* [PATCH 1/3] eml: reduce RE captures and possible side effects
2020-05-09 8:27 [PATCH 0/3] eml: some fixes and speedups Eric Wong
@ 2020-05-09 8:27 ` Eric Wong
2020-05-09 8:27 ` [PATCH 2/3] eml: speed up common LF-only emails Eric Wong
2020-05-09 8:27 ` [PATCH 3/3] emlcontentfoo: quiet warning on missing attributes Eric Wong
2 siblings, 0 replies; 4+ messages in thread
From: Eric Wong @ 2020-05-09 8:27 UTC (permalink / raw)
To: meta
Since Perl 5.6, the `@-' (aka @LAST_MATCH_START) and `@+' (aka
@LAST_MATCH_END) arrays provides integer offsets for every match
as documented in perlvar(1), regardless of regexp modifiers.
We can avoid relying on $1 in the epilogue scan, entirely.
So use these instead of relying on m//g and pos(), since the `g'
modifier can be affected by m//g matches performed in other
places.
Unrelated, but while we're in the area: remove some unnecessary
use of (?:...), too.
---
lib/PublicInbox/Eml.pm | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/lib/PublicInbox/Eml.pm b/lib/PublicInbox/Eml.pm
index 4508bd84..80e7c1af 100644
--- a/lib/PublicInbox/Eml.pm
+++ b/lib/PublicInbox/Eml.pm
@@ -71,11 +71,11 @@ sub re_memo ($) {
# compatible with our uses of Email::MIME
sub new {
my $ref = ref($_[1]) ? $_[1] : \(my $cpy = $_[1]);
- if ($$ref =~ /(?:\r?\n(\r?\n))/gs) { # likely
+ if ($$ref =~ /\r?\n(\r?\n)/s) { # likely
# This can modify $$ref in-place and to avoid memcpy/memmove
# on a potentially large $$ref. It does need to make a
# copy for $hdr, though. Idea stolen from Email::Simple
- my $hdr = substr($$ref, 0, pos($$ref), ''); # sv_chop on $$ref
+ my $hdr = substr($$ref, 0, $+[0], ''); # sv_chop on $$ref
substr($hdr, -(length($1))) = ''; # lower SvCUR
bless { hdr => \$hdr, crlf => $1, bdy => $ref }, __PACKAGE__;
} elsif ($$ref =~ /^[a-z0-9-]+[ \t]*:/ims && $$ref =~ /(\r?\n)\z/s) {
@@ -90,8 +90,8 @@ sub new {
sub new_sub {
my (undef, $ref) = @_;
# special case for messages like <85k5su9k59.fsf_-_@lola.goethe.zz>
- $$ref =~ /\A(?:(\r?\n))/gs or goto &new;
- my $hdr = substr($$ref, 0, pos($$ref), ''); # sv_chop on $$ref
+ $$ref =~ /\A(\r?\n)/s or goto &new;
+ my $hdr = substr($$ref, 0, $+[0], ''); # sv_chop on $$ref
bless { hdr => \$hdr, crlf => $1, bdy => $ref }, __PACKAGE__;
}
@@ -132,8 +132,8 @@ sub mp_descend ($$) {
# *sigh* just the regexp match alone seems to bump RSS by
# length($$bdy) on a ~30M string:
my $epilogue_missing;
- if ($$bdy =~ /((?:\r?\n)?^--$bnd--[ \t]*\r?$)/gsm) {
- substr($$bdy, pos($$bdy) - length($1)) = '';
+ if ($$bdy =~ /(?:\r?\n)?^--$bnd--[ \t]*\r?$/sm) {
+ substr($$bdy, $-[0]) = '';
} else {
$epilogue_missing = 1;
}
^ permalink raw reply related [flat|nested] 4+ messages in thread* [PATCH 2/3] eml: speed up common LF-only emails
2020-05-09 8:27 [PATCH 0/3] eml: some fixes and speedups Eric Wong
2020-05-09 8:27 ` [PATCH 1/3] eml: reduce RE captures and possible side effects Eric Wong
@ 2020-05-09 8:27 ` Eric Wong
2020-05-09 8:27 ` [PATCH 3/3] emlcontentfoo: quiet warning on missing attributes Eric Wong
2 siblings, 0 replies; 4+ messages in thread
From: Eric Wong @ 2020-05-09 8:27 UTC (permalink / raw)
To: meta
Emails a *nix MTA are typically LF-only, so we don't need the
complexity of the RE engine when a simple index() works. We
still need to ensure there's no "\r\n\r\n" before the first
"\n\n", but two calls to index() is still faster than a RE
match.
This gives a 2-5% speedup in some informal tests and saves ~30MB
when scanning a 30MB spam message on newer versions of Perl.
I'll have to diagnose why Perl wastes so much memory doing
RE matches on giant strings, though.
---
lib/PublicInbox/Eml.pm | 16 ++++++++++++----
1 file changed, 12 insertions(+), 4 deletions(-)
diff --git a/lib/PublicInbox/Eml.pm b/lib/PublicInbox/Eml.pm
index 80e7c1af..f022516c 100644
--- a/lib/PublicInbox/Eml.pm
+++ b/lib/PublicInbox/Eml.pm
@@ -71,10 +71,18 @@ sub re_memo ($) {
# compatible with our uses of Email::MIME
sub new {
my $ref = ref($_[1]) ? $_[1] : \(my $cpy = $_[1]);
- if ($$ref =~ /\r?\n(\r?\n)/s) { # likely
- # This can modify $$ref in-place and to avoid memcpy/memmove
- # on a potentially large $$ref. It does need to make a
- # copy for $hdr, though. Idea stolen from Email::Simple
+ # substr() can modify the first arg in-place and to avoid
+ # memcpy/memmove on a potentially large scalar. It does need
+ # to make a copy for $hdr, though. Idea stolen from Email::Simple.
+
+ # We also prefer index() on common LFLF emails since it's faster
+ # and re scan can bump RSS by length($$ref) on big strings
+ if (index($$ref, "\r\n") < 0 && (my $pos = index($$ref, "\n\n")) >= 0) {
+ # likely on *nix
+ my $hdr = substr($$ref, 0, $pos + 2, ''); # sv_chop on $$ref
+ chop($hdr); # lower SvCUR
+ bless { hdr => \$hdr, crlf => "\n", bdy => $ref }, __PACKAGE__;
+ } elsif ($$ref =~ /\r?\n(\r?\n)/s) {
my $hdr = substr($$ref, 0, $+[0], ''); # sv_chop on $$ref
substr($hdr, -(length($1))) = ''; # lower SvCUR
bless { hdr => \$hdr, crlf => $1, bdy => $ref }, __PACKAGE__;
^ permalink raw reply related [flat|nested] 4+ messages in thread