solver: break apart inbox blob retrieval

To avoid hogging the event loop in public-inbox-httpd when
many candidate messages match, we'll separate the steps to
ensure fairness on slow storage.

1 files changed, 83 insertions, 53 deletions

diff --git a/lib/PublicInbox/SolverGit.pm b/lib/PublicInbox/SolverGit.pm
index 12024dbc..c54d6d54 100644
--- a/lib/PublicInbox/SolverGit.pm
+++ b/lib/PublicInbox/SolverGit.pm
@@ -106,11 +106,16 @@ sub solve_existing ($$) {
 
 sub extract_diff ($$) {
         my ($p, $arg) = @_;
-        my ($self, $diffs, $pre, $post, $ibx, $smsg) = @$arg;
+        my ($self, $want, $smsg) = @$arg;
         my ($part) = @$p; # ignore $depth and @idx;
         my $ct = $part->content_type || 'text/plain';
         my ($s, undef) = msg_part_text($part, $ct);
         defined $s or return;
+        my $post = $want->{oid_b};
+        my $pre = $want->{oid_a};
+        if (!defined($pre) || $pre !~ /\A[a-f0-9]+\z/) {
+                $pre = '[a-f0-9]{7}'; # for RE below
+        }
 
         # Email::MIME::Encodings forces QP to be CRLF upon decoding,
         # change it back to LF:
@@ -192,10 +197,10 @@ sub extract_diff ($$) {
         close $tmp or die "close(tmp): $!";
 
         # for debugging/diagnostics:
-        $di->{ibx} = $ibx;
+        $di->{ibx} = $want->{cur_ibx};
         $di->{smsg} = $smsg;
 
-        push @$diffs, $di;
+        push @{$self->{tmp_diffs}}, $di;
 }
 
 sub path_searchable ($) { defined($_[0]) && $_[0] =~ m!\A[\w/\. \-]+\z! }
@@ -207,7 +212,7 @@ sub filename_query ($) {
         join('', map { qq( dfn:"$_") } split(/\.\./, $_[0]));
 }
 
-sub find_extract_diffs ($$$) {
+sub find_smsgs ($$$) {
         my ($self, $ibx, $want) = @_;
         my $srch = $ibx->search or return;
 
@@ -218,8 +223,6 @@ sub find_extract_diffs ($$$) {
         my $pre = $want->{oid_a};
         if (defined $pre && $pre =~ /\A[a-f0-9]+\z/) {
                 $q .= " dfpre:$pre";
-        } else {
-                $pre = '[a-f0-9]{7}'; # for $re below
         }
 
         my $path_b = $want->{path_b};
@@ -231,15 +234,8 @@ sub find_extract_diffs ($$$) {
                         $q .= filename_query($path_a);
                 }
         }
-
         my $mset = $srch->mset($q, { relevance => 1 });
-        my $diffs = [];
-        for my $smsg (@{$srch->mset_to_smsg($ibx, $mset)}) {
-                my $eml = $ibx->smsg_eml($smsg) or next;
-                $eml->each_part(\&extract_diff,
-                                [$self, $diffs, $pre, $post, $ibx, $smsg], 1);
-        }
-        @$diffs ? $diffs : undef;
+        $mset->size ? $srch->mset_to_smsg($ibx, $mset) : undef;
 }
 
 sub update_index_result ($$) {
@@ -264,7 +260,7 @@ sub prepare_index ($) {
         # no index creation for added files
         $oid_a =~ /\A0+\z/ and return next_step($self);
 
-        die "BUG: $oid_a not not found" unless $existing;
+        die "BUG: $oid_a not found" unless $existing;
 
         my $oid_full = $existing->[1];
         my $path_a = $di->{path_a} or die "BUG: path_a missing for $oid_full";
@@ -518,15 +514,78 @@ sub di_url ($$) {
         defined($url) ? "$url$mid/" : "<$mid>";
 }
 
+sub retry_current {
+        # my ($self, $want) = @_;
+        push @{$_[0]->{todo}}, $_[1];
+        goto \&next_step # retry solve_existing
+}
+
+sub try_harder {
+        my ($self, $want) = @_;
+
+        # do we have more inboxes to try?
+        goto \&retry_current if scalar @{$want->{try_ibxs}};
+
+        my $cur_want = $want->{oid_b};
+        if (length($cur_want) > $OID_MIN) { # maybe a shorter OID will work
+                delete $want->{try_ibxs}; # drop empty arrayref
+                chop($cur_want);
+                dbg($self, "retrying $want->{oid_b} as $cur_want");
+                $want->{oid_b} = $cur_want;
+                goto \&retry_current; # retry with shorter abbrev
+        }
+
+        dbg($self, "could not find $cur_want");
+        eval { done($self, undef) };
+        die "E: $@" if $@;
+}
+
 sub resolve_patch ($$) {
         my ($self, $want) = @_;
 
+        my $cur_want = $want->{oid_b};
         if (scalar(@{$self->{patches}}) > $MAX_PATCH) {
                 die "Aborting, too many steps to $self->{oid_want}";
         }
 
+        if (my $msgs = $want->{try_smsgs}) {
+                my $smsg = shift @$msgs;
+                if (my $eml = $want->{cur_ibx}->smsg_eml($smsg)) {
+                        $eml->each_part(\&extract_diff,
+                                        [ $self, $want, $smsg ], 1);
+                }
+
+                # try the remaining smsgs later
+                goto \&retry_current if scalar @$msgs;
+
+                delete $want->{try_smsgs};
+                delete $want->{cur_ibx};
+
+                my $diffs = delete $self->{tmp_diffs};
+                if (scalar @$diffs) {
+                        unshift @{$self->{patches}}, @$diffs;
+                        dbg($self, "found $cur_want in " .  join(" ||\n\t",
+                                map { di_url($self, $_) } @$diffs));
+
+                        # good, we can find a path to the oid we $want, now
+                        # lets see if we need to apply more patches:
+                        my $di = $diffs->[0];
+                        my $src = $di->{oid_a};
+
+                        unless ($src =~ /\A0+\z/) {
+                                # we have to solve it using another oid, fine:
+                                my $job = {
+                                        oid_b => $src,
+                                        path_b => $di->{path_a},
+                                };
+                                push @{$self->{todo}}, $job;
+                        }
+                        goto \&next_step; # onto the next todo item
+                }
+                goto \&try_harder;
+        }
+
         # see if we can find the blob in an existing git repo:
-        my $cur_want = $want->{oid_b};
         if (!$want->{try_ibxs} && $self->{seen_oid}->{$cur_want}++) {
                 die "Loop detected solving $cur_want\n";
         }
@@ -544,10 +603,9 @@ sub resolve_patch ($$) {
                         return;
                 }
                 mark_found($self, $cur_want, $existing);
-                return next_step($self); # onto patch application
+                goto \&next_step; # onto patch application
         } elsif ($existing > 0) {
-                push @{$self->{todo}}, $want;
-                return next_step($self); # retry solve_existing
+                goto \&retry_current;
         } else { # $existing == 0: we may retry if inbox scan (below) fails
                 delete $want->{try_gits};
         }
@@ -555,41 +613,13 @@ sub resolve_patch ($$) {
         # scan through inboxes to look for emails which results in
         # the oid we want:
         my $ibx = shift(@{$want->{try_ibxs}}) or die 'BUG: {try_ibxs} empty';
-        if (my $diffs = find_extract_diffs($self, $ibx, $want)) {
-                unshift @{$self->{patches}}, @$diffs;
-                dbg($self, "found $cur_want in ".
-                        join(" ||\n\t", map { di_url($self, $_) } @$diffs));
-
-                # good, we can find a path to the oid we $want, now
-                # lets see if we need to apply more patches:
-                my $di = $diffs->[0];
-                my $src = $di->{oid_a};
-
-                unless ($src =~ /\A0+\z/) {
-                        # we have to solve it using another oid, fine:
-                        my $job = { oid_b => $src, path_b => $di->{path_a} };
-                        push @{$self->{todo}}, $job;
-                }
-                return next_step($self); # onto the next todo item
-        }
-
-        if (scalar @{$want->{try_ibxs}}) { # do we have more inboxes to try?
-                push @{$self->{todo}}, $want;
-                return next_step($self);
+        if (my $msgs = find_smsgs($self, $ibx, $want)) {
+                $want->{try_smsgs} = $msgs;
+                $want->{cur_ibx} = $ibx;
+                $self->{tmp_diffs} = [];
+                goto \&retry_current;
         }
-
-        if (length($cur_want) > $OID_MIN) { # maybe a shorter OID will work
-                delete $want->{try_ibxs}; # drop empty arrayref
-                chop($cur_want);
-                dbg($self, "retrying $want->{oid_b} as $cur_want");
-                $want->{oid_b} = $cur_want;
-                push @{$self->{todo}}, $want;
-                return next_step($self); # retry with shorter abbrev
-        }
-
-        dbg($self, "could not find $cur_want");
-        eval { done($self, undef) };
-        die "E: $@" if $@;
+        goto \&try_harder;
 }
 
 # this API is designed to avoid creating self-referential structures;