]> Sergey Matveev's repositories - public-inbox.git/blobdiff - lib/PublicInbox/Search.pm
www: improve visibility of coderepos
[public-inbox.git] / lib / PublicInbox / Search.pm
index f42d70e3c6c8fda84bd3be15e5d4c2a30b5bbbde..c7d52dafc3f4f9e37491aea50c6ac13880ee8b62 100644 (file)
@@ -13,9 +13,9 @@ use POSIX qw(strftime);
 # values for searching, changing the numeric value breaks
 # compatibility with old indices (so don't change them it)
 use constant {
-       TS => 0, # Received: header in Unix time (IMAP INTERNALDATE)
+       TS => 0, # Received: in Unix time (IMAP INTERNALDATE, JMAP receivedAt)
        YYYYMMDD => 1, # Date: header for searching in the WWW UI
-       DT => 2, # Date: YYYYMMDDHHMMSS
+       DT => 2, # Date: YYYYMMDDHHMMSS (IMAP SENT*, JMAP sentAt)
 
        # added for public-inbox 1.6.0+
        BYTES => 3, # IMAP RFC822.SIZE
@@ -265,8 +265,10 @@ sub reopen {
 # and neither the SWIG nor XS bindings allow us to use custom code
 # to parse dates (and libgit2 doesn't expose git__date_parse, either,
 # so we're running git-rev-parse(1)).
-sub date_range {
-       my ($git, $pfx, $range) = @_;
+# This replaces things we need to send to $git->date_parse with
+# "\0".$strftime_format.['+'|$idx]."\0" placeholders
+sub date_parse_prepare {
+       my ($to_parse, $pfx, $range) = @_;
        # are we inside a parenthesized statement?
        my $end = $range =~ s/([\)\s]*)\z// ? $1 : '';
        my @r = split(/\.\./, $range, 2);
@@ -275,55 +277,93 @@ sub date_range {
        # n.b. git doesn't do YYYYMMDD w/o '-', it needs YYYY-MM-DD
        if ($pfx eq 'd') {
                if (!defined($r[1])) {
-                       $r[0] =~ s/\A([0-9]{4})([0-9]{2})([0-9]{2})\z/$1-$2-$3/;
-                       $r[0] = $git->date_parse($r[0]);
-                       $r[1] = $r[0] + 86400;
-                       for my $x (@r) {
-                               $x = strftime('%Y%m%d', gmtime($x));
+                       if ($r[0] =~ /\A([0-9]{4})([0-9]{2})([0-9]{2})\z/) {
+                               push @$to_parse, "$1-$2-$3";
+                               # we could've handled as-is, but we need
+                               # to parse anyways for "d+" below
+                       } else {
+                               push @$to_parse, $r[0];
                        }
+                       $r[0] = "\0%Y%m%d$#$to_parse\0";
+                       $r[1] = "\0%Y%m%d+\0";
                } else {
                        for my $x (@r) {
                                next if $x eq '' || $x =~ /\A[0-9]{8}\z/;
-                               $x = strftime('%Y%m%d',
-                                               gmtime($git->date_parse($x)));
+                               push @$to_parse, $x;
+                               $x = "\0%Y%m%d$#$to_parse\0";
                        }
                }
        } elsif ($pfx eq 'dt') {
                if (!defined($r[1])) { # git needs gaps and not /\d{14}/
-                       $r[0] =~ s/\A([0-9]{4})([0-9]{2})([0-9]{2})
-                                       ([0-9]{2})([0-9]{2})([0-9]{2})\z
-                               /$1-$2-$3 $4:$5:$6/x;
-                       $r[0] = $git->date_parse($r[0]);
-                       $r[1] = $r[0] + 86400;
-                       for my $x (@r) {
-                               $x = strftime('%Y%m%d%H%M%S', gmtime($x));
+                       if ($r[0] =~ /\A([0-9]{4})([0-9]{2})([0-9]{2})
+                                       ([0-9]{2})([0-9]{2})([0-9]{2})\z/x) {
+                               push @$to_parse, "$1-$2-$3 $4:$5:$6";
+                       } else {
+                               push @$to_parse, $r[0];
                        }
+                       $r[0] = "\0%Y%m%d%H%M%S$#$to_parse\0";
+                       $r[1] = "\0%Y%m%d%H%M%S+\0";
                } else {
                        for my $x (@r) {
                                next if $x eq '' || $x =~ /\A[0-9]{14}\z/;
-                               $x = strftime('%Y%m%d%H%M%S',
-                                               gmtime($git->date_parse($x)));
+                               push @$to_parse, $x;
+                               $x = "\0%Y%m%d%H%M%S$#$to_parse\0";
                        }
                }
        } else { # "rt", let git interpret "YYYY", deal with Y10K later :P
                for my $x (@r) {
                        next if $x eq '' || $x =~ /\A[0-9]{5,}\z/;
-                       $x = $git->date_parse($x);
+                       push @$to_parse, $x;
+                       $x = "\0%s$#$to_parse\0";
                }
-               $r[1] //= $r[0] + 86400;
+               $r[1] //= "\0%s+\0";
        }
        "$pfx:".join('..', @r).$end;
 }
 
+sub date_parse_finalize {
+       my ($git, $to_parse) = @_;
+       # git-rev-parse can handle any number of args up to system
+       # limits (around (4096*32) bytes on Linux).
+       my @r = $git->date_parse(@$to_parse);
+       my $i;
+       $_[2] =~ s/\0(%[%YmdHMSs]+)([0-9\+]+)\0/strftime($1,
+               gmtime($2 eq '+' ? ($r[$i]+86400) : $r[$i=$2+0]))/sge;
+}
+
+# n.b. argv never has NUL, though we'll need to filter it out
+# if this $argv isn't from a command execution
 sub query_argv_to_string {
        my (undef, $git, $argv) = @_;
-       join(' ', map {;
-               if (s!\b(d|rt|dt):(.+)\z!date_range($git, $1, $2)!sge) {
+       my $to_parse;
+       my $tmp = join(' ', map {;
+               if (s!\b(d|rt|dt):(\S+)\z!date_parse_prepare(
+                                               $to_parse //= [], $1, $2)!sge) {
                        $_;
+               } elsif (/\s/) {
+                       s/(.*?)\b(\w+:)// ? qq{$1$2"$_"} : qq{"$_"};
                } else {
-                       /\s/ ? (s/\A(\w+:)// ? qq{$1"$_"} : qq{"$_}) : $_
+                       $_
                }
        } @$argv);
+       date_parse_finalize($git, $to_parse, $tmp) if $to_parse;
+       $tmp
+}
+
+# this is for the WWW "q=" query parameter and "lei q --stdin"
+# it can't do d:"5 days ago", but it will do d:5.days.ago
+sub query_approxidate {
+       my (undef, $git) = @_; # $_[2] = $query_string (modified in-place)
+       my $DQ = qq<"\x{201c}\x{201d}>; # Xapian can use curly quotes
+       $_[2] =~ tr/\x00/ /; # Xapian doesn't do NUL, we use it as a placeholder
+       my ($terms, $phrase, $to_parse);
+       $_[2] =~ s{([^$DQ]*)([$DQ][^$DQ]*[$DQ])?}{
+               ($terms, $phrase) = ($1, $2);
+               $terms =~ s!\b(d|rt|dt):(\S+)!
+                       date_parse_prepare($to_parse //= [], $1, $2)!sge;
+               $terms.($phrase // '');
+               }sge;
+       date_parse_finalize($git, $to_parse, $_[2]) if $to_parse;
 }
 
 # read-only
@@ -388,18 +428,17 @@ sub _enquire_once { # retry_reopen callback
        my $enquire = $X{Enquire}->new($xdb);
        $enquire->set_query($query);
        $opts ||= {};
-        my $desc = !$opts->{asc};
        my $rel = $opts->{relevance} // 0;
-       if ($rel == -1) { # ORDER BY docid/UID
+       if ($rel == -2) { # ORDER BY docid/UID (highest first)
+               $enquire->set_weighting_scheme($X{BoolWeight}->new);
+               $enquire->set_docid_order($ENQ_DESCENDING);
+       } elsif ($rel == -1) { # ORDER BY docid/UID (lowest first)
                $enquire->set_weighting_scheme($X{BoolWeight}->new);
                $enquire->set_docid_order($ENQ_ASCENDING);
        } elsif ($rel == 0) {
-               $enquire->set_sort_by_value_then_relevance(TS, $desc);
-       } elsif ($rel == -2) {
-               $enquire->set_weighting_scheme($X{BoolWeight}->new);
-               $enquire->set_docid_order($ENQ_DESCENDING);
+               $enquire->set_sort_by_value_then_relevance(TS, !$opts->{asc});
        } else { # rel > 0
-               $enquire->set_sort_by_relevance_then_value(TS, $desc);
+               $enquire->set_sort_by_relevance_then_value(TS, !$opts->{asc});
        }
 
        # `mairix -t / --threads' or JMAP collapseThreads
@@ -442,7 +481,7 @@ sub qparse_new ($) {
        $cb->($qp, $NVRP->new(DT, 'dt:'));
 
        # for IMAP, undocumented for WWW and may be split off go away
-       $cb->($qp, $NVRP->new(BYTES, 'bytes:'));
+       $cb->($qp, $NVRP->new(BYTES, 'z:'));
        $cb->($qp, $NVRP->new(TS, 'rt:'));
        $cb->($qp, $NVRP->new(UID, 'uid:'));