On Mon, 26 Dec 2016 12:07:49 -0800
Paul Eggert <egg...@cs.ucla.edu> wrote:
> Norihiro Tanaka wrote:
> > Hmm, how about the following test cases, although it is extreame?
>
> I don't think we need to worry about performance for the case when -w
> is given, and a pattern matches data that contains non-word
> characters. In practice, such cases are rare. I expect that most
> users would be surprised that -w can match non-word characters, and
> that users wouldn't object to -w rejecting such matches (if this
> wouldn't hurt performance significantly).
>
> While looking into this I did find a very small performance tweak for
> the test case, and installed the attached.
Thanks.
BTW, with multiple patterns in current master, former uses fgrep matcher,
and later users grep matcher. I think that it is not reasonable.
env LC_ALL=C grep -w -f pat inp
env LC_ALL=C grep -F -w -f pat inp
So I wrote the patch to use fgrep matcher for both.
From 08d426d8a0bf8683b0e34ead2fd3561671414735 Mon Sep 17 00:00:00 2001
From: Norihiro Tanaka <nori...@kcn.ne.jp>
Date: Wed, 28 Dec 2016 08:57:54 +0900
Subject: [PATCH] grep: imorove performance with multiple patterns
* src/grep.c (main): Avoid fgrep-to-grep conversion for word matching
with multiple patterns in single byte locales.
---
src/grep.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/src/grep.c b/src/grep.c
index aebab20..8cf522c 100644
--- a/src/grep.c
+++ b/src/grep.c
@@ -2862,7 +2862,7 @@ main (int argc, char **argv)
for -iF. */
if (matcher == F_MATCHER_INDEX
&& (! localeinfo.multibyte
- ? match_words
+ ? (n_patterns == 1 && match_words)
: (contains_encoding_error (keys, keycc)
|| (match_icase && !fgrep_icase_available (keys, keycc)))))
{
--
1.7.1
