viewvcs: load blobs asynchronously

This actually leads to a nice 3-5% speedup under parallel loads
when using git(1) w/o SHA-1 collision detection enabled.  Gcf2
is slower since libgit2 has SHA-1 collision detection enabled
on my system.

Since we're in the area, improve location of comments w.r.t.
cgit CSS class names and note the reliance on scratchpad for
performance in a tight loop.

1 files changed, 16 insertions, 5 deletions

diff --git a/lib/PublicInbox/ViewVCS.pm b/lib/PublicInbox/ViewVCS.pm
index f740591d..915cf2c5 100644
--- a/lib/PublicInbox/ViewVCS.pm
+++ b/lib/PublicInbox/ViewVCS.pm
@@ -413,11 +413,21 @@ sub solve_result {
 blob $oid $size bytes $raw_link</pre>
 EOM
         }
+        @{$ctx->{-paths}} = ($path, $raw_link);
+        $ctx->{git} = $git;
+        if ($ctx->{env}->{'pi-httpd.async'}) {
+                ibx_async_cat($ctx, $oid, \&show_blob, $ctx);
+        } else { # synchronous
+                $git->cat_async($oid, \&show_blob, $ctx);
+                $git->cat_async_wait;
+        }
+}
 
-        my $blob = $git->cat_file($oid);
-        if (!$blob) { # WTF?
+sub show_blob { # git->cat_async callback
+        my ($blob, $oid, $type, $size, $ctx) = @_;
+        if (!$blob) {
                 my $e = "Failed to retrieve generated blob ($oid)";
-                warn "$e ($git->{git_dir})";
+                warn "$e ($ctx->{git}->{git_dir}) type=$type";
                 return html_page($ctx, 500, "<pre><b>$e</b></pre>".dbg_log($ctx))
         }
 
@@ -428,6 +438,7 @@ EOM
                 return delete($ctx->{-wcb})->([200, $h, [ $$blob ]]);
         }
 
+        my ($path, $raw_link) = @{delete $ctx->{-paths}};
         $bin and return html_page($ctx, 200,
                                 "<pre>blob $oid $size bytes (binary)" .
                                 " $raw_link</pre>".dbg_log($ctx));
@@ -445,14 +456,14 @@ EOM
                 $$blob = ascii_html($$blob);
         }
 
+        # using some of the same CSS class names and ids as cgit
         my $x = "<pre>blob $oid $size bytes $raw_link</pre>" .
                 "<hr /><table\nclass=blob>".
                 "<tr><td\nclass=linenumbers><pre>";
+        # scratchpad in this loop is faster here than `printf $zfh':
         $x .= sprintf("<a id=n$_ href=#n$_>% ${pad}u</a>\n", $_) for (1..$nl);
         $x .= '</pre></td><td><pre> </pre></td>'. # pad for non-CSS users
                 "<td\nclass=lines><pre\nstyle='white-space:pre'><code>";
-
-        # using some of the same CSS class names and ids as cgit
         html_page($ctx, 200, $x, $ctx->{-linkify}->linkify_2($$blob),
                 '</code></pre></td></tr></table>'.dbg_log($ctx));
 }