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]+)/) ? "#n$1" : '';
55 my $x = qq(@@ <a\nhref="$spfx$oid_a/s/$dctx->{Q}$n">$ca</a>);
57 $n = ($cb =~ /^\+([0-9]+)/) ? "#n$1" : '';
58 $x .= qq( <a\nhref="$spfx$oid_b/s/$dctx->{Q}$n">$cb</a> @@);
65 my ($dctx, $spfx, $oid) = @_;
66 defined($spfx) ? qq(<a\nhref="$spfx$oid/s/$dctx->{Q}">$oid</a>) : $oid;
69 # returns true if diffstat anchor written, false otherwise
71 my ($ctx, $fn, $rest) = @_;
75 # normal git diffstat output is impossible to parse reliably
76 # without --numstat, and that isn't the default for format-patch.
77 # So only do best-effort handling of renames for common cases;
78 # which works well in practice. If projects put "=>", or trailing
79 # spaces in filenames, oh well :P
80 $fn =~ s/$DIFFSTAT_COMMENT//;
81 $fn =~ s/\{(?:.+) => (.+)\}/$1/ or $fn =~ s/.* => (.+)/$1/;
82 $fn = git_unquote($fn);
84 # long filenames will require us to check in anchor1()
85 push(@{$ctx->{-long_path}}, $fn) if $fn =~ s!\A\.\.\./?!!;
87 my $attr = to_attr($ctx->{-apfx}.$fn) // return;
88 $ctx->{-anchors}->{$attr} = 1;
89 my $spaces = ($orig =~ s/( +)\z//) ? $1 : '';
90 print { $ctx->{zfh} } " <a\nid=i$attr\nhref=#$attr>",
91 ascii_html($orig), '</a>', $spaces,
92 $ctx->{-linkify}->to_html($rest);
95 # returns "diff --git" anchor destination, undef otherwise
98 my $attr = to_attr($ctx->{-apfx}.$pb) // return;
100 my $ok = delete $ctx->{-anchors}->{$attr};
102 # unlikely, check the end of long path names we captured,
103 # assume diffstat and diff output follow the same order,
104 # and ignore different ordering (could be malicious input)
106 my $fn = shift(@{$ctx->{-long_path}}) // return;
107 $pb =~ /\Q$fn\E\z/s or return;
108 $attr = to_attr($ctx->{-apfx}.$fn) // return;
109 $ok = delete $ctx->{-anchors}->{$attr} // return;
111 $ok ? "<a\nhref=#i$attr\nid=$attr>diff</a> --git" : undef
114 sub diff_header ($$$) {
115 my ($x, $ctx, $top) = @_;
116 my (undef, undef, $pa, $pb) = splice(@$top, 0, 4); # ignore oid_{a,b}
117 my $spfx = $ctx->{-spfx};
118 my $dctx = { spfx => $spfx };
120 # get rid of leading "a/" or "b/" (or whatever --{src,dst}-prefix are)
121 $pa = (split(m'/', git_unquote($pa), 2))[1] if $pa ne '/dev/null';
122 $pb = (split(m'/', git_unquote($pb), 2))[1] if $pb ne '/dev/null';
123 if ($pa eq $pb && $pb ne '/dev/null') {
124 $dctx->{Q} = '?b='.uri_escape_path($pb);
127 if ($pb ne '/dev/null') {
128 push @q, 'b='.uri_escape_path($pb);
130 if ($pa ne '/dev/null') {
131 push @q, 'a='.uri_escape_path($pa);
133 $dctx->{Q} = '?'.join('&', @q);
136 # linkify early and all at once, since we know the following
137 # subst ops on $$x won't need further escaping:
138 $$x = $ctx->{-linkify}->to_html($$x);
140 # no need to capture oid_a and oid_b on add/delete,
141 # we just linkify OIDs directly via s///e in conditional
142 if ($$x =~ s/$NULL_TO_BLOB/$1 . oid($dctx, $spfx, $2)/e) {
143 push @{$ctx->{-qry}->{dfpost}}, $2;
144 } elsif ($$x =~ s/$BLOB_TO_NULL/'index '.oid($dctx, $spfx, $1).$2/e) {
145 push @{$ctx->{-qry}->{dfpre}}, $1;
146 } elsif ($$x =~ $BLOB_TO_BLOB) {
147 # modification-only, not add/delete:
148 # linkify hunk headers later using oid_a and oid_b
149 @$dctx{qw(oid_a oid_b)} = ($1, $2);
150 push @{$ctx->{-qry}->{dfpre}}, $1;
151 push @{$ctx->{-qry}->{dfpost}}, $2;
153 warn "BUG? <$$x> had no ^index line";
155 $$x =~ s!^diff --git!anchor1($ctx, $pb) // 'diff --git'!ems;
156 print { $ctx->{zfh} } qq(<span\nclass="head">), $$x, '</span>';
160 sub diff_before_or_after ($$) {
162 if (exists $ctx->{-anchors} && $$x =~ /\A(.*?) # likely "---\n" # \$1
164 ((?:^\x20(?:[^\n]+?)(?:\x20+\|\x20[^\n]*\n))+)
165 (\x20[0-9]+\x20files?\x20)changed,
166 (.*?)\z/msx) { # notes, commit message, etc
167 my @x = ($4, $3, $2, $1);
169 my $lnk = $ctx->{-linkify};
170 my $zfh = $ctx->{zfh};
171 print $zfh $lnk->to_html(pop @x); # $1 uninteresting prefix
172 for my $l (split(/^/m, pop(@x))) { # $2 per-file stat lines
173 $l =~ /^ (.+)( +\| .*\z)/s and
174 anchor0($ctx, $1, $2) and next;
175 print $zfh $lnk->to_html($l);
177 my $ch = $ctx->{changed_href} // '#related';
178 print $zfh pop(@x), # $3 /^ \d+ files? /
179 qq(<a href="$ch">changed</a>,),
180 # insertions/deletions, notes, commit message, etc:
183 print { $ctx->{zfh} } $ctx->{-linkify}->to_html($$x);
187 # callers must do CRLF => LF conversion before calling this
188 sub flush_diff ($$) {
189 my ($ctx, $cur) = @_;
191 my @top = split($EXTRACT_DIFFS, $$cur);
192 undef $$cur; # free memory
194 my $lnk = $ctx->{-linkify};
195 my $dctx; # {}, keys: Q, oid_a, oid_b
198 while (defined(my $x = shift @top)) {
199 if (scalar(@top) >= 4 &&
200 $top[1] =~ $IS_OID &&
201 $top[0] =~ $IS_OID) {
202 $dctx = diff_header(\$x, $ctx, \@top);
204 open(my $afh, '>>', \(my $after='')) or die "open: $!";
206 # Quiet "Complex regular subexpression recursion limit"
207 # warning. Perl will truncate matches upon hitting
208 # that limit, giving us more (and shorter) scalars than
209 # would be ideal, but otherwise it's harmless.
211 # We could replace the `+' metacharacter with `{1,100}'
212 # to limit the matches ourselves to 100, but we can
213 # let Perl do it for us, quietly.
214 no warnings 'regexp';
216 for my $s (split(/((?:(?:^\+[^\n]*\n)+)|
218 (?:^@@ [^\n]+\n))/xsm, $x)) {
219 if (!defined($dctx)) {
221 } elsif ($s =~ s/\A@@ (\S+) (\S+) @@//) {
222 print $zfh qq(<span\nclass="hunk">),
223 diff_hunk($dctx, $1, $2),
226 } elsif ($s =~ /\A\+/) { # $s may be huge
227 print $zfh qq(<span\nclass="add">),
230 } elsif ($s =~ /\A-- $/sm) { # email sig starts
233 } elsif ($s =~ /\A-/) { # $s may be huge
234 print $zfh qq(<span\nclass="del">),
237 } else { # $s may be huge
238 print $zfh $lnk->to_html($s);
241 diff_before_or_after($ctx, \$after) if !$dctx;
243 diff_before_or_after($ctx, \$x);