ref $groups eq 'ARRAY') {
698 foreach (@$groups) {
699 if ($_ eq 'heads') {
700 push (@args, '--heads');
701 } elsif ($_ eq 'tags') {
702 push (@args, '--tags');
703 } else {
704 # Ignore unknown groups for future
705 # compatibility
706 }
707 }
708 }
709 push (@args, $repo);
710 if (ref $refglobs eq 'ARRAY') {
711 push (@args, @$refglobs);
712 }
713
714 my @self = $self ? ($self) : (); # Ultra trickery
715 my ($fh, $ctx) = Git::command_output_pipe(@self, 'ls-remote', @args);
716 my %refs;
717 while (<$fh>) {
718 chomp;
719 my ($hash, $ref) = split(/\t/, $_, 2);
720 $refs{$ref} = $hash;
721 }
722 Git::command_close_pipe(@self, $fh, $ctx);
723 return \%refs;
724 }
725
726
727 =item ident ( TYPE | IDENTSTR )
728
729 =item ident_person ( TYPE | IDENTSTR | IDENTARRAY )
730
731 This suite of functions retrieves and parses ident information, as stored
732 in the commit and tag objects or produced by C<var GIT_type_IDENT> (thus
733 C<TYPE> can be either I<author> or I<committer>; case is insignificant).
734
735 The C<ident> method retrieves the ident information from C<git var>
736 and either returns it as a scalar string or as an array with the fields parsed.
737 Alternatively, it can take a prepared ident string (e.g. from the commit
738 object) and just parse it.
739
740 C<ident_person> returns the person part of the ident - name and email;
741 it can take the same arguments as C<ident> or the array returned by C<ident>.
742
743 The synopsis is like:
744
745 my ($name, $email, $time_tz) = ident('author');
746 "$name <$email>" eq ident_person('author');
747 "$name <$email>" eq ident_person($name);
748 $time_tz =~ /^\d+ [+-]\d{4}$/;
749
750 =cut
751
752 sub ident {
753 my ($self, $type) = _maybe_self(@_);
754 my $identstr;
755 if (lc $type eq lc 'committer' or lc $type eq lc 'author') {
756 my @cmd = ('var', 'GIT_'.uc($type).'_IDENT');
757 unshift @cmd, $self if $self;
758 $identstr = command_oneline(@cmd);
759 } else {
760 $identstr = $type;
761 }
762 if (wantarray) {
763 return $identstr =~ /^(.*) <(.*)> (\d+ [+-]\d{4})$/;
764 } else {
765 return $identstr;
766 }
767 }
768
769 sub ident_person {
770 my ($self, @ident) = _maybe_self(@_);
771 $#ident == 0 and @ident = $self ? $self->ident($ident[0]) : ident($ident[0]);
772 return "$ident[0] <$ident[1]>";
773 }
774
775
776 =item hash_object ( TYPE, FILENAME )
777
778 Compute the SHA1 object id of the given C<FILENAME> considering it is
779 of the C<TYPE> object type (C<blob>, C<commit>, C<tree>).
780
781 The method can be called without any instance or on a specified Git repository,
782 it makes zero difference.
783
784 The function returns the SHA1 hash.
785
786 =cut
787
788 # TODO: Support for passing FILEHANDLE instead of FILENAME
789 sub hash_object {
790 my ($self, $type, $file) = _maybe_self(@_);
791 command_oneline('hash-object', '-t', $type, $file);
792 }
793
794
795 =item hash_and_insert_object ( FILENAME )
796
797 Compute the SHA1 object id of the given C<FILENAME> and add the object to the
798 object database.
799
800 The function returns the SHA1 hash.
801
802 =cut
803
804 # TODO: Support for passing FILEHANDLE instead of FILENAME
805 sub hash_and_insert_object {
806 my ($self, $filename) = @_;
807
808 carp "Bad filename \"$filename\"" if $filename =~ /[\r\n]/;
809
810 $self->_open_hash_and_insert_object_if_needed();
811 my ($in, $out) = ($self->{hash_object_in}, $self->{hash_object_out});
812
813 unless (print $out $filename, "\n") {
814 $self->_close_hash_and_insert_object();
815 throw Error::Simple("out pipe went bad");
816 }
817
818 chomp(my $hash = <$in>);
819 unless (defined($hash)) {
820 $self->_close_hash_and_insert_object();
821 throw Error::Simple("in pipe went bad");
822 }
823
824 return $hash;
825 }
826
827 sub _open_hash_and_insert_object_if_needed {
828 my ($self) = @_;
829
830 return if defined($self->{hash_object_pid});
831
832 ($self->{hash_object_pid}, $self->{hash_object_in},
833 $self->{hash_object_out}, $self->{hash_object_ctx}) =
834 command_bidi_pipe(qw(hash-object -w --stdin-paths));
835 }
836
837 sub _close_hash_and_insert_object {
838 my ($self) = @_;
839
840 return unless defined($self->{hash_object_pid});
841
842 my @vars = map { 'hash_object_' . $_ } qw(pid in out ctx);
843
844 command_close_bidi_pipe(@$self{@vars});
845 delete @$self{@vars};
846 }
847
848 =item cat_blob ( SHA1, FILEHANDLE )
849
850 Prints the contents of the blob identified by C<SHA1> to C<FILEHANDLE> and
851 returns the number of bytes printed.
852
853 =cut
854
855 sub cat_blob {
856 my ($self, $sha1, $fh) = @_;
857
858 $self->_open_cat_blob_if_needed();
859 my ($in, $out) = ($self->{cat_blob_in}, $self->{cat_blob_out});
860
861 unless (print $out $sha1, "\n") {
862 $self->_close_cat_blob();
863 throw Error::Simple("out pipe went bad");
864 }
865
866 my $description = <$in>;
867 if ($description =~ / missing$/) {
868 carp "$sha1 doesn't exist in the repository";
869 return -1;
870 }
871
872 if ($description !~ /^[0-9a-fA-F]{40} \S+ (\d+)$/) {
873 carp "Unexpected result returned from git cat-file";
874 return -1;
875 }
876
877 my $size = $1;
878
879 my $blob;
880 my $bytesRead = 0;
881
882 while (1) {
883 my $bytesLeft = $size - $bytesRead;
884 last unless $bytesLeft;
885
886 my $bytesToRead = $bytesLeft < 1024 ? $bytesLeft : 1024;
887 my $read = read($in, $blob, $bytesToRead, $bytesRead);
888 unless (defined($read)) {
889 $self->_close_cat_blob();
890 throw Error::Simple("in pipe went bad");
891 }
892
893 $bytesRead += $read;
894 }
895
896 # Skip past the trailing newline.
897 my $newline;
898 my $read = read($in, $newline, 1);
899 unless (defined($read)) {
900 $self->_close_cat_blob();
901 throw Error::Simple("in pipe went bad");
902 }
903 unless ($read == 1 && $newline eq "\n") {
904 $self->_close_cat_blob();
905 throw Error::Simple("didn't find newline after blob");
906 }
907
908 unless (print $fh $blob) {
909 $self->_close_cat_blob();
910 throw Error::Simple("couldn't write to passed in filehandle");
911 }
912
913 return $size;
914 }
915
916 sub _open_cat_blob_if_needed {
917 my ($self) = @_;
918
919 return if defined($self->{cat_blob_pid});
920
921 ($self->{cat_blob_pid}, $self->{cat_blob_in},
922 $self->{cat_blob_out}, $self->{cat_blob_ctx}) =
923 command_bidi_pipe(qw(cat-file --batch));
924 }
925
926 sub _close_cat_blob {
927 my ($self) = @_;
928
929 return unless defined($self->{cat_blob_pid});
930
931 my @vars = map { 'cat_blob_' . $_ } qw(pid in out ctx);
932
933 command_close_bidi_pipe(@$self{@vars});
934 delete @$self{@vars};
935 }
936
937
938 { # %TEMP_* Lexical Context
939
940 my (%TEMP_FILEMAP, %TEMP_FILES);
941
942 =item temp_acquire ( NAME )
943
944 Attempts to retreive the temporary file mapped to the string C<NAME>. If an
945 associated temp file has not been created this session or was closed, it is
946 created, cached, and set for autoflush and binmode.
947
948 Internally locks the file mapped to C<NAME>. This lock must be released with
949 C<temp_release()> when the temp file is no longer needed. Subsequent attempts
950 to retrieve temporary files mapped to the same C<NAME> while still locked will
951 cause an error. This locking mechanism provides a weak guarantee and is not
952 threadsafe. It does provide some error checking to help prevent temp file refs
953 writing over one another.
954
955 In general, the L<File::Handle> returned should not be closed by consumers as
956 it defeats the purpose of this caching mechanism. If you need to close the temp
957 file handle, then you should use L<File::Temp> or another temp file faculty
958 directly. If a handle is closed and then requested again, then a warning will
959 issue.
960
961 =cut
962
963 sub temp_acquire {
964 my ($self, $name) = _maybe_self(@_);
965
966 my $temp_fd = _temp_cache($name);
967
968 $TEMP_FILES{$temp_fd}{locked} = 1;
969 $temp_fd;
970 }
971
972 =item temp_release ( NAME )
973
974 =item temp_release ( FILEHANDLE )
975
976 Releases a lock acquired through C<temp_acquire()>. Can be called either with
977 the C<NAME> mapping used when acquiring the temp file or with the C<FILEHANDLE>
978 referencing a locked temp file.
979
980 Warns if an attempt is made to release a file that is not locked.
981
982 The temp file will be truncated before being released. This can help to reduce
983 disk I/O where the system is smart enough to detect the truncation while data
984 is in the output buffers. Beware that after the temp file is released and
985 truncated, any operations on that file may fail miserably until it is
986 re-acquired. All contents are lost between each release and acquire mapped to
987 the same string.
988
989 =cut
990
991 sub temp_release {
992 my ($self, $temp_fd, $trunc) = _maybe_self(@_);
993
994 if (exists $TEMP_FILEMAP{$temp_fd}) {
995 $temp_fd = $TEMP_FILES{$temp_fd};
996 }
997 unless ($TEMP_FILES{$temp_fd}{locked}) {
998 carp "Attempt to release temp file '",
999 $temp_fd, "' that has not been locked";
1000 }
1001 temp_reset($temp_fd) if $trunc and $temp_fd->opened;
1002
1003 $TEMP_FILES{$temp_fd}{locked} = 0;
1004 undef;
1005 }
1006
1007 sub _temp_cache {
1008 my ($name) = @_;
1009
1010 _verify_require();
1011
1012 my $temp_fd = \$TEMP_FILEMAP{$name};
1013 if (defined $$temp_fd and $$temp_fd->opened) {
1014 if ($TEMP_FILES{$$temp_fd}{locked}) {
1015 throw Error::Simple("Temp file with moniker '" .
1016 $name . "' already in use");
1017 }
1018 } else {
1019 if (defined $$temp_fd) {
1020 # then we're here because of a closed handle.
1021 carp "Temp file '", $name,
1022 "' was closed. Opening replacement.";
1023 }
1024 my $fname;
1025 ($$temp_fd, $fname) = File::Temp->tempfile(
1026 'Git_XXXXXX', UNLINK => 1
1027 ) or throw Error::Simple("couldn't open new temp file");
1028 $$temp_fd->autoflush;
1029 binmode $$temp_fd;
1030 $TEMP_FILES{$$temp_fd}{fname} = $fname;
1031 }
1032 $$temp_fd;
1033 }
1034
1035 sub _verify_require {
1036 eval { require File::Temp; require File::Spec; };
1037 $@ and throw Error::Simple($@);
1038 }
1039
1040 =item temp_reset ( FILEHANDLE )
1041
1042 Truncates and resets the position of the C<FILEHANDLE>.
1043
1044 =cut
1045
1046 sub temp_reset {
1047 my ($self, $temp_fd) = _maybe_self(@_);
1048
1049 truncate $temp_fd, 0
1050 or throw Error::Simple("couldn't truncate file");
1051 sysseek($temp_fd, 0, SEEK_SET) and seek($temp_fd, 0, SEEK_SET)
1052 or throw Error::Simple("couldn't seek to beginning of file");
1053 sysseek($temp_fd, 0, SEEK_CUR) == 0 and tell($temp_fd) == 0
1054 or throw Error::Simple("expected file position to be reset");
1055 }
1056
1057 =item temp_path ( NAME )
1058
1059 =item temp_path ( FILEHANDLE )
1060
1061 Returns the filename associated with the given tempfile.
1062
1063 =cut
1064
1065 sub temp_path {
1066 my ($self, $temp_fd) = _maybe_self(@_);
1067
1068 if (exists $TEMP_FILEMAP{$temp_fd}) {
1069 $temp_fd = $TEMP_FILEMAP{$temp_fd};
1070 }
1071 $TEMP_FILES{$temp_fd}{fname};
1072 }
1073
1074 sub END {
1075 unlink values %TEMP_FILEMAP if %TEMP_FILEMAP;
1076 }
1077
1078 } # %TEMP_* Lexical Context
1079
1080 =back
1081
1082 =head1 ERROR HANDLING
1083
1084 All functions are supposed to throw Perl exceptions in case of errors.
1085 See the L<Error> module on how to catch those. Most exceptions are mere
1086 L<Error::Simple> instances.
1087
1088 However, the C<command()>, C<command_oneline()> and C<command_noisy()>
1089 functions suite can throw C<Git::Error::Command> exceptions as well: those are
1090 thrown when the external command returns an error code and contain the error
1091 code as well as access to the captured command's output. The exception class
1092 provides the usual C<stringify> and C<value> (command's exit code) methods and
1093 in addition also a C<cmd_output> method that returns either an array or a
1094 string with the captured command output (depending on the original function
1095 call context; C<command_noisy()> returns C<undef>) and $<cmdline> which
1096 returns the command and its arguments (but without proper quoting).
1097
1098 Note that the C<command_*_pipe()> functions cannot throw this exception since
1099 it has no idea whether the command failed or not. You will only find out
1100 at the time you C<close> the pipe; if you want to have that automated,
1101 use C<command_close_pipe()>, which can throw the exception.
1102
1103 =cut
1104
1105 {
1106 package Git::Error::Command;
1107
1108 @Git::Error::Command::ISA = qw(Error);
1109
1110 sub new {
1111 my $self = shift;
1112 my $cmdline = '' . shift;
1113 my $value = 0 + shift;
1114 my $outputref = shift;
1115 my(@args) = ();
1116
1117 local $Error::Depth = $Error::Depth + 1;
1118
1119 push(@args, '-cmdline', $cmdline);
1120 push(@args, '-value', $value);
1121 push(@args, '-outputref', $outputref);
1122
1123 $self->SUPER::new(-text => 'command returned error', @args);
1124 }
1125
1126 sub stringify {
1127 my $self = shift;
1128 my $text = $self->SUPER::stringify;
1129 $self->cmdline() . ': ' . $text . ': ' . $self->value() . "\n";
1130 }
1131
1132 sub cmdline {
1133 my $self = shift;
1134 $self->{'-cmdline'};
1135 }
1136
1137 sub cmd_output {
1138 my $self = shift;
1139 my $ref = $self->{'-outputref'};
1140 defined $ref or undef;
1141 if (ref $ref eq 'ARRAY') {
1142 return @$ref;
1143 } else { # SCALAR
1144 return $$ref;
1145 }
1146 }
1147 }
1148
1149 =over 4
1150
1151 =item git_cmd_try { CODE } ERRMSG
1152
1153 This magical statement will automatically catch any C<Git::Error::Command>
1154 exceptions thrown by C<CODE> and make your program die with C<ERRMSG>
1155 on its lips; the message will have %s substituted for the command line
1156 and %d for the exit status. This statement is useful mostly for producing
1157 more user-friendly error messages.
1158
1159 In case of no exception caught the statement returns C<CODE>'s return value.
1160
1161 Note that this is the only auto-exported function.
1162
1163 =cut
1164
1165 sub git_cmd_try(&$) {
1166 my ($code, $errmsg) = @_;
1167 my @result;
1168 my $err;
1169 my $array = wantarray;
1170 try {
1171 if ($array) {
1172 @result = &$code;
1173 } else {
1174 $result[0] = &$code;
1175 }
1176 } catch Git::Error::Command with {
1177 my $E = shift;
1178 $err = $errmsg;
1179 $err =~ s/\%s/$E->cmdline()/ge;
1180 $err =~ s/\%d/$E->value()/ge;
1181 # We can't croak here since Error.pm would mangle
1182 # that to Error::Simple.
1183 };
1184 $err and croak $err;
1185 return $array ? @result : $result[0];
1186 }
1187
1188
1189 =back
1190
1191 =head1 COPYRIGHT
1192
1193 Copyright 2006 by Petr Baudis E<lt>pasky@suse.czE<gt>.
1194
1195 This module is free software; it may be used, copied, modified
1196 and distributed under the terms of the GNU General Public Licence,
1197 either version 2, or (at your option) any later version.
1198
1199 =cut
1200
1201
1202 # Take raw method argument list and return ($obj, @args) in case
1203 # the method was called upon an instance and (undef, @args) if
1204 # it was called directly.
1205 sub _maybe_self {
1206 UNIVERSAL::isa($_[0], 'Git') ? @_ : (undef, @_);
1207 }
1208
1209 # Check if the command id is something reasonable.
1210 sub _check_valid_cmd {
1211 my ($cmd) = @_;
1212 $cmd =~ /^[a-z0-9A-Z_-]+$/ or throw Error::Simple("bad command: $cmd");
1213 }
1214
1215 # Common backend for the pipe creators.
1216 sub _command_common_pipe {
1217 my $direction = shift;
1218 my ($self, @p) = _maybe_self(@_);
1219 my (%opts, $cmd, @args);
1220 if (ref $p[0]) {
1221 ($cmd, @args) = @{shift @p};
1222 %opts = ref $p[0] ? %{$p[0]} : @p;
1223 } else {
1224 ($cmd, @args) = @p;
1225 }
1226 _check_valid_cmd($cmd);
1227
1228 my $fh;
1229 if ($^O eq 'MSWin32') {
1230 # ActiveState Perl
1231 #defined $opts{STDERR} and
1232 # warn 'ignoring STDERR option - running w/ ActiveState';
1233 $direction eq '-|' or
1234 die 'input pipe for ActiveState not implemented';
1235 # the strange construction with *ACPIPE is just to
1236 # explain the tie below that we want to bind to
1237 # a handle class, not scalar. It is not known if
1238 # it is something specific to ActiveState Perl or
1239 # just a Perl quirk.
1240 tie (*ACPIPE, 'Git::activestate_pipe', $cmd, @args);
1241 $fh = *ACPIPE;
1242
1243 } else {
1244 my $pid = open($fh, $direction);
1245 if (not defined $pid) {
1246 throw Error::Simple("open failed: $!");
1247 } elsif ($pid == 0) {
1248 if (defined $opts{STDERR}) {
1249 close STDERR;
1250 }
1251 if ($opts{STDERR}) {
1252 open (STDERR, '>&', $opts{STDERR})
1253 or die "dup failed: $!";
1254 }
1255 _cmd_exec($self, $cmd, @args);
1256 }
1257 }
1258 return wantarray ? ($fh, join(' ', $cmd, @args)) : $fh;
1259 }
1260
1261 # When already in the subprocess, set up the appropriate state
1262 # for the given repository and execute the git command.
1263 sub _cmd_exec {
1264 my ($self, @args) = @_;
1265 if ($self) {
1266 $self->repo_path() and $ENV{'GIT_DIR'} = $self->repo_path();
1267 $self->wc_path() and chdir($self->wc_path());
1268 $self->wc_subdir() and chdir($self->wc_subdir());
1269 }
1270 _execv_git_cmd(@args);
1271 die qq[exec "@args" failed: $!];
1272 }
1273
1274 # Execute the given Git command ($_[0]) with arguments ($_[1..])
1275 # by searching for it at proper places.
1276 sub _execv_git_cmd { exec('git', @_); }
1277
1278 # Close pipe to a subprocess.
1279 sub _cmd_close {
1280 my ($fh, $ctx) = @_;
1281 if (not close $fh) {
1282 if ($!) {
1283 # It's just close, no point in fatalities
1284 carp "error closing pipe: $!";
1285 } elsif ($? >> 8) {
1286 # The caller should pepper this.
1287 throw Git::Error::Command($ctx, $? >> 8);
1288 }
1289 # else we might e.g. closed a live stream; the command
1290 # dying of SIGPIPE would drive us here.
1291 }
1292 }
1293
1294
1295 sub DESTROY {
1296 my ($self) = @_;
1297 $self->_close_hash_and_insert_object();
1298 $self->_close_cat_blob();
1299 }
1300
1301
1302 # Pipe implementation for ActiveState Perl.
1303
1304 package Git::activestate_pipe;
1305 use strict;
1306
1307 sub TIEHANDLE {
1308 my ($class, @params) = @_;
1309 # FIXME: This is probably horrible idea and the thing will explode
1310 # at the moment you give it arguments that require some quoting,
1311 # but I have no ActiveState clue... --pasky
1312 # Let's just hope ActiveState Perl does at least the quoting
1313 # correctly.
1314 my @data = qx{git @params};
1315 bless { i => 0, data => \@data }, $class;
1316 }
1317
1318 sub READLINE {
1319 my $self = shift;
1320 if ($self->{i} >= scalar @{$self->{data}}) {
1321 return undef;
1322 }
1323 my $i = $self->{i};
1324 if (wantarray) {
1325 $self->{i} = $#{$self->{'data'}} + 1;
1326 return splice(@{$self->{'data'}}, $i);
1327 }
1328 $self->{i} = $i + 1;
1329 return $self->{'data'}->[ $i ];
1330 }
1331
1332 sub CLOSE {
1333 my $self = shift;
1334 delete $self->{data};
1335 delete $self->{i};
1336 }
1337
1338 sub EOF {
1339 my $self = shift;
1340 return ($self->{i} >= scalar @{$self->{data}});
1341 }
1342
1343
1344 1; # Famous last words