File Coverage

git-autofixup

Criterion	Covered	Total	%
statement	284	346	82.0
branch	99	156	63.4
condition	27	36	75.0
subroutine	28	31	90.3
pod			n/a
total	438	569	76.9

#!/usr/bin/perl
package main;
use 5.008004;
use strict;
use warnings FATAL => 'all';

use Carp qw(croak);
use Pod::Usage;
use Getopt::Long qw(:config bundling);
use File::Temp;
use File::Spec ();

our $VERSION = 0.003000; # X.YYYZZZ

my $VERBOSE;
my @GIT_OPTIONS;

# Strictness levels.
my ($CONTEXT, $ADJACENT, $SURROUNDED) = (0..10);

my $usage =<<'END';
usage: git-autofixup []

-h                 show usage
--help             show manpage
--version          show version
-v, --verbose      increase verbosity (use up to 2 times)
-c N, --context N  set number of diff context lines (default 3)
-e, --exit-code    use more detailed exit codes (see --help)
-s N, --strict N   set strictness (default 0)
    Assign a hunk to fixup a topic branch commit if:

    0: either only one topic branch commit is blamed in the hunk context or
       blocks of added lines are adjacent to exactly one topic branch commit.
       Removing upstream lines is allowed for this level.
    1: blocks of added lines are adjacent to exactly one topic branch commit
    2: blocks of added lines are surrounded by exactly one topic branch commit

    Regardless of strictness level, removed lines are correlated with the
    commit they're blamed on, and all the blocks of changed lines in a hunk
    must be correlated with the same topic branch commit in order to be
    assigned to it. See the --help for more details.
-g ARG, --gitopt ARG
    Specify option for git. Can be used multiple times.
END

# Parse hunks out of `git diff` output. Return an array of hunk hashrefs.
sub parse_hunks {
    my $fh = shift;
    my ($file_a, $file_b);
    my @hunks;
    my $line;
    while ($line = <$fh>) {
        if ($line =~ /^--- (.*)/) {
            $file_a = $1;
        } elsif ($line =~ /^\+\+\+ (.*)/) {
            $file_b = $1;
        } elsif ($line =~ /^@@ -(\d+)(?:,(\d+))? \+\d+(?:,\d+)? @@/) {
            my $header = $line;

            for ($file_a, $file_b) {
                s#^[abiwco]/##;
            }

            next if $file_a ne $file_b; # Ignore creations and deletions.

            my $lines = [];
            while (1) {
                $line = <$fh>;
                if (!defined($line) || $line =~ /^[^ +\\-]/) {
                    last;
                }
                push @{$lines}, $line;
            }

            push(@hunks, {
                file => $file_a,
                start => $1,
                count => defined($2) ? $2 : 1,
                header => $header,
                lines => $lines,
            });
            # The next line after a hunk could be a header for the next commit
            # or hunk.
            redo if defined $line;
        }
    }
    return @hunks;
}

sub git_cmd {
    return ('git', @GIT_OPTIONS, @_);
}

sub summary_for_commits {
    my $rev = shift;
    my %commits;
    for (qx(git log --no-merges --format=%H:%s $rev..)) {
        chomp;
        my ($sha, $msg) = split ':', $_, 2;
        $commits{$sha} = $msg;
    }
    return \%commits;
}

# Return targets of fixup!/squash! commits.
sub sha_aliases {
    my $summary_for = shift;
    my %aliases;
    my @targets = keys(%{$summary_for});
    for my $sha (@targets) {
        my $summary = $summary_for->{$sha};
        next if $summary !~ /^(?:fixup|squash)! (.*)/;
        my $prefix = $1;
        if ($prefix =~ /^(?:(?:fixup|squash)! ){2}/) {
            die "fixup commits for fixup commits aren't supported: $sha";
        }
        my @matches = grep {startswith($summary_for->{$_}, $prefix)} @targets;
        if (@matches > 1) {
            die "ambiguous fixup commit target: multiple commit summaries start with: $prefix\n";
        } elsif (@matches == 0) {
            die "no fixup target in topic branch: $sha\n";
        } elsif (@matches == 1) {
            $aliases{$sha} = $matches[0];
        }
    }
    return \%aliases;
}

sub fixup_sha {
    my $args = shift;
    my $hunk = $args->{hunk};
    my $blame = $args->{blame};
    my $summary_for = $args->{summary_for};
    my $strict = $args->{strict};
    if (grep {!defined} ($hunk, $blame, $summary_for, $strict)) {
        croak 'missing argument';
    }

    my @targets;
    if ($args->{strict} == $CONTEXT) {
        @targets = fixup_targets_from_all_context($args);
        my @topic_targets = grep {defined $summary_for->{$_}} @targets;
        if (@topic_targets > 1) {
            # The context assignment is ambiguous, but an adjacency assignment
            # might not be.
            @targets = fixup_targets_from_adjacent_context($args);
        }
    } else {
        @targets = fixup_targets_from_adjacent_context($args);
    }

    my $upstream_is_blamed = grep {!defined $summary_for->{$_}} @targets;
    my @topic_targets = grep {defined $summary_for->{$_}} @targets;
    if ($strict && $upstream_is_blamed) {
        $VERBOSE && print hunk_desc($hunk), " changes lines blamed on upstream\n";
        return;
    } elsif (@topic_targets > 1) {
        $VERBOSE && print hunk_desc($hunk), " has multiple targets\n";
        return;
    } elsif (@topic_targets == 0) {
        $VERBOSE && print hunk_desc($hunk), " has no targets\n";
        return;
    }
    return $topic_targets[0];
}

sub hunk_desc {
    my $hunk = shift;
    return join " ", (
        $hunk->{file},
        $hunk->{header} =~ /(@@[^@]*@@)/,
    );
}

sub fixup_targets_from_all_context {
    my $args = shift;
    my ($hunk, $blame, $summary_for) = @{$args}{qw(hunk blame summary_for)};
    croak 'missing argument' if grep {!defined} ($hunk, $blame, $summary_for);

    my @targets = uniq(map {$_->{sha}} values(%{$blame}));
    return wantarray ? @targets : \@targets;
}

sub uniq {
    my %seen;
    return grep {!$seen{$_}++} @_;
}

sub fixup_targets_from_adjacent_context {
    my $args = shift;
    my $hunk = $args->{hunk};
    my $blame = $args->{blame};
    my $summary_for = $args->{summary_for};
    my $strict = $args->{strict};
    if (grep {!defined} ($hunk, $blame, $summary_for, $strict)) {
        croak 'missing argument';
    }

    my $blame_indexes = blame_indexes($hunk);

    my %blamed;
    my $diff = $hunk->{lines};
    for (my $di = 0; $di < @{$diff}; $di++) { # diff index
        my $bi = $blame_indexes->[$di];
        my $line = $diff->[$di];
        if (startswith($line, '-')) {
            my $sha = $blame->{$bi}{sha};
            $blamed{$sha} = 1;
        } elsif (startswith($line, '+')) {
            my @lines;
            if ($di > 0 && defined $blame->{$bi-1}) {
                push @lines, $bi-1;
            }
            if (defined $blame->{$bi}) {
                push @lines, $bi;
            }
            my @adjacent_shas = uniq(map {$_->{sha}} @{$blame}{@lines});
            my @target_shas = grep {defined $summary_for->{$_}} @adjacent_shas;
            # Note that lines at the beginning or end of a file can be
            # "surrounded" by a single line.
            my $is_surrounded = @target_shas > 0
                && @target_shas == @adjacent_shas
                && $target_shas[0] eq $target_shas[-1];
            my $is_adjacent = @target_shas == 1;
            if ($is_surrounded || ($strict < $SURROUNDED && $is_adjacent)) {
                $blamed{$target_shas[0]} = 1;
            }
            while ($di < @$diff-1 && startswith($diff->[$di+1], '+')) {
                $di++;
            }
        }
    }
    my @targets = keys %blamed;
    return wantarray ? @targets : \@targets;
}

sub startswith {
    my ($haystack, $needle) = @_;
    return index($haystack, $needle, 0) == 0;
}

# Map lines in a hunk's diff to the corresponding `git blame HEAD` output.
sub blame_indexes {
    my $hunk = shift;
    my @indexes;
    my $bi = $hunk->{start};
    for (my $di = 0; $di < @{$hunk->{lines}}; $di++) {
        push @indexes, $bi;
        my $first = substr($hunk->{lines}[$di], 0, 1);
        if ($first eq '-' or $first eq ' ') {
            $bi++;
        }
        # Don't increment $bi for added lines.
    }
    return \@indexes;
}

sub print_hunk_blamediff {
    my $args = shift;
    my $fh = $args->{fh};
    my $hunk = $args->{hunk};
    my $summary_for = $args->{summary_for};
    my $blame = $args->{blame};
    my $blame_indexes = $args->{blame_indexes};
    if (grep {!defined} ($fh, $hunk, $summary_for, $blame, $blame_indexes)) {
        croak 'missing argument';
    }

    my $format = "%-8.8s|%4.4s|%-30.30s|%-30.30s\n";
    for (my $i = 0; $i < @{$hunk->{lines}}; $i++) {
        my $line_r = $hunk->{lines}[$i];
        my $bi = $blame_indexes->[$i];
        my $sha = defined $blame->{$bi} ? $blame->{$bi}{sha} : undef;

        my $display_sha = defined($sha) ? $sha : q{};
        my $display_bi = $bi;
        if (startswith($line_r, '+')) {
            $display_sha = q{}; # For added lines.
            $display_bi = q{};
        }
        if (defined($sha) && !defined($summary_for->{$sha})) {
            # For lines from before the given upstream revision.
            $display_sha = '^';
        }

        my $line_l = '';
        if (defined $blame->{$bi} && !startswith($line_r, '+')) {
            $line_l = $blame->{$bi}{text};
        }

        for ($line_l, $line_r) {
            # For the table to line up, tabs need to be converted to a string of fixed width.
            s/\t/^I/g;
            # Remove trailing newlines and carriage returns. If more trailing
            # whitespace is removed, that's fine.
            $_ = rtrim($_);
        }

        printf {$fh} $format, $display_sha, $display_bi, $line_l, $line_r;
    }
    print {$fh} "\n";
    return;
}

sub rtrim {
    my $s = shift;
    $s =~ s/\s+\z//;
    return $s;
}

sub blame {
    my ($hunk, $alias_for) = @_;
    if ($hunk->{count} == 0) {
        return {};
    }
    my @cmd = git_cmd(
        'blame', '--porcelain',
        '-L' => "$hunk->{start},+$hunk->{count}",
        'HEAD', '--',
        "$hunk->{file}");
    my %blame;
    my ($sha, $line_num);
    open(my $fh, '-|', @cmd) or die "git blame: $!\n";
    while (my $line = <$fh>) {
        if ($line =~ /^([0-9a-f]{40}) \d+ (\d+)/) {
             ($sha, $line_num) = ($1, $2);
        }
        if (startswith($line, "\t")) {
            if (defined $alias_for->{$sha}) {
                $sha = $alias_for->{$sha};
            }
            $blame{$line_num} = {sha => $sha, text => substr($line, 1)};
        }
    }
    close($fh) or die "git blame: non-zero exit code";
    return \%blame;
}

sub diff_hunks {
    my $num_context_lines = shift;
    my @cmd = git_cmd(qw(-c diff.noprefix=false diff --no-ext-diff --ignore-submodules), "-U$num_context_lines");
    if (is_index_dirty()) {
        push @cmd, "--cached";
    }
    open(my $fh, '-|', @cmd) or die $!;
    my @hunks = parse_hunks($fh, keep_lines => 1);
    close($fh) or die "git diff: non-zero exit code";
    return wantarray ? @hunks : \@hunks;
}

sub commit_fixup {
    my ($sha, $hunks) = @_;
    open my $fh, '|-', git_cmd(qw(apply --unidiff-zero --cached -)) or die "git apply: $!\n";
    for my $hunk (@{$hunks}) {
        print({$fh}
            "--- a/$hunk->{file}\n",
            "+++ b/$hunk->{file}\n",
            $hunk->{header},
            @{$hunk->{lines}},
        );
    }
    close $fh or die "git apply: non-zero exit code\n";
    system(git_cmd('commit', "--fixup=$sha")) == 0 or die "git commit: $!\n";
    return;
}

sub is_index_dirty {
    open(my $fh, '-|', git_cmd(qw(status --porcelain))) or die "git status: $!\n";
    my $dirty;
    while (my $line = <$fh>) {
        if ($line =~ /^[^?! ]/) {
            $dirty = 1;
            last;
        }
    }
    close $fh or die "git status: non-zero exit code\n";
    return $dirty;
}

sub fixup_hunks_by_sha {
    my $args = shift;
    my $hunks = $args->{hunks};
    my $blame_for = $args->{blame_for};
    my $summary_for = $args->{summary_for};
    my $strict = $args->{strict};
    if (grep {!defined} ($hunks, $blame_for, $summary_for, $strict)) {
        croak 'missing argument';
    }

    my %hunks_for;
    for my $hunk (@{$hunks}) {
        my $blame = $blame_for->{$hunk};
        my $sha = fixup_sha({
            hunk => $hunk,
            blame => $blame,
            summary_for => $summary_for,
            strict => $strict,
        });
        if ($sha && $VERBOSE) {
            printf "%s fixes %s %s\n",
                hunk_desc($hunk),
                substr($sha, 0, 8),
                $summary_for->{$sha};
        }
        if ($VERBOSE > 1) {
            print_hunk_blamediff({
                fh => *STDOUT,
                hunk => $hunk,
                summary_for => $summary_for,
                blame => $blame,
                blame_indexes => blame_indexes($hunk)
            });
        }
        next if !$sha;
        push @{$hunks_for{$sha}}, $hunk;
    }
    return \%hunks_for;
}

# Return SHAs in some consistent order.
#
# Currently they're ordered by how early their assigned hunks appear in the
# diff output. This assumes $hunks is in the order it was parsed from the diff.
# This ordering seems nice since it'd be similar to the order a human would
# make commits in if they were working their way down the diff.
sub ordered_shas {
    my $hunks = shift;
    my $sha_for = shift;
    my @ordered = ();
    for my $hunk (@{$hunks}) {
        if (defined $sha_for->{$hunk}) {
            push @ordered, $sha_for->{$hunk};
        }
    }
    return uniq(@ordered);
}

# Reverse the sha->hunks hashef and return a hunk->sha hashref.
sub sha_for_hunk_map {
    my $hunks_for = shift;
    my %sha_for;
    for my $sha (keys %{$hunks_for}) {
        for my $hunk (@{$hunks_for->{$sha}}) {
            if (defined $sha_for{$hunk}) {
                die "multiple SHAs for hunk";  # This should never happen.
            }
            $sha_for{$hunk} = $sha;
        }
    }
    return \%sha_for;
}

sub exit_code {
    my ($hunks, $hunks_for) = @_;
    my $hunk_count = scalar @{$hunks};

    my $assigned_hunk_count = 0;
    for (values %{$hunks_for}) {
        $assigned_hunk_count += @{$_};
    }

    my $rc;
    if ($hunk_count == 0) {
        $rc = 3;  # no hunks to assign
    } elsif ($assigned_hunk_count == 0) {
        $rc = 2;  # hunks exist, but none assigned
    } elsif ($assigned_hunk_count < $hunk_count) {
        $rc = 1;  # not all hunks assigned
    } elsif ($hunk_count == $assigned_hunk_count) {
        $rc = 0;  # all hunks assigned
    } else {
        die "unexpected conditions when choosing exit code";
    }
    return $rc;
}

# Create a temporary index so we can craft commits with already-staged hunks.
# Return a File::Temp object so the caller has control over its lifetime.
sub create_temp_index {
    my $tempfile = File::Temp->new(
        TEMPLATE => 'git-autofixup_index.XXXXXX',
        DIR => File::Spec->tmpdir());
    $ENV{GIT_INDEX_FILE} = $tempfile->filename();
    # A blank index makes it look like we're deleting everything, so read
    # HEAD's tree into it.
    qx(git read-tree HEAD^{tree});
    $? == 0 or die "Can't read HEAD's tree into temp index.\n";
    return $tempfile;
}

sub main {
    $VERBOSE = 0;
    my $help;
    my $man;
    my $show_version;
    my $strict = $CONTEXT;
    my $num_context_lines = 3;
    my $dryrun;
    my $use_detailed_exit_codes;

    GetOptions(
        'h' => \$help,
        'help' => \$man,
        'version' => \$show_version,
        'verbose|v+' => \$VERBOSE,
        'strict|s=i' => \$strict,
        'context|c=i' => \$num_context_lines,
        'dryrun|n' => \$dryrun,
        'gitopt|g=s' => \@GIT_OPTIONS,
        'exit-code' => \$use_detailed_exit_codes,
    ) or return 1;
    if ($help) {
        print $usage;
        return 0;
    }
    if ($show_version) {
        print "$VERSION\n";
        return 0;
    }
    if ($man) {
        pod2usage(-exitval => 0, -verbose => 2);
    }

    @ARGV == 1 or die "No upstream commit given.\n";
    my $upstream = shift @ARGV;
    qx(git rev-parse --verify ${upstream}^{commit});
    $? == 0 or die "Can't resolve given commit.\n";

    if ($num_context_lines < 0) {
        die "invalid number of context lines: $num_context_lines\n";
    }

    if ($strict < 0) {
        die "invalid strictness level: $strict\n";
    } elsif ($strict > 0 && $num_context_lines == 0) {
        die "strict hunk assignment requires context\n";
    }


    my $toplevel = qx(git rev-parse --show-toplevel);
    chomp $toplevel;
    $? == 0 or die "Can't get repo toplevel dir\n";
    chdir $toplevel or die $!;

    my $hunks = diff_hunks($num_context_lines);
    my $summary_for = summary_for_commits($upstream);
    my $alias_for = sha_aliases($summary_for);
    my %blame_for = map {$_ => blame($_, $alias_for)} @{$hunks};
    my $hunks_for = fixup_hunks_by_sha({
        hunks => $hunks,
        blame_for => \%blame_for,
        summary_for => $summary_for,
        strict => $strict,
    });
    my @ordered_shas = ordered_shas($hunks, sha_for_hunk_map($hunks_for));

    if ($dryrun) {
        if ($use_detailed_exit_codes) {
            return exit_code($hunks, $hunks_for);
        }
        return 0;
    }

    # Limit the tempfile's lifetime to the execution of main().
    local $ENV{GIT_INDEX_FILE};  # Throw away ENV changes between main() calls.
    my $tempfile = create_temp_index();

    for my $sha (@ordered_shas) {
        my $fixup_hunks = $hunks_for->{$sha};
        commit_fixup($sha, $fixup_hunks);
    }
    if ($use_detailed_exit_codes) {
        return exit_code($hunks, $hunks_for);
    }
    return 0;
}

if (!caller()) {
    exit main();
}
1;

__END__