perl / Git.pmon commit Merge branch 'jk/maint-stash-oob' into maint (be44a20)
   1=head1 NAME
   2
   3Git - Perl interface to the Git version control system
   4
   5=cut
   6
   7
   8package Git;
   9
  10use 5.008;
  11use strict;
  12
  13
  14BEGIN {
  15
  16our ($VERSION, @ISA, @EXPORT, @EXPORT_OK);
  17
  18# Totally unstable API.
  19$VERSION = '0.01';
  20
  21
  22=head1 SYNOPSIS
  23
  24  use Git;
  25
  26  my $version = Git::command_oneline('version');
  27
  28  git_cmd_try { Git::command_noisy('update-server-info') }
  29              '%s failed w/ code %d';
  30
  31  my $repo = Git->repository (Directory => '/srv/git/cogito.git');
  32
  33
  34  my @revs = $repo->command('rev-list', '--since=last monday', '--all');
  35
  36  my ($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);
  39
  40  my $lastrev = $repo->command_oneline( [ 'rev-list', '--all' ],
  41                                        STDERR => 0 );
  42
  43  my $sha1 = $repo->hash_and_insert_object('file.txt');
  44  my $tempfile = tempfile();
  45  my $size = $repo->cat_blob($sha1, $tempfile);
  46
  47=cut
  48
  49
  50require Exporter;
  51
  52@ISA = qw(Exporter);
  53
  54@EXPORT = qw(git_cmd_try);
  55
  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
  61                remote_refs
  62                temp_acquire temp_release temp_reset temp_path);
  63
  64
  65=head1 DESCRIPTION
  66
  67This module provides Perl scripts easy way to interface the Git version control
  68system. The modules have an easy and well-tested way to call arbitrary Git
  69commands; in the future, the interface will also provide specialized methods
  70for doing easily operations which are not totally trivial to do over
  71the generic command interface.
  72
  73While some commands can be executed outside of any context (e.g. 'version'
  74or 'init'), most operations require a repository context, which in practice
  75means getting an instance of the Git object using the repository() constructor.
  76(In the future, we will also get a new_repository() constructor.) All commands
  77called as methods of the object are then executed in the context of the
  78repository.
  79
  80Part of the "repository state" is also information about path to the attached
  81working copy (unless you work with a bare repository). You can also navigate
  82inside of the working copy using the C<wc_chdir()> method. (Note that
  83the repository object is self-contained and will not change working directory
  84of your process.)
  85
  86TODO: In the future, we might also do
  87
  88        my $remoterepo = $repo->remote_repository (Name => 'cogito', Branch => 'master');
  89        $remoterepo ||= Git->remote_repository ('http://git.or.cz/cogito.git/');
  90        my @refs = $remoterepo->refs();
  91
  92Currently, the module merely wraps calls to external Git tools. In the future,
  93it will provide a much faster way to interact with Git by linking directly
  94to libgit. This should be completely opaque to the user, though (performance
  95increase notwithstanding).
  96
  97=cut
  98
  99
 100use Carp qw(carp croak); # but croak is bad - throw instead
 101use Error qw(:try);
 102use Cwd qw(abs_path cwd);
 103use IPC::Open2 qw(open2);
 104use Fcntl qw(SEEK_SET SEEK_CUR);
 105}
 106
 107
 108=head1 CONSTRUCTORS
 109
 110=over 4
 111
 112=item repository ( OPTIONS )
 113
 114=item repository ( DIRECTORY )
 115
 116=item repository ()
 117
 118Construct a new repository object.
 119C<OPTIONS> are passed in a hash like fashion, using key and value pairs.
 120Possible options are:
 121
 122B<Repository> - Path to the Git repository.
 123
 124B<WorkingCopy> - Path to the associated working copy; not strictly required
 125as many commands will happily crunch on a bare repository.
 126
 127B<WorkingSubdir> - Subdirectory in the working copy to work inside.
 128Just left undefined if you do not want to limit the scope of operations.
 129
 130B<Directory> - Path to the Git working directory in its usual setup.
 131The C<.git> directory is searched in the directory and all the parent
 132directories; if found, C<WorkingCopy> is set to the directory containing
 133it and C<Repository> to the C<.git> directory itself. If no C<.git>
 134directory was found, the C<Directory> is assumed to be a bare repository,
 135C<Repository> is set to point at it and C<WorkingCopy> is left undefined.
 136If the C<$GIT_DIR> environment variable is set, things behave as expected
 137as well.
 138
 139You should not use both C<Directory> and either of C<Repository> and
 140C<WorkingCopy> - the results of that are undefined.
 141
 142Alternatively, a directory path may be passed as a single scalar argument
 143to the constructor; it is equivalent to setting only the C<Directory> option
 144field.
 145
 146Calling the constructor with no options whatsoever is equivalent to
 147calling it with C<< Directory => '.' >>. In general, if you are building
 148a standard porcelain command, simply doing C<< Git->repository() >> should
 149do the right thing and setup the object to reflect exactly where the user
 150is right now.
 151
 152=cut
 153
 154sub repository {
 155        my $class = shift;
 156        my @args = @_;
 157        my %opts = ();
 158        my $self;
 159
 160        if (defined $args[0]) {
 161                if ($#args % 2 != 1) {
 162                        # Not a hash.
 163                        $#args == 0 or throw Error::Simple("bad usage");
 164                        %opts = ( Directory => $args[0] );
 165                } else {
 166                        %opts = @args;
 167                }
 168        }
 169
 170        if (not defined $opts{Repository} and not defined $opts{WorkingCopy}
 171                and not defined $opts{Directory}) {
 172                $opts{Directory} = '.';
 173        }
 174
 175        if (defined $opts{Directory}) {
 176                -d $opts{Directory} or throw Error::Simple("Directory not found: $opts{Directory} $!");
 177
 178                my $search = Git->repository(WorkingCopy => $opts{Directory});
 179                my $dir;
 180                try {
 181                        $dir = $search->command_oneline(['rev-parse', '--git-dir'],
 182                                                        STDERR => 0);
 183                } catch Git::Error::Command with {
 184                        $dir = undef;
 185                };
 186
 187                if ($dir) {
 188                        $dir =~ m#^/# or $dir = $opts{Directory} . '/' . $dir;
 189                        $opts{Repository} = abs_path($dir);
 190
 191                        # If --git-dir went ok, this shouldn't die either.
 192                        my $prefix = $search->command_oneline('rev-parse', '--show-prefix');
 193                        $dir = abs_path($opts{Directory}) . '/';
 194                        if ($prefix) {
 195                                if (substr($dir, -length($prefix)) ne $prefix) {
 196                                        throw Error::Simple("rev-parse confused me - $dir does not have trailing $prefix");
 197                                }
 198                                substr($dir, -length($prefix)) = '';
 199                        }
 200                        $opts{WorkingCopy} = $dir;
 201                        $opts{WorkingSubdir} = $prefix;
 202
 203                } else {
 204                        # A bare repository? Let's see...
 205                        $dir = $opts{Directory};
 206
 207                        unless (-d "$dir/refs" and -d "$dir/objects" and -e "$dir/HEAD") {
 208                                # Mimic git-rev-parse --git-dir error message:
 209                                throw Error::Simple("fatal: Not a git repository: $dir");
 210                        }
 211                        my $search = Git->repository(Repository => $dir);
 212                        try {
 213                                $search->command('symbolic-ref', 'HEAD');
 214                        } catch Git::Error::Command with {
 215                                # Mimic git-rev-parse --git-dir error message:
 216                                throw Error::Simple("fatal: Not a git repository: $dir");
 217                        }
 218
 219                        $opts{Repository} = abs_path($dir);
 220                }
 221
 222                delete $opts{Directory};
 223        }
 224
 225        $self = { opts => \%opts };
 226        bless $self, $class;
 227}
 228
 229=back
 230
 231=head1 METHODS
 232
 233=over 4
 234
 235=item command ( COMMAND [, ARGUMENTS... ] )
 236
 237=item command ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 238
 239Execute the given Git C<COMMAND> (specify it without the 'git-'
 240prefix), optionally with the specified extra C<ARGUMENTS>.
 241
 242The second more elaborate form can be used if you want to further adjust
 243the command execution. Currently, only one option is supported:
 244
 245B<STDERR> - How to deal with the command's error output. By default (C<undef>)
 246it is delivered to the caller's C<STDERR>. A false value (0 or '') will cause
 247it to be thrown away. If you want to process it, you can get it in a filehandle
 248you specify, but you must be extremely careful; if the error output is not
 249very short and you want to read it in the same process as where you called
 250C<command()>, you are set up for a nice deadlock!
 251
 252The method can be called without any instance or on a specified Git repository
 253(in that case the command will be run in the repository context).
 254
 255In scalar context, it returns all the command output in a single string
 256(verbatim).
 257
 258In array context, it returns an array containing lines printed to the
 259command's stdout (without trailing newlines).
 260
 261In both cases, the command's stdin and stderr are the same as the caller's.
 262
 263=cut
 264
 265sub command {
 266        my ($fh, $ctx) = command_output_pipe(@_);
 267
 268        if (not defined wantarray) {
 269                # Nothing to pepper the possible exception with.
 270                _cmd_close($fh, $ctx);
 271
 272        } elsif (not wantarray) {
 273                local $/;
 274                my $text = <$fh>;
 275                try {
 276                        _cmd_close($fh, $ctx);
 277                } catch Git::Error::Command with {
 278                        # Pepper with the output:
 279                        my $E = shift;
 280                        $E->{'-outputref'} = \$text;
 281                        throw $E;
 282                };
 283                return $text;
 284
 285        } else {
 286                my @lines = <$fh>;
 287                defined and chomp for @lines;
 288                try {
 289                        _cmd_close($fh, $ctx);
 290                } catch Git::Error::Command with {
 291                        my $E = shift;
 292                        $E->{'-outputref'} = \@lines;
 293                        throw $E;
 294                };
 295                return @lines;
 296        }
 297}
 298
 299
 300=item command_oneline ( COMMAND [, ARGUMENTS... ] )
 301
 302=item command_oneline ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 303
 304Execute the given C<COMMAND> in the same way as command()
 305does but always return a scalar string containing the first line
 306of the command's standard output.
 307
 308=cut
 309
 310sub command_oneline {
 311        my ($fh, $ctx) = command_output_pipe(@_);
 312
 313        my $line = <$fh>;
 314        defined $line and chomp $line;
 315        try {
 316                _cmd_close($fh, $ctx);
 317        } catch Git::Error::Command with {
 318                # Pepper with the output:
 319                my $E = shift;
 320                $E->{'-outputref'} = \$line;
 321                throw $E;
 322        };
 323        return $line;
 324}
 325
 326
 327=item command_output_pipe ( COMMAND [, ARGUMENTS... ] )
 328
 329=item command_output_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 330
 331Execute the given C<COMMAND> in the same way as command()
 332does but return a pipe filehandle from which the command output can be
 333read.
 334
 335The function can return C<($pipe, $ctx)> in array context.
 336See C<command_close_pipe()> for details.
 337
 338=cut
 339
 340sub command_output_pipe {
 341        _command_common_pipe('-|', @_);
 342}
 343
 344
 345=item command_input_pipe ( COMMAND [, ARGUMENTS... ] )
 346
 347=item command_input_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 348
 349Execute the given C<COMMAND> in the same way as command_output_pipe()
 350does but return an input pipe filehandle instead; the command output
 351is not captured.
 352
 353The function can return C<($pipe, $ctx)> in array context.
 354See C<command_close_pipe()> for details.
 355
 356=cut
 357
 358sub command_input_pipe {
 359        _command_common_pipe('|-', @_);
 360}
 361
 362
 363=item command_close_pipe ( PIPE [, CTX ] )
 364
 365Close the C<PIPE> as returned from C<command_*_pipe()>, checking
 366whether the command finished successfully. The optional C<CTX> argument
 367is required if you want to see the command name in the error message,
 368and it is the second value returned by C<command_*_pipe()> when
 369called in array context. The call idiom is:
 370
 371        my ($fh, $ctx) = $r->command_output_pipe('status');
 372        while (<$fh>) { ... }
 373        $r->command_close_pipe($fh, $ctx);
 374
 375Note that you should not rely on whatever actually is in C<CTX>;
 376currently it is simply the command name but in future the context might
 377have more complicated structure.
 378
 379=cut
 380
 381sub command_close_pipe {
 382        my ($self, $fh, $ctx) = _maybe_self(@_);
 383        $ctx ||= '<unknown>';
 384        _cmd_close($fh, $ctx);
 385}
 386
 387=item command_bidi_pipe ( COMMAND [, ARGUMENTS... ] )
 388
 389Execute the given C<COMMAND> in the same way as command_output_pipe()
 390does but return both an input pipe filehandle and an output pipe filehandle.
 391
 392The function will return return C<($pid, $pipe_in, $pipe_out, $ctx)>.
 393See C<command_close_bidi_pipe()> for details.
 394
 395=cut
 396
 397sub command_bidi_pipe {
 398        my ($pid, $in, $out);
 399        my ($self) = _maybe_self(@_);
 400        local %ENV = %ENV;
 401        my $cwd_save = undef;
 402        if ($self) {
 403                shift;
 404                $cwd_save = cwd();
 405                _setup_git_cmd_env($self);
 406        }
 407        $pid = open2($in, $out, 'git', @_);
 408        chdir($cwd_save) if $cwd_save;
 409        return ($pid, $in, $out, join(' ', @_));
 410}
 411
 412=item command_close_bidi_pipe ( PID, PIPE_IN, PIPE_OUT [, CTX] )
 413
 414Close the C<PIPE_IN> and C<PIPE_OUT> as returned from C<command_bidi_pipe()>,
 415checking whether the command finished successfully. The optional C<CTX>
 416argument is required if you want to see the command name in the error message,
 417and it is the fourth value returned by C<command_bidi_pipe()>.  The call idiom
 418is:
 419
 420        my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
 421        print "000000000\n" $out;
 422        while (<$in>) { ... }
 423        $r->command_close_bidi_pipe($pid, $in, $out, $ctx);
 424
 425Note that you should not rely on whatever actually is in C<CTX>;
 426currently it is simply the command name but in future the context might
 427have more complicated structure.
 428
 429=cut
 430
 431sub command_close_bidi_pipe {
 432        local $?;
 433        my ($pid, $in, $out, $ctx) = @_;
 434        foreach my $fh ($in, $out) {
 435                unless (close $fh) {
 436                        if ($!) {
 437                                carp "error closing pipe: $!";
 438                        } elsif ($? >> 8) {
 439                                throw Git::Error::Command($ctx, $? >>8);
 440                        }
 441                }
 442        }
 443
 444        waitpid $pid, 0;
 445
 446        if ($? >> 8) {
 447                throw Git::Error::Command($ctx, $? >>8);
 448        }
 449}
 450
 451
 452=item command_noisy ( COMMAND [, ARGUMENTS... ] )
 453
 454Execute the given C<COMMAND> in the same way as command() does but do not
 455capture the command output - the standard output is not redirected and goes
 456to the standard output of the caller application.
 457
 458While the method is called command_noisy(), you might want to as well use
 459it for the most silent Git commands which you know will never pollute your
 460stdout but you want to avoid the overhead of the pipe setup when calling them.
 461
 462The function returns only after the command has finished running.
 463
 464=cut
 465
 466sub command_noisy {
 467        my ($self, $cmd, @args) = _maybe_self(@_);
 468        _check_valid_cmd($cmd);
 469
 470        my $pid = fork;
 471        if (not defined $pid) {
 472                throw Error::Simple("fork failed: $!");
 473        } elsif ($pid == 0) {
 474                _cmd_exec($self, $cmd, @args);
 475        }
 476        if (waitpid($pid, 0) > 0 and $?>>8 != 0) {
 477                throw Git::Error::Command(join(' ', $cmd, @args), $? >> 8);
 478        }
 479}
 480
 481
 482=item version ()
 483
 484Return the Git version in use.
 485
 486=cut
 487
 488sub version {
 489        my $verstr = command_oneline('--version');
 490        $verstr =~ s/^git version //;
 491        $verstr;
 492}
 493
 494
 495=item exec_path ()
 496
 497Return path to the Git sub-command executables (the same as
 498C<git --exec-path>). Useful mostly only internally.
 499
 500=cut
 501
 502sub exec_path { command_oneline('--exec-path') }
 503
 504
 505=item html_path ()
 506
 507Return path to the Git html documentation (the same as
 508C<git --html-path>). Useful mostly only internally.
 509
 510=cut
 511
 512sub html_path { command_oneline('--html-path') }
 513
 514
 515=item repo_path ()
 516
 517Return path to the git repository. Must be called on a repository instance.
 518
 519=cut
 520
 521sub repo_path { $_[0]->{opts}->{Repository} }
 522
 523
 524=item wc_path ()
 525
 526Return path to the working copy. Must be called on a repository instance.
 527
 528=cut
 529
 530sub wc_path { $_[0]->{opts}->{WorkingCopy} }
 531
 532
 533=item wc_subdir ()
 534
 535Return path to the subdirectory inside of a working copy. Must be called
 536on a repository instance.
 537
 538=cut
 539
 540sub wc_subdir { $_[0]->{opts}->{WorkingSubdir} ||= '' }
 541
 542
 543=item wc_chdir ( SUBDIR )
 544
 545Change the working copy subdirectory to work within. The C<SUBDIR> is
 546relative to the working copy root directory (not the current subdirectory).
 547Must be called on a repository instance attached to a working copy
 548and the directory must exist.
 549
 550=cut
 551
 552sub wc_chdir {
 553        my ($self, $subdir) = @_;
 554        $self->wc_path()
 555                or throw Error::Simple("bare repository");
 556
 557        -d $self->wc_path().'/'.$subdir
 558                or throw Error::Simple("subdir not found: $subdir $!");
 559        # Of course we will not "hold" the subdirectory so anyone
 560        # can delete it now and we will never know. But at least we tried.
 561
 562        $self->{opts}->{WorkingSubdir} = $subdir;
 563}
 564
 565
 566=item config ( VARIABLE )
 567
 568Retrieve the configuration C<VARIABLE> in the same manner as C<config>
 569does. In scalar context requires the variable to be set only one time
 570(exception is thrown otherwise), in array context returns allows the
 571variable to be set multiple times and returns all the values.
 572
 573This currently wraps command('config') so it is not so fast.
 574
 575=cut
 576
 577sub config {
 578        my ($self, $var) = _maybe_self(@_);
 579
 580        try {
 581                my @cmd = ('config');
 582                unshift @cmd, $self if $self;
 583                if (wantarray) {
 584                        return command(@cmd, '--get-all', $var);
 585                } else {
 586                        return command_oneline(@cmd, '--get', $var);
 587                }
 588        } catch Git::Error::Command with {
 589                my $E = shift;
 590                if ($E->value() == 1) {
 591                        # Key not found.
 592                        return;
 593                } else {
 594                        throw $E;
 595                }
 596        };
 597}
 598
 599
 600=item config_bool ( VARIABLE )
 601
 602Retrieve the bool configuration C<VARIABLE>. The return value
 603is usable as a boolean in perl (and C<undef> if it's not defined,
 604of course).
 605
 606This currently wraps command('config') so it is not so fast.
 607
 608=cut
 609
 610sub config_bool {
 611        my ($self, $var) = _maybe_self(@_);
 612
 613        try {
 614                my @cmd = ('config', '--bool', '--get', $var);
 615                unshift @cmd, $self if $self;
 616                my $val = command_oneline(@cmd);
 617                return undef unless defined $val;
 618                return $val eq 'true';
 619        } catch Git::Error::Command with {
 620                my $E = shift;
 621                if ($E->value() == 1) {
 622                        # Key not found.
 623                        return undef;
 624                } else {
 625                        throw $E;
 626                }
 627        };
 628}
 629
 630=item config_int ( VARIABLE )
 631
 632Retrieve the integer configuration C<VARIABLE>. The return value
 633is simple decimal number.  An optional value suffix of 'k', 'm',
 634or 'g' in the config file will cause the value to be multiplied
 635by 1024, 1048576 (1024^2), or 1073741824 (1024^3) prior to output.
 636It would return C<undef> if configuration variable is not defined,
 637
 638This currently wraps command('config') so it is not so fast.
 639
 640=cut
 641
 642sub config_int {
 643        my ($self, $var) = _maybe_self(@_);
 644
 645        try {
 646                my @cmd = ('config', '--int', '--get', $var);
 647                unshift @cmd, $self if $self;
 648                return command_oneline(@cmd);
 649        } catch Git::Error::Command with {
 650                my $E = shift;
 651                if ($E->value() == 1) {
 652                        # Key not found.
 653                        return undef;
 654                } else {
 655                        throw $E;
 656                }
 657        };
 658}
 659
 660=item get_colorbool ( NAME )
 661
 662Finds if color should be used for NAMEd operation from the configuration,
 663and returns boolean (true for "use color", false for "do not use color").
 664
 665=cut
 666
 667sub get_colorbool {
 668        my ($self, $var) = @_;
 669        my $stdout_to_tty = (-t STDOUT) ? "true" : "false";
 670        my $use_color = $self->command_oneline('config', '--get-colorbool',
 671                                               $var, $stdout_to_tty);
 672        return ($use_color eq 'true');
 673}
 674
 675=item get_color ( SLOT, COLOR )
 676
 677Finds color for SLOT from the configuration, while defaulting to COLOR,
 678and returns the ANSI color escape sequence:
 679
 680        print $repo->get_color("color.interactive.prompt", "underline blue white");
 681        print "some text";
 682        print $repo->get_color("", "normal");
 683
 684=cut
 685
 686sub get_color {
 687        my ($self, $slot, $default) = @_;
 688        my $color = $self->command_oneline('config', '--get-color', $slot, $default);
 689        if (!defined $color) {
 690                $color = "";
 691        }
 692        return $color;
 693}
 694
 695=item remote_refs ( REPOSITORY [, GROUPS [, REFGLOBS ] ] )
 696
 697This function returns a hashref of refs stored in a given remote repository.
 698The hash is in the format C<refname =\> hash>. For tags, the C<refname> entry
 699contains the tag object while a C<refname^{}> entry gives the tagged objects.
 700
 701C<REPOSITORY> has the same meaning as the appropriate C<git-ls-remote>
 702argument; either an URL or a remote name (if called on a repository instance).
 703C<GROUPS> is an optional arrayref that can contain 'tags' to return all the
 704tags and/or 'heads' to return all the heads. C<REFGLOB> is an optional array
 705of strings containing a shell-like glob to further limit the refs returned in
 706the hash; the meaning is again the same as the appropriate C<git-ls-remote>
 707argument.
 708
 709This function may or may not be called on a repository instance. In the former
 710case, remote names as defined in the repository are recognized as repository
 711specifiers.
 712
 713=cut
 714
 715sub remote_refs {
 716        my ($self, $repo, $groups, $refglobs) = _maybe_self(@_);
 717        my @args;
 718        if (ref $groups eq 'ARRAY') {
 719                foreach (@$groups) {
 720                        if ($_ eq 'heads') {
 721                                push (@args, '--heads');
 722                        } elsif ($_ eq 'tags') {
 723                                push (@args, '--tags');
 724                        } else {
 725                                # Ignore unknown groups for future
 726                                # compatibility
 727                        }
 728                }
 729        }
 730        push (@args, $repo);
 731        if (ref $refglobs eq 'ARRAY') {
 732                push (@args, @$refglobs);
 733        }
 734
 735        my @self = $self ? ($self) : (); # Ultra trickery
 736        my ($fh, $ctx) = Git::command_output_pipe(@self, 'ls-remote', @args);
 737        my %refs;
 738        while (<$fh>) {
 739                chomp;
 740                my ($hash, $ref) = split(/\t/, $_, 2);
 741                $refs{$ref} = $hash;
 742        }
 743        Git::command_close_pipe(@self, $fh, $ctx);
 744        return \%refs;
 745}
 746
 747
 748=item ident ( TYPE | IDENTSTR )
 749
 750=item ident_person ( TYPE | IDENTSTR | IDENTARRAY )
 751
 752This suite of functions retrieves and parses ident information, as stored
 753in the commit and tag objects or produced by C<var GIT_type_IDENT> (thus
 754C<TYPE> can be either I<author> or I<committer>; case is insignificant).
 755
 756The C<ident> method retrieves the ident information from C<git var>
 757and either returns it as a scalar string or as an array with the fields parsed.
 758Alternatively, it can take a prepared ident string (e.g. from the commit
 759object) and just parse it.
 760
 761C<ident_person> returns the person part of the ident - name and email;
 762it can take the same arguments as C<ident> or the array returned by C<ident>.
 763
 764The synopsis is like:
 765
 766        my ($name, $email, $time_tz) = ident('author');
 767        "$name <$email>" eq ident_person('author');
 768        "$name <$email>" eq ident_person($name);
 769        $time_tz =~ /^\d+ [+-]\d{4}$/;
 770
 771=cut
 772
 773sub ident {
 774        my ($self, $type) = _maybe_self(@_);
 775        my $identstr;
 776        if (lc $type eq lc 'committer' or lc $type eq lc 'author') {
 777                my @cmd = ('var', 'GIT_'.uc($type).'_IDENT');
 778                unshift @cmd, $self if $self;
 779                $identstr = command_oneline(@cmd);
 780        } else {
 781                $identstr = $type;
 782        }
 783        if (wantarray) {
 784                return $identstr =~ /^(.*) <(.*)> (\d+ [+-]\d{4})$/;
 785        } else {
 786                return $identstr;
 787        }
 788}
 789
 790sub ident_person {
 791        my ($self, @ident) = _maybe_self(@_);
 792        $#ident == 0 and @ident = $self ? $self->ident($ident[0]) : ident($ident[0]);
 793        return "$ident[0] <$ident[1]>";
 794}
 795
 796
 797=item hash_object ( TYPE, FILENAME )
 798
 799Compute the SHA1 object id of the given C<FILENAME> considering it is
 800of the C<TYPE> object type (C<blob>, C<commit>, C<tree>).
 801
 802The method can be called without any instance or on a specified Git repository,
 803it makes zero difference.
 804
 805The function returns the SHA1 hash.
 806
 807=cut
 808
 809# TODO: Support for passing FILEHANDLE instead of FILENAME
 810sub hash_object {
 811        my ($self, $type, $file) = _maybe_self(@_);
 812        command_oneline('hash-object', '-t', $type, $file);
 813}
 814
 815
 816=item hash_and_insert_object ( FILENAME )
 817
 818Compute the SHA1 object id of the given C<FILENAME> and add the object to the
 819object database.
 820
 821The function returns the SHA1 hash.
 822
 823=cut
 824
 825# TODO: Support for passing FILEHANDLE instead of FILENAME
 826sub hash_and_insert_object {
 827        my ($self, $filename) = @_;
 828
 829        carp "Bad filename \"$filename\"" if $filename =~ /[\r\n]/;
 830
 831        $self->_open_hash_and_insert_object_if_needed();
 832        my ($in, $out) = ($self->{hash_object_in}, $self->{hash_object_out});
 833
 834        unless (print $out $filename, "\n") {
 835                $self->_close_hash_and_insert_object();
 836                throw Error::Simple("out pipe went bad");
 837        }
 838
 839        chomp(my $hash = <$in>);
 840        unless (defined($hash)) {
 841                $self->_close_hash_and_insert_object();
 842                throw Error::Simple("in pipe went bad");
 843        }
 844
 845        return $hash;
 846}
 847
 848sub _open_hash_and_insert_object_if_needed {
 849        my ($self) = @_;
 850
 851        return if defined($self->{hash_object_pid});
 852
 853        ($self->{hash_object_pid}, $self->{hash_object_in},
 854         $self->{hash_object_out}, $self->{hash_object_ctx}) =
 855                $self->command_bidi_pipe(qw(hash-object -w --stdin-paths --no-filters));
 856}
 857
 858sub _close_hash_and_insert_object {
 859        my ($self) = @_;
 860
 861        return unless defined($self->{hash_object_pid});
 862
 863        my @vars = map { 'hash_object_' . $_ } qw(pid in out ctx);
 864
 865        command_close_bidi_pipe(@$self{@vars});
 866        delete @$self{@vars};
 867}
 868
 869=item cat_blob ( SHA1, FILEHANDLE )
 870
 871Prints the contents of the blob identified by C<SHA1> to C<FILEHANDLE> and
 872returns the number of bytes printed.
 873
 874=cut
 875
 876sub cat_blob {
 877        my ($self, $sha1, $fh) = @_;
 878
 879        $self->_open_cat_blob_if_needed();
 880        my ($in, $out) = ($self->{cat_blob_in}, $self->{cat_blob_out});
 881
 882        unless (print $out $sha1, "\n") {
 883                $self->_close_cat_blob();
 884                throw Error::Simple("out pipe went bad");
 885        }
 886
 887        my $description = <$in>;
 888        if ($description =~ / missing$/) {
 889                carp "$sha1 doesn't exist in the repository";
 890                return -1;
 891        }
 892
 893        if ($description !~ /^[0-9a-fA-F]{40} \S+ (\d+)$/) {
 894                carp "Unexpected result returned from git cat-file";
 895                return -1;
 896        }
 897
 898        my $size = $1;
 899
 900        my $blob;
 901        my $bytesRead = 0;
 902
 903        while (1) {
 904                my $bytesLeft = $size - $bytesRead;
 905                last unless $bytesLeft;
 906
 907                my $bytesToRead = $bytesLeft < 1024 ? $bytesLeft : 1024;
 908                my $read = read($in, $blob, $bytesToRead, $bytesRead);
 909                unless (defined($read)) {
 910                        $self->_close_cat_blob();
 911                        throw Error::Simple("in pipe went bad");
 912                }
 913
 914                $bytesRead += $read;
 915        }
 916
 917        # Skip past the trailing newline.
 918        my $newline;
 919        my $read = read($in, $newline, 1);
 920        unless (defined($read)) {
 921                $self->_close_cat_blob();
 922                throw Error::Simple("in pipe went bad");
 923        }
 924        unless ($read == 1 && $newline eq "\n") {
 925                $self->_close_cat_blob();
 926                throw Error::Simple("didn't find newline after blob");
 927        }
 928
 929        unless (print $fh $blob) {
 930                $self->_close_cat_blob();
 931                throw Error::Simple("couldn't write to passed in filehandle");
 932        }
 933
 934        return $size;
 935}
 936
 937sub _open_cat_blob_if_needed {
 938        my ($self) = @_;
 939
 940        return if defined($self->{cat_blob_pid});
 941
 942        ($self->{cat_blob_pid}, $self->{cat_blob_in},
 943         $self->{cat_blob_out}, $self->{cat_blob_ctx}) =
 944                $self->command_bidi_pipe(qw(cat-file --batch));
 945}
 946
 947sub _close_cat_blob {
 948        my ($self) = @_;
 949
 950        return unless defined($self->{cat_blob_pid});
 951
 952        my @vars = map { 'cat_blob_' . $_ } qw(pid in out ctx);
 953
 954        command_close_bidi_pipe(@$self{@vars});
 955        delete @$self{@vars};
 956}
 957
 958
 959{ # %TEMP_* Lexical Context
 960
 961my (%TEMP_FILEMAP, %TEMP_FILES);
 962
 963=item temp_acquire ( NAME )
 964
 965Attempts to retreive the temporary file mapped to the string C<NAME>. If an
 966associated temp file has not been created this session or was closed, it is
 967created, cached, and set for autoflush and binmode.
 968
 969Internally locks the file mapped to C<NAME>. This lock must be released with
 970C<temp_release()> when the temp file is no longer needed. Subsequent attempts
 971to retrieve temporary files mapped to the same C<NAME> while still locked will
 972cause an error. This locking mechanism provides a weak guarantee and is not
 973threadsafe. It does provide some error checking to help prevent temp file refs
 974writing over one another.
 975
 976In general, the L<File::Handle> returned should not be closed by consumers as
 977it defeats the purpose of this caching mechanism. If you need to close the temp
 978file handle, then you should use L<File::Temp> or another temp file faculty
 979directly. If a handle is closed and then requested again, then a warning will
 980issue.
 981
 982=cut
 983
 984sub temp_acquire {
 985        my $temp_fd = _temp_cache(@_);
 986
 987        $TEMP_FILES{$temp_fd}{locked} = 1;
 988        $temp_fd;
 989}
 990
 991=item temp_release ( NAME )
 992
 993=item temp_release ( FILEHANDLE )
 994
 995Releases a lock acquired through C<temp_acquire()>. Can be called either with
 996the C<NAME> mapping used when acquiring the temp file or with the C<FILEHANDLE>
 997referencing a locked temp file.
 998
 999Warns if an attempt is made to release a file that is not locked.
1000
1001The temp file will be truncated before being released. This can help to reduce
1002disk I/O where the system is smart enough to detect the truncation while data
1003is in the output buffers. Beware that after the temp file is released and
1004truncated, any operations on that file may fail miserably until it is
1005re-acquired. All contents are lost between each release and acquire mapped to
1006the same string.
1007
1008=cut
1009
1010sub temp_release {
1011        my ($self, $temp_fd, $trunc) = _maybe_self(@_);
1012
1013        if (exists $TEMP_FILEMAP{$temp_fd}) {
1014                $temp_fd = $TEMP_FILES{$temp_fd};
1015        }
1016        unless ($TEMP_FILES{$temp_fd}{locked}) {
1017                carp "Attempt to release temp file '",
1018                        $temp_fd, "' that has not been locked";
1019        }
1020        temp_reset($temp_fd) if $trunc and $temp_fd->opened;
1021
1022        $TEMP_FILES{$temp_fd}{locked} = 0;
1023        undef;
1024}
1025
1026sub _temp_cache {
1027        my ($self, $name) = _maybe_self(@_);
1028
1029        _verify_require();
1030
1031        my $temp_fd = \$TEMP_FILEMAP{$name};
1032        if (defined $$temp_fd and $$temp_fd->opened) {
1033                if ($TEMP_FILES{$$temp_fd}{locked}) {
1034                        throw Error::Simple("Temp file with moniker '" .
1035                                $name . "' already in use");
1036                }
1037        } else {
1038                if (defined $$temp_fd) {
1039                        # then we're here because of a closed handle.
1040                        carp "Temp file '", $name,
1041                                "' was closed. Opening replacement.";
1042                }
1043                my $fname;
1044
1045                my $tmpdir;
1046                if (defined $self) {
1047                        $tmpdir = $self->repo_path();
1048                }
1049
1050                ($$temp_fd, $fname) = File::Temp->tempfile(
1051                        'Git_XXXXXX', UNLINK => 1, DIR => $tmpdir,
1052                        ) or throw Error::Simple("couldn't open new temp file");
1053
1054                $$temp_fd->autoflush;
1055                binmode $$temp_fd;
1056                $TEMP_FILES{$$temp_fd}{fname} = $fname;
1057        }
1058        $$temp_fd;
1059}
1060
1061sub _verify_require {
1062        eval { require File::Temp; require File::Spec; };
1063        $@ and throw Error::Simple($@);
1064}
1065
1066=item temp_reset ( FILEHANDLE )
1067
1068Truncates and resets the position of the C<FILEHANDLE>.
1069
1070=cut
1071
1072sub temp_reset {
1073        my ($self, $temp_fd) = _maybe_self(@_);
1074
1075        truncate $temp_fd, 0
1076                or throw Error::Simple("couldn't truncate file");
1077        sysseek($temp_fd, 0, SEEK_SET) and seek($temp_fd, 0, SEEK_SET)
1078                or throw Error::Simple("couldn't seek to beginning of file");
1079        sysseek($temp_fd, 0, SEEK_CUR) == 0 and tell($temp_fd) == 0
1080                or throw Error::Simple("expected file position to be reset");
1081}
1082
1083=item temp_path ( NAME )
1084
1085=item temp_path ( FILEHANDLE )
1086
1087Returns the filename associated with the given tempfile.
1088
1089=cut
1090
1091sub temp_path {
1092        my ($self, $temp_fd) = _maybe_self(@_);
1093
1094        if (exists $TEMP_FILEMAP{$temp_fd}) {
1095                $temp_fd = $TEMP_FILEMAP{$temp_fd};
1096        }
1097        $TEMP_FILES{$temp_fd}{fname};
1098}
1099
1100sub END {
1101        unlink values %TEMP_FILEMAP if %TEMP_FILEMAP;
1102}
1103
1104} # %TEMP_* Lexical Context
1105
1106=back
1107
1108=head1 ERROR HANDLING
1109
1110All functions are supposed to throw Perl exceptions in case of errors.
1111See the L<Error> module on how to catch those. Most exceptions are mere
1112L<Error::Simple> instances.
1113
1114However, the C<command()>, C<command_oneline()> and C<command_noisy()>
1115functions suite can throw C<Git::Error::Command> exceptions as well: those are
1116thrown when the external command returns an error code and contain the error
1117code as well as access to the captured command's output. The exception class
1118provides the usual C<stringify> and C<value> (command's exit code) methods and
1119in addition also a C<cmd_output> method that returns either an array or a
1120string with the captured command output (depending on the original function
1121call context; C<command_noisy()> returns C<undef>) and $<cmdline> which
1122returns the command and its arguments (but without proper quoting).
1123
1124Note that the C<command_*_pipe()> functions cannot throw this exception since
1125it has no idea whether the command failed or not. You will only find out
1126at the time you C<close> the pipe; if you want to have that automated,
1127use C<command_close_pipe()>, which can throw the exception.
1128
1129=cut
1130
1131{
1132        package Git::Error::Command;
1133
1134        @Git::Error::Command::ISA = qw(Error);
1135
1136        sub new {
1137                my $self = shift;
1138                my $cmdline = '' . shift;
1139                my $value = 0 + shift;
1140                my $outputref = shift;
1141                my(@args) = ();
1142
1143                local $Error::Depth = $Error::Depth + 1;
1144
1145                push(@args, '-cmdline', $cmdline);
1146                push(@args, '-value', $value);
1147                push(@args, '-outputref', $outputref);
1148
1149                $self->SUPER::new(-text => 'command returned error', @args);
1150        }
1151
1152        sub stringify {
1153                my $self = shift;
1154                my $text = $self->SUPER::stringify;
1155                $self->cmdline() . ': ' . $text . ': ' . $self->value() . "\n";
1156        }
1157
1158        sub cmdline {
1159                my $self = shift;
1160                $self->{'-cmdline'};
1161        }
1162
1163        sub cmd_output {
1164                my $self = shift;
1165                my $ref = $self->{'-outputref'};
1166                defined $ref or undef;
1167                if (ref $ref eq 'ARRAY') {
1168                        return @$ref;
1169                } else { # SCALAR
1170                        return $$ref;
1171                }
1172        }
1173}
1174
1175=over 4
1176
1177=item git_cmd_try { CODE } ERRMSG
1178
1179This magical statement will automatically catch any C<Git::Error::Command>
1180exceptions thrown by C<CODE> and make your program die with C<ERRMSG>
1181on its lips; the message will have %s substituted for the command line
1182and %d for the exit status. This statement is useful mostly for producing
1183more user-friendly error messages.
1184
1185In case of no exception caught the statement returns C<CODE>'s return value.
1186
1187Note that this is the only auto-exported function.
1188
1189=cut
1190
1191sub git_cmd_try(&$) {
1192        my ($code, $errmsg) = @_;
1193        my @result;
1194        my $err;
1195        my $array = wantarray;
1196        try {
1197                if ($array) {
1198                        @result = &$code;
1199                } else {
1200                        $result[0] = &$code;
1201                }
1202        } catch Git::Error::Command with {
1203                my $E = shift;
1204                $err = $errmsg;
1205                $err =~ s/\%s/$E->cmdline()/ge;
1206                $err =~ s/\%d/$E->value()/ge;
1207                # We can't croak here since Error.pm would mangle
1208                # that to Error::Simple.
1209        };
1210        $err and croak $err;
1211        return $array ? @result : $result[0];
1212}
1213
1214
1215=back
1216
1217=head1 COPYRIGHT
1218
1219Copyright 2006 by Petr Baudis E<lt>pasky@suse.czE<gt>.
1220
1221This module is free software; it may be used, copied, modified
1222and distributed under the terms of the GNU General Public Licence,
1223either version 2, or (at your option) any later version.
1224
1225=cut
1226
1227
1228# Take raw method argument list and return ($obj, @args) in case
1229# the method was called upon an instance and (undef, @args) if
1230# it was called directly.
1231sub _maybe_self {
1232        UNIVERSAL::isa($_[0], 'Git') ? @_ : (undef, @_);
1233}
1234
1235# Check if the command id is something reasonable.
1236sub _check_valid_cmd {
1237        my ($cmd) = @_;
1238        $cmd =~ /^[a-z0-9A-Z_-]+$/ or throw Error::Simple("bad command: $cmd");
1239}
1240
1241# Common backend for the pipe creators.
1242sub _command_common_pipe {
1243        my $direction = shift;
1244        my ($self, @p) = _maybe_self(@_);
1245        my (%opts, $cmd, @args);
1246        if (ref $p[0]) {
1247                ($cmd, @args) = @{shift @p};
1248                %opts = ref $p[0] ? %{$p[0]} : @p;
1249        } else {
1250                ($cmd, @args) = @p;
1251        }
1252        _check_valid_cmd($cmd);
1253
1254        my $fh;
1255        if ($^O eq 'MSWin32') {
1256                # ActiveState Perl
1257                #defined $opts{STDERR} and
1258                #       warn 'ignoring STDERR option - running w/ ActiveState';
1259                $direction eq '-|' or
1260                        die 'input pipe for ActiveState not implemented';
1261                # the strange construction with *ACPIPE is just to
1262                # explain the tie below that we want to bind to
1263                # a handle class, not scalar. It is not known if
1264                # it is something specific to ActiveState Perl or
1265                # just a Perl quirk.
1266                tie (*ACPIPE, 'Git::activestate_pipe', $cmd, @args);
1267                $fh = *ACPIPE;
1268
1269        } else {
1270                my $pid = open($fh, $direction);
1271                if (not defined $pid) {
1272                        throw Error::Simple("open failed: $!");
1273                } elsif ($pid == 0) {
1274                        if (defined $opts{STDERR}) {
1275                                close STDERR;
1276                        }
1277                        if ($opts{STDERR}) {
1278                                open (STDERR, '>&', $opts{STDERR})
1279                                        or die "dup failed: $!";
1280                        }
1281                        _cmd_exec($self, $cmd, @args);
1282                }
1283        }
1284        return wantarray ? ($fh, join(' ', $cmd, @args)) : $fh;
1285}
1286
1287# When already in the subprocess, set up the appropriate state
1288# for the given repository and execute the git command.
1289sub _cmd_exec {
1290        my ($self, @args) = @_;
1291        _setup_git_cmd_env($self);
1292        _execv_git_cmd(@args);
1293        die qq[exec "@args" failed: $!];
1294}
1295
1296# set up the appropriate state for git command
1297sub _setup_git_cmd_env {
1298        my $self = shift;
1299        if ($self) {
1300                $self->repo_path() and $ENV{'GIT_DIR'} = $self->repo_path();
1301                $self->repo_path() and $self->wc_path()
1302                        and $ENV{'GIT_WORK_TREE'} = $self->wc_path();
1303                $self->wc_path() and chdir($self->wc_path());
1304                $self->wc_subdir() and chdir($self->wc_subdir());
1305        }
1306}
1307
1308# Execute the given Git command ($_[0]) with arguments ($_[1..])
1309# by searching for it at proper places.
1310sub _execv_git_cmd { exec('git', @_); }
1311
1312# Close pipe to a subprocess.
1313sub _cmd_close {
1314        my ($fh, $ctx) = @_;
1315        if (not close $fh) {
1316                if ($!) {
1317                        # It's just close, no point in fatalities
1318                        carp "error closing pipe: $!";
1319                } elsif ($? >> 8) {
1320                        # The caller should pepper this.
1321                        throw Git::Error::Command($ctx, $? >> 8);
1322                }
1323                # else we might e.g. closed a live stream; the command
1324                # dying of SIGPIPE would drive us here.
1325        }
1326}
1327
1328
1329sub DESTROY {
1330        my ($self) = @_;
1331        $self->_close_hash_and_insert_object();
1332        $self->_close_cat_blob();
1333}
1334
1335
1336# Pipe implementation for ActiveState Perl.
1337
1338package Git::activestate_pipe;
1339use strict;
1340
1341sub TIEHANDLE {
1342        my ($class, @params) = @_;
1343        # FIXME: This is probably horrible idea and the thing will explode
1344        # at the moment you give it arguments that require some quoting,
1345        # but I have no ActiveState clue... --pasky
1346        # Let's just hope ActiveState Perl does at least the quoting
1347        # correctly.
1348        my @data = qx{git @params};
1349        bless { i => 0, data => \@data }, $class;
1350}
1351
1352sub READLINE {
1353        my $self = shift;
1354        if ($self->{i} >= scalar @{$self->{data}}) {
1355                return undef;
1356        }
1357        my $i = $self->{i};
1358        if (wantarray) {
1359                $self->{i} = $#{$self->{'data'}} + 1;
1360                return splice(@{$self->{'data'}}, $i);
1361        }
1362        $self->{i} = $i + 1;
1363        return $self->{'data'}->[ $i ];
1364}
1365
1366sub CLOSE {
1367        my $self = shift;
1368        delete $self->{data};
1369        delete $self->{i};
1370}
1371
1372sub EOF {
1373        my $self = shift;
1374        return ($self->{i} >= scalar @{$self->{data}});
1375}
1376
1377
13781; # Famous last words