perl / Git.pmon commit Merge branch 'maint' (5bcde30)
   1=head1 NAME
   2
   3Git - Perl interface to the Git version control system
   4
   5=cut
   6
   7
   8package Git;
   9
  10use strict;
  11
  12
  13BEGIN {
  14
  15our ($VERSION, @ISA, @EXPORT, @EXPORT_OK);
  16
  17# Totally unstable API.
  18$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
  49require 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
  61
  62=head1 DESCRIPTION
  63
  64This module provides Perl scripts easy way to interface the Git version control
  65system. The modules have an easy and well-tested way to call arbitrary Git
  66commands; in the future, the interface will also provide specialized methods
  67for doing easily operations which are not totally trivial to do over
  68the generic command interface.
  69
  70While some commands can be executed outside of any context (e.g. 'version'
  71or 'init'), most operations require a repository context, which in practice
  72means getting an instance of the Git object using the repository() constructor.
  73(In the future, we will also get a new_repository() constructor.) All commands
  74called as methods of the object are then executed in the context of the
  75repository.
  76
  77Part of the "repository state" is also information about path to the attached
  78working copy (unless you work with a bare repository). You can also navigate
  79inside of the working copy using the C<wc_chdir()> method. (Note that
  80the repository object is self-contained and will not change working directory
  81of your process.)
  82
  83TODO: In the future, we might also do
  84
  85        my $remoterepo = $repo->remote_repository (Name => 'cogito', Branch => 'master');
  86        $remoterepo ||= Git->remote_repository ('http://git.or.cz/cogito.git/');
  87        my @refs = $remoterepo->refs();
  88
  89Currently, the module merely wraps calls to external Git tools. In the future,
  90it will provide a much faster way to interact with Git by linking directly
  91to libgit. This should be completely opaque to the user, though (performance
  92increate nonwithstanding).
  93
  94=cut
  95
  96
  97use Carp qw(carp croak); # but croak is bad - throw instead
  98use Error qw(:try);
  99use Cwd qw(abs_path);
 100use IPC::Open2 qw(open2);
 101
 102}
 103
 104
 105=head1 CONSTRUCTORS
 106
 107=over 4
 108
 109=item repository ( OPTIONS )
 110
 111=item repository ( DIRECTORY )
 112
 113=item repository ()
 114
 115Construct a new repository object.
 116C<OPTIONS> are passed in a hash like fashion, using key and value pairs.
 117Possible options are:
 118
 119B<Repository> - Path to the Git repository.
 120
 121B<WorkingCopy> - Path to the associated working copy; not strictly required
 122as many commands will happily crunch on a bare repository.
 123
 124B<WorkingSubdir> - Subdirectory in the working copy to work inside.
 125Just left undefined if you do not want to limit the scope of operations.
 126
 127B<Directory> - Path to the Git working directory in its usual setup.
 128The C<.git> directory is searched in the directory and all the parent
 129directories; if found, C<WorkingCopy> is set to the directory containing
 130it and C<Repository> to the C<.git> directory itself. If no C<.git>
 131directory was found, the C<Directory> is assumed to be a bare repository,
 132C<Repository> is set to point at it and C<WorkingCopy> is left undefined.
 133If the C<$GIT_DIR> environment variable is set, things behave as expected
 134as well.
 135
 136You should not use both C<Directory> and either of C<Repository> and
 137C<WorkingCopy> - the results of that are undefined.
 138
 139Alternatively, a directory path may be passed as a single scalar argument
 140to the constructor; it is equivalent to setting only the C<Directory> option
 141field.
 142
 143Calling the constructor with no options whatsoever is equivalent to
 144calling it with C<< Directory => '.' >>. In general, if you are building
 145a standard porcelain command, simply doing C<< Git->repository() >> should
 146do the right thing and setup the object to reflect exactly where the user
 147is right now.
 148
 149=cut
 150
 151sub repository {
 152        my $class = shift;
 153        my @args = @_;
 154        my %opts = ();
 155        my $self;
 156
 157        if (defined $args[0]) {
 158                if ($#args % 2 != 1) {
 159                        # Not a hash.
 160                        $#args == 0 or throw Error::Simple("bad usage");
 161                        %opts = ( Directory => $args[0] );
 162                } else {
 163                        %opts = @args;
 164                }
 165        }
 166
 167        if (not defined $opts{Repository} and not defined $opts{WorkingCopy}) {
 168                $opts{Directory} ||= '.';
 169        }
 170
 171        if ($opts{Directory}) {
 172                -d $opts{Directory} or throw Error::Simple("Directory not found: $!");
 173
 174                my $search = Git->repository(WorkingCopy => $opts{Directory});
 175                my $dir;
 176                try {
 177                        $dir = $search->command_oneline(['rev-parse', '--git-dir'],
 178                                                        STDERR => 0);
 179                } catch Git::Error::Command with {
 180                        $dir = undef;
 181                };
 182
 183                if ($dir) {
 184                        $dir =~ m#^/# or $dir = $opts{Directory} . '/' . $dir;
 185                        $opts{Repository} = $dir;
 186
 187                        # If --git-dir went ok, this shouldn't die either.
 188                        my $prefix = $search->command_oneline('rev-parse', '--show-prefix');
 189                        $dir = abs_path($opts{Directory}) . '/';
 190                        if ($prefix) {
 191                                if (substr($dir, -length($prefix)) ne $prefix) {
 192                                        throw Error::Simple("rev-parse confused me - $dir does not have trailing $prefix");
 193                                }
 194                                substr($dir, -length($prefix)) = '';
 195                        }
 196                        $opts{WorkingCopy} = $dir;
 197                        $opts{WorkingSubdir} = $prefix;
 198
 199                } else {
 200                        # A bare repository? Let's see...
 201                        $dir = $opts{Directory};
 202
 203                        unless (-d "$dir/refs" and -d "$dir/objects" and -e "$dir/HEAD") {
 204                                # Mimick git-rev-parse --git-dir error message:
 205                                throw Error::Simple('fatal: Not a git repository');
 206                        }
 207                        my $search = Git->repository(Repository => $dir);
 208                        try {
 209                                $search->command('symbolic-ref', 'HEAD');
 210                        } catch Git::Error::Command with {
 211                                # Mimick git-rev-parse --git-dir error message:
 212                                throw Error::Simple('fatal: Not a git repository');
 213                        }
 214
 215                        $opts{Repository} = abs_path($dir);
 216                }
 217
 218                delete $opts{Directory};
 219        }
 220
 221        $self = { opts => \%opts };
 222        bless $self, $class;
 223}
 224
 225=back
 226
 227=head1 METHODS
 228
 229=over 4
 230
 231=item command ( COMMAND [, ARGUMENTS... ] )
 232
 233=item command ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 234
 235Execute the given Git C<COMMAND> (specify it without the 'git-'
 236prefix), optionally with the specified extra C<ARGUMENTS>.
 237
 238The second more elaborate form can be used if you want to further adjust
 239the command execution. Currently, only one option is supported:
 240
 241B<STDERR> - How to deal with the command's error output. By default (C<undef>)
 242it is delivered to the caller's C<STDERR>. A false value (0 or '') will cause
 243it to be thrown away. If you want to process it, you can get it in a filehandle
 244you specify, but you must be extremely careful; if the error output is not
 245very short and you want to read it in the same process as where you called
 246C<command()>, you are set up for a nice deadlock!
 247
 248The method can be called without any instance or on a specified Git repository
 249(in that case the command will be run in the repository context).
 250
 251In scalar context, it returns all the command output in a single string
 252(verbatim).
 253
 254In array context, it returns an array containing lines printed to the
 255command's stdout (without trailing newlines).
 256
 257In both cases, the command's stdin and stderr are the same as the caller's.
 258
 259=cut
 260
 261sub command {
 262        my ($fh, $ctx) = command_output_pipe(@_);
 263
 264        if (not defined wantarray) {
 265                # Nothing to pepper the possible exception with.
 266                _cmd_close($fh, $ctx);
 267
 268        } elsif (not wantarray) {
 269                local $/;
 270                my $text = <$fh>;
 271                try {
 272                        _cmd_close($fh, $ctx);
 273                } catch Git::Error::Command with {
 274                        # Pepper with the output:
 275                        my $E = shift;
 276                        $E->{'-outputref'} = \$text;
 277                        throw $E;
 278                };
 279                return $text;
 280
 281        } else {
 282                my @lines = <$fh>;
 283                defined and chomp for @lines;
 284                try {
 285                        _cmd_close($fh, $ctx);
 286                } catch Git::Error::Command with {
 287                        my $E = shift;
 288                        $E->{'-outputref'} = \@lines;
 289                        throw $E;
 290                };
 291                return @lines;
 292        }
 293}
 294
 295
 296=item command_oneline ( COMMAND [, ARGUMENTS... ] )
 297
 298=item command_oneline ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 299
 300Execute the given C<COMMAND> in the same way as command()
 301does but always return a scalar string containing the first line
 302of the command's standard output.
 303
 304=cut
 305
 306sub command_oneline {
 307        my ($fh, $ctx) = command_output_pipe(@_);
 308
 309        my $line = <$fh>;
 310        defined $line and chomp $line;
 311        try {
 312                _cmd_close($fh, $ctx);
 313        } catch Git::Error::Command with {
 314                # Pepper with the output:
 315                my $E = shift;
 316                $E->{'-outputref'} = \$line;
 317                throw $E;
 318        };
 319        return $line;
 320}
 321
 322
 323=item command_output_pipe ( COMMAND [, ARGUMENTS... ] )
 324
 325=item command_output_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 326
 327Execute the given C<COMMAND> in the same way as command()
 328does but return a pipe filehandle from which the command output can be
 329read.
 330
 331The function can return C<($pipe, $ctx)> in array context.
 332See C<command_close_pipe()> for details.
 333
 334=cut
 335
 336sub command_output_pipe {
 337        _command_common_pipe('-|', @_);
 338}
 339
 340
 341=item command_input_pipe ( COMMAND [, ARGUMENTS... ] )
 342
 343=item command_input_pipe ( [ COMMAND, ARGUMENTS... ], { Opt => Val ... } )
 344
 345Execute the given C<COMMAND> in the same way as command_output_pipe()
 346does but return an input pipe filehandle instead; the command output
 347is not captured.
 348
 349The function can return C<($pipe, $ctx)> in array context.
 350See C<command_close_pipe()> for details.
 351
 352=cut
 353
 354sub command_input_pipe {
 355        _command_common_pipe('|-', @_);
 356}
 357
 358
 359=item command_close_pipe ( PIPE [, CTX ] )
 360
 361Close the C<PIPE> as returned from C<command_*_pipe()>, checking
 362whether the command finished successfully. The optional C<CTX> argument
 363is required if you want to see the command name in the error message,
 364and it is the second value returned by C<command_*_pipe()> when
 365called in array context. The call idiom is:
 366
 367        my ($fh, $ctx) = $r->command_output_pipe('status');
 368        while (<$fh>) { ... }
 369        $r->command_close_pipe($fh, $ctx);
 370
 371Note that you should not rely on whatever actually is in C<CTX>;
 372currently it is simply the command name but in future the context might
 373have more complicated structure.
 374
 375=cut
 376
 377sub command_close_pipe {
 378        my ($self, $fh, $ctx) = _maybe_self(@_);
 379        $ctx ||= '<unknown>';
 380        _cmd_close($fh, $ctx);
 381}
 382
 383=item command_bidi_pipe ( COMMAND [, ARGUMENTS... ] )
 384
 385Execute the given C<COMMAND> in the same way as command_output_pipe()
 386does but return both an input pipe filehandle and an output pipe filehandle.
 387
 388The function will return return C<($pid, $pipe_in, $pipe_out, $ctx)>.
 389See C<command_close_bidi_pipe()> for details.
 390
 391=cut
 392
 393sub command_bidi_pipe {
 394        my ($pid, $in, $out);
 395        $pid = open2($in, $out, 'git', @_);
 396        return ($pid, $in, $out, join(' ', @_));
 397}
 398
 399=item command_close_bidi_pipe ( PID, PIPE_IN, PIPE_OUT [, CTX] )
 400
 401Close the C<PIPE_IN> and C<PIPE_OUT> as returned from C<command_bidi_pipe()>,
 402checking whether the command finished successfully. The optional C<CTX>
 403argument is required if you want to see the command name in the error message,
 404and it is the fourth value returned by C<command_bidi_pipe()>.  The call idiom
 405is:
 406
 407        my ($pid, $in, $out, $ctx) = $r->command_bidi_pipe('cat-file --batch-check');
 408        print "000000000\n" $out;
 409        while (<$in>) { ... }
 410        $r->command_close_bidi_pipe($pid, $in, $out, $ctx);
 411
 412Note that you should not rely on whatever actually is in C<CTX>;
 413currently it is simply the command name but in future the context might
 414have more complicated structure.
 415
 416=cut
 417
 418sub command_close_bidi_pipe {
 419        my ($pid, $in, $out, $ctx) = @_;
 420        foreach my $fh ($in, $out) {
 421                unless (close $fh) {
 422                        if ($!) {
 423                                carp "error closing pipe: $!";
 424                        } elsif ($? >> 8) {
 425                                throw Git::Error::Command($ctx, $? >>8);
 426                        }
 427                }
 428        }
 429
 430        waitpid $pid, 0;
 431
 432        if ($? >> 8) {
 433                throw Git::Error::Command($ctx, $? >>8);
 434        }
 435}
 436
 437
 438=item command_noisy ( COMMAND [, ARGUMENTS... ] )
 439
 440Execute the given C<COMMAND> in the same way as command() does but do not
 441capture the command output - the standard output is not redirected and goes
 442to the standard output of the caller application.
 443
 444While the method is called command_noisy(), you might want to as well use
 445it for the most silent Git commands which you know will never pollute your
 446stdout but you want to avoid the overhead of the pipe setup when calling them.
 447
 448The function returns only after the command has finished running.
 449
 450=cut
 451
 452sub command_noisy {
 453        my ($self, $cmd, @args) = _maybe_self(@_);
 454        _check_valid_cmd($cmd);
 455
 456        my $pid = fork;
 457        if (not defined $pid) {
 458                throw Error::Simple("fork failed: $!");
 459        } elsif ($pid == 0) {
 460                _cmd_exec($self, $cmd, @args);
 461        }
 462        if (waitpid($pid, 0) > 0 and $?>>8 != 0) {
 463                throw Git::Error::Command(join(' ', $cmd, @args), $? >> 8);
 464        }
 465}
 466
 467
 468=item version ()
 469
 470Return the Git version in use.
 471
 472=cut
 473
 474sub version {
 475        my $verstr = command_oneline('--version');
 476        $verstr =~ s/^git version //;
 477        $verstr;
 478}
 479
 480
 481=item exec_path ()
 482
 483Return path to the Git sub-command executables (the same as
 484C<git --exec-path>). Useful mostly only internally.
 485
 486=cut
 487
 488sub exec_path { command_oneline('--exec-path') }
 489
 490
 491=item repo_path ()
 492
 493Return path to the git repository. Must be called on a repository instance.
 494
 495=cut
 496
 497sub repo_path { $_[0]->{opts}->{Repository} }
 498
 499
 500=item wc_path ()
 501
 502Return path to the working copy. Must be called on a repository instance.
 503
 504=cut
 505
 506sub wc_path { $_[0]->{opts}->{WorkingCopy} }
 507
 508
 509=item wc_subdir ()
 510
 511Return path to the subdirectory inside of a working copy. Must be called
 512on a repository instance.
 513
 514=cut
 515
 516sub wc_subdir { $_[0]->{opts}->{WorkingSubdir} ||= '' }
 517
 518
 519=item wc_chdir ( SUBDIR )
 520
 521Change the working copy subdirectory to work within. The C<SUBDIR> is
 522relative to the working copy root directory (not the current subdirectory).
 523Must be called on a repository instance attached to a working copy
 524and the directory must exist.
 525
 526=cut
 527
 528sub wc_chdir {
 529        my ($self, $subdir) = @_;
 530        $self->wc_path()
 531                or throw Error::Simple("bare repository");
 532
 533        -d $self->wc_path().'/'.$subdir
 534                or throw Error::Simple("subdir not found: $!");
 535        # Of course we will not "hold" the subdirectory so anyone
 536        # can delete it now and we will never know. But at least we tried.
 537
 538        $self->{opts}->{WorkingSubdir} = $subdir;
 539}
 540
 541
 542=item config ( VARIABLE )
 543
 544Retrieve the configuration C<VARIABLE> in the same manner as C<config>
 545does. In scalar context requires the variable to be set only one time
 546(exception is thrown otherwise), in array context returns allows the
 547variable to be set multiple times and returns all the values.
 548
 549This currently wraps command('config') so it is not so fast.
 550
 551=cut
 552
 553sub config {
 554        my ($self, $var) = _maybe_self(@_);
 555
 556        try {
 557                my @cmd = ('config');
 558                unshift @cmd, $self if $self;
 559                if (wantarray) {
 560                        return command(@cmd, '--get-all', $var);
 561                } else {
 562                        return command_oneline(@cmd, '--get', $var);
 563                }
 564        } catch Git::Error::Command with {
 565                my $E = shift;
 566                if ($E->value() == 1) {
 567                        # Key not found.
 568                        return;
 569                } else {
 570                        throw $E;
 571                }
 572        };
 573}
 574
 575
 576=item config_bool ( VARIABLE )
 577
 578Retrieve the bool configuration C<VARIABLE>. The return value
 579is usable as a boolean in perl (and C<undef> if it's not defined,
 580of course).
 581
 582This currently wraps command('config') so it is not so fast.
 583
 584=cut
 585
 586sub config_bool {
 587        my ($self, $var) = _maybe_self(@_);
 588
 589        try {
 590                my @cmd = ('config', '--bool', '--get', $var);
 591                unshift @cmd, $self if $self;
 592                my $val = command_oneline(@cmd);
 593                return undef unless defined $val;
 594                return $val eq 'true';
 595        } catch Git::Error::Command with {
 596                my $E = shift;
 597                if ($E->value() == 1) {
 598                        # Key not found.
 599                        return undef;
 600                } else {
 601                        throw $E;
 602                }
 603        };
 604}
 605
 606=item config_int ( VARIABLE )
 607
 608Retrieve the integer configuration C<VARIABLE>. The return value
 609is simple decimal number.  An optional value suffix of 'k', 'm',
 610or 'g' in the config file will cause the value to be multiplied
 611by 1024, 1048576 (1024^2), or 1073741824 (1024^3) prior to output.
 612It would return C<undef> if configuration variable is not defined,
 613
 614This currently wraps command('config') so it is not so fast.
 615
 616=cut
 617
 618sub config_int {
 619        my ($self, $var) = _maybe_self(@_);
 620
 621        try {
 622                my @cmd = ('config', '--int', '--get', $var);
 623                unshift @cmd, $self if $self;
 624                return command_oneline(@cmd);
 625        } catch Git::Error::Command with {
 626                my $E = shift;
 627                if ($E->value() == 1) {
 628                        # Key not found.
 629                        return undef;
 630                } else {
 631                        throw $E;
 632                }
 633        };
 634}
 635
 636=item get_colorbool ( NAME )
 637
 638Finds if color should be used for NAMEd operation from the configuration,
 639and returns boolean (true for "use color", false for "do not use color").
 640
 641=cut
 642
 643sub get_colorbool {
 644        my ($self, $var) = @_;
 645        my $stdout_to_tty = (-t STDOUT) ? "true" : "false";
 646        my $use_color = $self->command_oneline('config', '--get-colorbool',
 647                                               $var, $stdout_to_tty);
 648        return ($use_color eq 'true');
 649}
 650
 651=item get_color ( SLOT, COLOR )
 652
 653Finds color for SLOT from the configuration, while defaulting to COLOR,
 654and returns the ANSI color escape sequence:
 655
 656        print $repo->get_color("color.interactive.prompt", "underline blue white");
 657        print "some text";
 658        print $repo->get_color("", "normal");
 659
 660=cut
 661
 662sub get_color {
 663        my ($self, $slot, $default) = @_;
 664        my $color = $self->command_oneline('config', '--get-color', $slot, $default);
 665        if (!defined $color) {
 666                $color = "";
 667        }
 668        return $color;
 669}
 670
 671=item ident ( TYPE | IDENTSTR )
 672
 673=item ident_person ( TYPE | IDENTSTR | IDENTARRAY )
 674
 675This suite of functions retrieves and parses ident information, as stored
 676in the commit and tag objects or produced by C<var GIT_type_IDENT> (thus
 677C<TYPE> can be either I<author> or I<committer>; case is insignificant).
 678
 679The C<ident> method retrieves the ident information from C<git-var>
 680and either returns it as a scalar string or as an array with the fields parsed.
 681Alternatively, it can take a prepared ident string (e.g. from the commit
 682object) and just parse it.
 683
 684C<ident_person> returns the person part of the ident - name and email;
 685it can take the same arguments as C<ident> or the array returned by C<ident>.
 686
 687The synopsis is like:
 688
 689        my ($name, $email, $time_tz) = ident('author');
 690        "$name <$email>" eq ident_person('author');
 691        "$name <$email>" eq ident_person($name);
 692        $time_tz =~ /^\d+ [+-]\d{4}$/;
 693
 694=cut
 695
 696sub ident {
 697        my ($self, $type) = _maybe_self(@_);
 698        my $identstr;
 699        if (lc $type eq lc 'committer' or lc $type eq lc 'author') {
 700                my @cmd = ('var', 'GIT_'.uc($type).'_IDENT');
 701                unshift @cmd, $self if $self;
 702                $identstr = command_oneline(@cmd);
 703        } else {
 704                $identstr = $type;
 705        }
 706        if (wantarray) {
 707                return $identstr =~ /^(.*) <(.*)> (\d+ [+-]\d{4})$/;
 708        } else {
 709                return $identstr;
 710        }
 711}
 712
 713sub ident_person {
 714        my ($self, @ident) = _maybe_self(@_);
 715        $#ident == 0 and @ident = $self ? $self->ident($ident[0]) : ident($ident[0]);
 716        return "$ident[0] <$ident[1]>";
 717}
 718
 719
 720=item hash_object ( TYPE, FILENAME )
 721
 722Compute the SHA1 object id of the given C<FILENAME> considering it is
 723of the C<TYPE> object type (C<blob>, C<commit>, C<tree>).
 724
 725The method can be called without any instance or on a specified Git repository,
 726it makes zero difference.
 727
 728The function returns the SHA1 hash.
 729
 730=cut
 731
 732# TODO: Support for passing FILEHANDLE instead of FILENAME
 733sub hash_object {
 734        my ($self, $type, $file) = _maybe_self(@_);
 735        command_oneline('hash-object', '-t', $type, $file);
 736}
 737
 738
 739=item hash_and_insert_object ( FILENAME )
 740
 741Compute the SHA1 object id of the given C<FILENAME> and add the object to the
 742object database.
 743
 744The function returns the SHA1 hash.
 745
 746=cut
 747
 748# TODO: Support for passing FILEHANDLE instead of FILENAME
 749sub hash_and_insert_object {
 750        my ($self, $filename) = @_;
 751
 752        carp "Bad filename \"$filename\"" if $filename =~ /[\r\n]/;
 753
 754        $self->_open_hash_and_insert_object_if_needed();
 755        my ($in, $out) = ($self->{hash_object_in}, $self->{hash_object_out});
 756
 757        unless (print $out $filename, "\n") {
 758                $self->_close_hash_and_insert_object();
 759                throw Error::Simple("out pipe went bad");
 760        }
 761
 762        chomp(my $hash = <$in>);
 763        unless (defined($hash)) {
 764                $self->_close_hash_and_insert_object();
 765                throw Error::Simple("in pipe went bad");
 766        }
 767
 768        return $hash;
 769}
 770
 771sub _open_hash_and_insert_object_if_needed {
 772        my ($self) = @_;
 773
 774        return if defined($self->{hash_object_pid});
 775
 776        ($self->{hash_object_pid}, $self->{hash_object_in},
 777         $self->{hash_object_out}, $self->{hash_object_ctx}) =
 778                command_bidi_pipe(qw(hash-object -w --stdin-paths));
 779}
 780
 781sub _close_hash_and_insert_object {
 782        my ($self) = @_;
 783
 784        return unless defined($self->{hash_object_pid});
 785
 786        my @vars = map { 'hash_object_' . $_ } qw(pid in out ctx);
 787
 788        command_close_bidi_pipe($self->{@vars});
 789        delete $self->{@vars};
 790}
 791
 792=item cat_blob ( SHA1, FILEHANDLE )
 793
 794Prints the contents of the blob identified by C<SHA1> to C<FILEHANDLE> and
 795returns the number of bytes printed.
 796
 797=cut
 798
 799sub cat_blob {
 800        my ($self, $sha1, $fh) = @_;
 801
 802        $self->_open_cat_blob_if_needed();
 803        my ($in, $out) = ($self->{cat_blob_in}, $self->{cat_blob_out});
 804
 805        unless (print $out $sha1, "\n") {
 806                $self->_close_cat_blob();
 807                throw Error::Simple("out pipe went bad");
 808        }
 809
 810        my $description = <$in>;
 811        if ($description =~ / missing$/) {
 812                carp "$sha1 doesn't exist in the repository";
 813                return -1;
 814        }
 815
 816        if ($description !~ /^[0-9a-fA-F]{40} \S+ (\d+)$/) {
 817                carp "Unexpected result returned from git cat-file";
 818                return -1;
 819        }
 820
 821        my $size = $1;
 822
 823        my $blob;
 824        my $bytesRead = 0;
 825
 826        while (1) {
 827                my $bytesLeft = $size - $bytesRead;
 828                last unless $bytesLeft;
 829
 830                my $bytesToRead = $bytesLeft < 1024 ? $bytesLeft : 1024;
 831                my $read = read($in, $blob, $bytesToRead, $bytesRead);
 832                unless (defined($read)) {
 833                        $self->_close_cat_blob();
 834                        throw Error::Simple("in pipe went bad");
 835                }
 836
 837                $bytesRead += $read;
 838        }
 839
 840        # Skip past the trailing newline.
 841        my $newline;
 842        my $read = read($in, $newline, 1);
 843        unless (defined($read)) {
 844                $self->_close_cat_blob();
 845                throw Error::Simple("in pipe went bad");
 846        }
 847        unless ($read == 1 && $newline eq "\n") {
 848                $self->_close_cat_blob();
 849                throw Error::Simple("didn't find newline after blob");
 850        }
 851
 852        unless (print $fh $blob) {
 853                $self->_close_cat_blob();
 854                throw Error::Simple("couldn't write to passed in filehandle");
 855        }
 856
 857        return $size;
 858}
 859
 860sub _open_cat_blob_if_needed {
 861        my ($self) = @_;
 862
 863        return if defined($self->{cat_blob_pid});
 864
 865        ($self->{cat_blob_pid}, $self->{cat_blob_in},
 866         $self->{cat_blob_out}, $self->{cat_blob_ctx}) =
 867                command_bidi_pipe(qw(cat-file --batch));
 868}
 869
 870sub _close_cat_blob {
 871        my ($self) = @_;
 872
 873        return unless defined($self->{cat_blob_pid});
 874
 875        my @vars = map { 'cat_blob_' . $_ } qw(pid in out ctx);
 876
 877        command_close_bidi_pipe($self->{@vars});
 878        delete $self->{@vars};
 879}
 880
 881=back
 882
 883=head1 ERROR HANDLING
 884
 885All functions are supposed to throw Perl exceptions in case of errors.
 886See the L<Error> module on how to catch those. Most exceptions are mere
 887L<Error::Simple> instances.
 888
 889However, the C<command()>, C<command_oneline()> and C<command_noisy()>
 890functions suite can throw C<Git::Error::Command> exceptions as well: those are
 891thrown when the external command returns an error code and contain the error
 892code as well as access to the captured command's output. The exception class
 893provides the usual C<stringify> and C<value> (command's exit code) methods and
 894in addition also a C<cmd_output> method that returns either an array or a
 895string with the captured command output (depending on the original function
 896call context; C<command_noisy()> returns C<undef>) and $<cmdline> which
 897returns the command and its arguments (but without proper quoting).
 898
 899Note that the C<command_*_pipe()> functions cannot throw this exception since
 900it has no idea whether the command failed or not. You will only find out
 901at the time you C<close> the pipe; if you want to have that automated,
 902use C<command_close_pipe()>, which can throw the exception.
 903
 904=cut
 905
 906{
 907        package Git::Error::Command;
 908
 909        @Git::Error::Command::ISA = qw(Error);
 910
 911        sub new {
 912                my $self = shift;
 913                my $cmdline = '' . shift;
 914                my $value = 0 + shift;
 915                my $outputref = shift;
 916                my(@args) = ();
 917
 918                local $Error::Depth = $Error::Depth + 1;
 919
 920                push(@args, '-cmdline', $cmdline);
 921                push(@args, '-value', $value);
 922                push(@args, '-outputref', $outputref);
 923
 924                $self->SUPER::new(-text => 'command returned error', @args);
 925        }
 926
 927        sub stringify {
 928                my $self = shift;
 929                my $text = $self->SUPER::stringify;
 930                $self->cmdline() . ': ' . $text . ': ' . $self->value() . "\n";
 931        }
 932
 933        sub cmdline {
 934                my $self = shift;
 935                $self->{'-cmdline'};
 936        }
 937
 938        sub cmd_output {
 939                my $self = shift;
 940                my $ref = $self->{'-outputref'};
 941                defined $ref or undef;
 942                if (ref $ref eq 'ARRAY') {
 943                        return @$ref;
 944                } else { # SCALAR
 945                        return $$ref;
 946                }
 947        }
 948}
 949
 950=over 4
 951
 952=item git_cmd_try { CODE } ERRMSG
 953
 954This magical statement will automatically catch any C<Git::Error::Command>
 955exceptions thrown by C<CODE> and make your program die with C<ERRMSG>
 956on its lips; the message will have %s substituted for the command line
 957and %d for the exit status. This statement is useful mostly for producing
 958more user-friendly error messages.
 959
 960In case of no exception caught the statement returns C<CODE>'s return value.
 961
 962Note that this is the only auto-exported function.
 963
 964=cut
 965
 966sub git_cmd_try(&$) {
 967        my ($code, $errmsg) = @_;
 968        my @result;
 969        my $err;
 970        my $array = wantarray;
 971        try {
 972                if ($array) {
 973                        @result = &$code;
 974                } else {
 975                        $result[0] = &$code;
 976                }
 977        } catch Git::Error::Command with {
 978                my $E = shift;
 979                $err = $errmsg;
 980                $err =~ s/\%s/$E->cmdline()/ge;
 981                $err =~ s/\%d/$E->value()/ge;
 982                # We can't croak here since Error.pm would mangle
 983                # that to Error::Simple.
 984        };
 985        $err and croak $err;
 986        return $array ? @result : $result[0];
 987}
 988
 989
 990=back
 991
 992=head1 COPYRIGHT
 993
 994Copyright 2006 by Petr Baudis E<lt>pasky@suse.czE<gt>.
 995
 996This module is free software; it may be used, copied, modified
 997and distributed under the terms of the GNU General Public Licence,
 998either version 2, or (at your option) any later version.
 999
1000=cut
1001
1002
1003# Take raw method argument list and return ($obj, @args) in case
1004# the method was called upon an instance and (undef, @args) if
1005# it was called directly.
1006sub _maybe_self {
1007        # This breaks inheritance. Oh well.
1008        ref $_[0] eq 'Git' ? @_ : (undef, @_);
1009}
1010
1011# Check if the command id is something reasonable.
1012sub _check_valid_cmd {
1013        my ($cmd) = @_;
1014        $cmd =~ /^[a-z0-9A-Z_-]+$/ or throw Error::Simple("bad command: $cmd");
1015}
1016
1017# Common backend for the pipe creators.
1018sub _command_common_pipe {
1019        my $direction = shift;
1020        my ($self, @p) = _maybe_self(@_);
1021        my (%opts, $cmd, @args);
1022        if (ref $p[0]) {
1023                ($cmd, @args) = @{shift @p};
1024                %opts = ref $p[0] ? %{$p[0]} : @p;
1025        } else {
1026                ($cmd, @args) = @p;
1027        }
1028        _check_valid_cmd($cmd);
1029
1030        my $fh;
1031        if ($^O eq 'MSWin32') {
1032                # ActiveState Perl
1033                #defined $opts{STDERR} and
1034                #       warn 'ignoring STDERR option - running w/ ActiveState';
1035                $direction eq '-|' or
1036                        die 'input pipe for ActiveState not implemented';
1037                # the strange construction with *ACPIPE is just to
1038                # explain the tie below that we want to bind to
1039                # a handle class, not scalar. It is not known if
1040                # it is something specific to ActiveState Perl or
1041                # just a Perl quirk.
1042                tie (*ACPIPE, 'Git::activestate_pipe', $cmd, @args);
1043                $fh = *ACPIPE;
1044
1045        } else {
1046                my $pid = open($fh, $direction);
1047                if (not defined $pid) {
1048                        throw Error::Simple("open failed: $!");
1049                } elsif ($pid == 0) {
1050                        if (defined $opts{STDERR}) {
1051                                close STDERR;
1052                        }
1053                        if ($opts{STDERR}) {
1054                                open (STDERR, '>&', $opts{STDERR})
1055                                        or die "dup failed: $!";
1056                        }
1057                        _cmd_exec($self, $cmd, @args);
1058                }
1059        }
1060        return wantarray ? ($fh, join(' ', $cmd, @args)) : $fh;
1061}
1062
1063# When already in the subprocess, set up the appropriate state
1064# for the given repository and execute the git command.
1065sub _cmd_exec {
1066        my ($self, @args) = @_;
1067        if ($self) {
1068                $self->repo_path() and $ENV{'GIT_DIR'} = $self->repo_path();
1069                $self->wc_path() and chdir($self->wc_path());
1070                $self->wc_subdir() and chdir($self->wc_subdir());
1071        }
1072        _execv_git_cmd(@args);
1073        die qq[exec "@args" failed: $!];
1074}
1075
1076# Execute the given Git command ($_[0]) with arguments ($_[1..])
1077# by searching for it at proper places.
1078sub _execv_git_cmd { exec('git', @_); }
1079
1080# Close pipe to a subprocess.
1081sub _cmd_close {
1082        my ($fh, $ctx) = @_;
1083        if (not close $fh) {
1084                if ($!) {
1085                        # It's just close, no point in fatalities
1086                        carp "error closing pipe: $!";
1087                } elsif ($? >> 8) {
1088                        # The caller should pepper this.
1089                        throw Git::Error::Command($ctx, $? >> 8);
1090                }
1091                # else we might e.g. closed a live stream; the command
1092                # dying of SIGPIPE would drive us here.
1093        }
1094}
1095
1096
1097sub DESTROY {
1098        my ($self) = @_;
1099        $self->_close_hash_and_insert_object();
1100        $self->_close_cat_blob();
1101}
1102
1103
1104# Pipe implementation for ActiveState Perl.
1105
1106package Git::activestate_pipe;
1107use strict;
1108
1109sub TIEHANDLE {
1110        my ($class, @params) = @_;
1111        # FIXME: This is probably horrible idea and the thing will explode
1112        # at the moment you give it arguments that require some quoting,
1113        # but I have no ActiveState clue... --pasky
1114        # Let's just hope ActiveState Perl does at least the quoting
1115        # correctly.
1116        my @data = qx{git @params};
1117        bless { i => 0, data => \@data }, $class;
1118}
1119
1120sub READLINE {
1121        my $self = shift;
1122        if ($self->{i} >= scalar @{$self->{data}}) {
1123                return undef;
1124        }
1125        my $i = $self->{i};
1126        if (wantarray) {
1127                $self->{i} = $#{$self->{'data'}} + 1;
1128                return splice(@{$self->{'data'}}, $i);
1129        }
1130        $self->{i} = $i + 1;
1131        return $self->{'data'}->[ $i ];
1132}
1133
1134sub CLOSE {
1135        my $self = shift;
1136        delete $self->{data};
1137        delete $self->{i};
1138}
1139
1140sub EOF {
1141        my $self = shift;
1142        return ($self->{i} >= scalar @{$self->{data}});
1143}
1144
1145
11461; # Famous last words