1 # Copyright (C) all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # used by PublicInbox::View
5 # This adds CSS spans for diff highlighting.
6 # It also generates links for ViewVCS + SolverGit to show
7 # (or reconstruct) blobs.
9 package PublicInbox::ViewDiff;
12 use parent qw(Exporter);
13 our @EXPORT_OK = qw(flush_diff uri_escape_path);
14 use URI::Escape qw(uri_escape_utf8);
15 use PublicInbox::Hval qw(ascii_html to_attr);
16 use PublicInbox::Git qw(git_unquote);
18 my $OID_NULL = '0{7,}';
19 my $OID_BLOB = '[a-f0-9]{7,}';
22 my $FN = qr!(?:"?[^/\n]+/[^\n]+|/dev/null)!;
24 # cf. git diff.c :: get_compact_summary
25 my $DIFFSTAT_COMMENT =
26 qr/(?: *\((?:new|gone|(?:(?:new|mode) [\+\-][lx]))\))? *\z/s;
27 my $NULL_TO_BLOB = qr/^(index $OID_NULL\.\.)($OID_BLOB)\b/ms;
28 my $BLOB_TO_NULL = qr/^index ($OID_BLOB)(\.\.$OID_NULL)\b/ms;
29 my $BLOB_TO_BLOB = qr/^index ($OID_BLOB)\.\.($OID_BLOB)/ms;
30 our $EXTRACT_DIFFS = qr/(
31 (?: # begin header stuff, don't capture filenames, here,
32 # but instead wait for the --- and +++ lines.
33 (?:^diff\x20--git\x20$FN\x20$FN$LF)
35 # old mode || new mode || copy|rename|deleted|...
37 )? # end of optional stuff, everything below is required
38 ^index\x20($OID_BLOB)\.\.($OID_BLOB)$ANY*$LF
40 ^\+{3}\x20($FN)$LF)/msx;
41 our $IS_OID = qr/\A$OID_BLOB\z/s;
44 # '/' + $URI::Escape::Unsafe{RFC3986}
45 uri_escape_utf8($_[0], "^A-Za-z0-9\-\._~/");
48 # link to line numbers in blobs
50 my ($dctx, $ca, $cb) = @_;
51 my ($oid_a, $oid_b, $spfx) = @$dctx{qw(oid_a oid_b spfx)};
53 if (defined($spfx) && defined($oid_a) && defined($oid_b)) {
54 my ($n) = ($ca =~ /^-([0-9]+)/);
55 $n = defined($n) ? "#n$n" : '';
57 my $x = qq(@@ <a\nhref="$spfx$oid_a/s/$dctx->{Q}$n">$ca</a>);
59 ($n) = ($cb =~ /^\+([0-9]+)/);
60 $n = defined($n) ? "#n$n" : '';
61 $x .= qq( <a\nhref="$spfx$oid_b/s/$dctx->{Q}$n">$cb</a> @@);
68 my ($dctx, $spfx, $oid) = @_;
69 defined($spfx) ? qq(<a\nhref="$spfx$oid/s/$dctx->{Q}">$oid</a>) : $oid;
72 # returns true if diffstat anchor written, false otherwise
74 my ($dst, $ctx, $fn, $rest) = @_;
78 # normal git diffstat output is impossible to parse reliably
79 # without --numstat, and that isn't the default for format-patch.
80 # So only do best-effort handling of renames for common cases;
81 # which works well in practice. If projects put "=>", or trailing
82 # spaces in filenames, oh well :P
83 $fn =~ s/$DIFFSTAT_COMMENT//;
84 $fn =~ s/\{(?:.+) => (.+)\}/$1/ or $fn =~ s/.* => (.+)/$1/;
85 $fn = git_unquote($fn);
87 # long filenames will require us to check in anchor1()
88 push(@{$ctx->{-long_path}}, $fn) if $fn =~ s!\A\.\.\./?!!;
90 if (defined(my $attr = to_attr($ctx->{-apfx}.$fn))) {
91 $ctx->{-anchors}->{$attr} = 1;
92 my $spaces = ($orig =~ s/( +)\z//) ? $1 : '';
93 $$dst .= " <a\nid=i$attr\nhref=#$attr>" .
94 ascii_html($orig) . '</a>' . $spaces .
95 $ctx->{-linkify}->to_html($rest);
101 # returns "diff --git" anchor destination, undef otherwise
104 my $attr = to_attr($ctx->{-apfx}.$pb) // return;
106 my $ok = delete $ctx->{-anchors}->{$attr};
108 # unlikely, check the end of long path names we captured,
109 # assume diffstat and diff output follow the same order,
110 # and ignore different ordering (could be malicious input)
112 my $fn = shift(@{$ctx->{-long_path}}) // return;
113 $pb =~ /\Q$fn\E\z/s or return;
114 $attr = to_attr($ctx->{-apfx}.$fn) // return;
115 $ok = delete $ctx->{-anchors}->{$attr} // return;
117 $ok ? "<a\nhref=#i$attr\nid=$attr>diff</a> --git" : undef
120 sub diff_header ($$$) {
121 my ($x, $ctx, $top) = @_;
122 my (undef, undef, $pa, $pb) = splice(@$top, 0, 4); # ignore oid_{a,b}
123 my $spfx = $ctx->{-spfx};
124 my $dctx = { spfx => $spfx };
126 # get rid of leading "a/" or "b/" (or whatever --{src,dst}-prefix are)
127 $pa = (split(m'/', git_unquote($pa), 2))[1] if $pa ne '/dev/null';
128 $pb = (split(m'/', git_unquote($pb), 2))[1] if $pb ne '/dev/null';
129 if ($pa eq $pb && $pb ne '/dev/null') {
130 $dctx->{Q} = '?b='.uri_escape_path($pb);
133 if ($pb ne '/dev/null') {
134 push @q, 'b='.uri_escape_path($pb);
136 if ($pa ne '/dev/null') {
137 push @q, 'a='.uri_escape_path($pa);
139 $dctx->{Q} = '?'.join('&', @q);
142 # linkify early and all at once, since we know the following
143 # subst ops on $$x won't need further escaping:
144 $$x = $ctx->{-linkify}->to_html($$x);
146 # no need to capture oid_a and oid_b on add/delete,
147 # we just linkify OIDs directly via s///e in conditional
148 if ($$x =~ s/$NULL_TO_BLOB/$1 . oid($dctx, $spfx, $2)/e) {
149 push @{$ctx->{-qry}->{dfpost}}, $2;
150 } elsif ($$x =~ s/$BLOB_TO_NULL/'index '.oid($dctx, $spfx, $1).$2/e) {
151 push @{$ctx->{-qry}->{dfpre}}, $1;
152 } elsif ($$x =~ $BLOB_TO_BLOB) {
153 # modification-only, not add/delete:
154 # linkify hunk headers later using oid_a and oid_b
155 @$dctx{qw(oid_a oid_b)} = ($1, $2);
156 push @{$ctx->{-qry}->{dfpre}}, $1;
157 push @{$ctx->{-qry}->{dfpost}}, $2;
159 warn "BUG? <$$x> had no ^index line";
161 $$x =~ s!^diff --git!anchor1($ctx, $pb) // 'diff --git'!ems;
162 my $dst = $ctx->{obuf};
163 $$dst .= qq(<span\nclass="head">);
169 sub diff_before_or_after ($$) {
171 if (exists $ctx->{-anchors} && $$x =~ /\A(.*?) # likely "---\n"
173 ((?:^\x20(?:[^\n]+?)(?:\x20+\|\x20[^\n]*\n))+)
174 (\x20[0-9]+\x20files?\x20)changed,([^\n]+\n)
175 (.*?)\z/msx) { # notes, commit message, etc
176 my @x = ($5, $4, $3, $2, $1);
177 my $lnk = $ctx->{-linkify};
178 $$x = $lnk->to_html(pop @x); # uninteresting prefix
179 for my $l (split(/^/m, pop(@x))) { # per-file diffstat lines
180 $l =~ /^ (.+)( +\| .*\z)/s and
181 anchor0($x, $ctx, $1, $2) and next;
182 $$x .= $lnk->to_html($l);
184 $$x .= pop @x; # $3 /^ \d+ files? /
185 my $ch = $ctx->{changed_href} // '#related';
186 $$x .= qq(<a href="$ch">changed</a>,);
187 $$x .= ascii_html(pop @x); # $4: insertions/deletions
188 $$x .= $lnk->to_html(pop @x); # notes, commit message, etc
190 $ctx->{-linkify}->to_html($$x);
194 # callers must do CRLF => LF conversion before calling this
195 sub flush_diff ($$) {
196 my ($ctx, $cur) = @_;
198 my @top = split($EXTRACT_DIFFS, $$cur);
199 undef $$cur; # free memory
201 my $linkify = $ctx->{-linkify};
202 my $dst = $ctx->{obuf};
203 my $dctx; # {}, keys: Q, oid_a, oid_b
205 while (defined(my $x = shift @top)) {
206 if (scalar(@top) >= 4 &&
207 $top[1] =~ $IS_OID &&
208 $top[0] =~ $IS_OID) {
209 $dctx = diff_header(\$x, $ctx, \@top);
213 # Quiet "Complex regular subexpression recursion limit"
214 # warning. Perl will truncate matches upon hitting
215 # that limit, giving us more (and shorter) scalars than
216 # would be ideal, but otherwise it's harmless.
218 # We could replace the `+' metacharacter with `{1,100}'
219 # to limit the matches ourselves to 100, but we can
220 # let Perl do it for us, quietly.
221 no warnings 'regexp';
223 for my $s (split(/((?:(?:^\+[^\n]*\n)+)|
225 (?:^@@ [^\n]+\n))/xsm, $x)) {
226 if (!defined($dctx)) {
228 } elsif ($s =~ s/\A@@ (\S+) (\S+) @@//) {
229 $$dst .= qq(<span\nclass="hunk">);
230 $$dst .= diff_hunk($dctx, $1, $2);
231 $$dst .= $linkify->to_html($s);
233 } elsif ($s =~ /\A\+/) {
234 $$dst .= qq(<span\nclass="add">);
235 $$dst .= $linkify->to_html($s);
237 } elsif ($s =~ /\A-- $/sm) { # email sig starts
240 } elsif ($s =~ /\A-/) {
241 $$dst .= qq(<span\nclass="del">);
242 $$dst .= $linkify->to_html($s);
245 $$dst .= $linkify->to_html($s);
248 $$dst .= diff_before_or_after($ctx, \$after) if !$dctx;
250 $$dst .= diff_before_or_after($ctx, \$x);