Paul Eggert wrote:
Attached are some proposed patches which should improve the performance
of grep -P when applied to binary files, among other things. I have
some other ideas for boosting performance further but thought I'd
publish these first.
I pushed those patches, along with the attached further patches to fix
up some porting glitches and bugs I encountered in subsequent testing.
I plan to follow up soon on Bug#18454 with more performance-related
patches in this area.
From 53c5d9fd50b6895b886c1d19d0851562fc03e00c Mon Sep 17 00:00:00 2001
From: Paul Eggert <egg...@cs.ucla.edu>
Date: Tue, 16 Sep 2014 17:29:40 -0700
Subject: [PATCH 07/10] grep: avoid false alarms for mb_clen and to_uchar
* cfg.mk (_gl_TS_unmarked_extern_functions): New var,
to bypass the tight_scope false alarms on mb_clen and to_uchar.
---
cfg.mk | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/cfg.mk b/cfg.mk
index 947d184..3316b5d 100644
--- a/cfg.mk
+++ b/cfg.mk
@@ -28,6 +28,10 @@ local-checks-to-skip = \
# Tools used to bootstrap this package, used for "announcement".
bootstrap-tools = autoconf,automake,gnulib
+# The tight_scope test gets confused about inline functions.
+# like 'to_uchar'.
+_gl_TS_unmarked_extern_functions = main usage mb_clen to_uchar
+
# Now that we have better tests, make this the default.
export VERBOSE = yes
--
1.9.3
From 493ddec2e61d48953600575896a5d3ce1d1a582b Mon Sep 17 00:00:00 2001
From: Paul Eggert <egg...@cs.ucla.edu>
Date: Mon, 15 Sep 2014 22:25:21 -0700
Subject: [PATCH 08/10] grep: use mbclen cache in one more place
* src/grep.c (fgrep_to_grep_pattern): Use mb_clen here, too.
---
src/grep.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/src/grep.c b/src/grep.c
index 72a811e..e4379bc 100644
--- a/src/grep.c
+++ b/src/grep.c
@@ -1912,8 +1912,7 @@ fgrep_to_grep_pattern (size_t len, char const *keys,
for (; len; keys += n, len -= n)
{
- wchar_t wc;
- n = mbrtowc (&wc, keys, len, &mb_state);
+ n = mb_clen (keys, len, &mb_state);
switch (n)
{
case (size_t) -2:
--
1.9.3
From 219f10596c17e38b2716673a140c2b3827549862 Mon Sep 17 00:00:00 2001
From: Paul Eggert <egg...@cs.ucla.edu>
Date: Mon, 15 Sep 2014 17:27:58 -0700
Subject: [PATCH 09/10] grep: port -P speedup to hosts lacking
PCRE_STUDY_JIT_COMPILE
* src/pcresearch.c (Pcompile): Do not assume that
PCRE_STUDY_JIT_COMPILE is defined.
(empty_match): Define on all platforms.
---
src/pcresearch.c | 14 +++++++-------
1 file changed, 7 insertions(+), 7 deletions(-)
diff --git a/src/pcresearch.c b/src/pcresearch.c
index 95877e3..ce65758 100644
--- a/src/pcresearch.c
+++ b/src/pcresearch.c
@@ -33,10 +33,6 @@ static pcre *cre;
/* Additional information about the pattern. */
static pcre_extra *extra;
-/* Table, indexed by ! (flag & PCRE_NOTBOL), of whether the empty
- string matches when that flag is used. */
-static int empty_match[2];
-
# ifdef PCRE_STUDY_JIT_COMPILE
static pcre_jit_stack *jit_stack;
# else
@@ -44,6 +40,10 @@ static pcre_jit_stack *jit_stack;
# endif
#endif
+/* Table, indexed by ! (flag & PCRE_NOTBOL), of whether the empty
+ string matches when that flag is used. */
+static int empty_match[2];
+
void
Pcompile (char const *pattern, size_t size)
{
@@ -129,11 +129,11 @@ Pcompile (char const *pattern, size_t size)
pcre_assign_jit_stack (extra, NULL, jit_stack);
}
- empty_match[false] = pcre_exec (cre, extra, "", 0, 0, PCRE_NOTBOL, NULL, 0);
- empty_match[true] = pcre_exec (cre, extra, "", 0, 0, 0, NULL, 0);
-
# endif
free (re);
+
+ empty_match[false] = pcre_exec (cre, extra, "", 0, 0, PCRE_NOTBOL, NULL, 0);
+ empty_match[true] = pcre_exec (cre, extra, "", 0, 0, 0, NULL, 0);
#endif /* HAVE_LIBPCRE */
}
--
1.9.3
From 530fd765922b16643c78652ef036024fc4dd72eb Mon Sep 17 00:00:00 2001
From: Paul Eggert <egg...@cs.ucla.edu>
Date: Mon, 15 Sep 2014 18:33:19 -0700
Subject: [PATCH 10/10] grep: fix -P speedup bug with empty match
* src/pcresearch.c (NSUB): New top-level constant, replacing
'nsub' within Pexecute.
(Pcompile, Pexecute): Use it.
(Pexecute): Don't assume sub[1] is zero after a PCRE_ERROR_BADUTF8
match failure.
* tests/pcre-invalid-utf8-input: Test for this bug.
---
src/pcresearch.c | 32 +++++++++++++++++++-------------
tests/pcre-invalid-utf8-input | 5 +++++
2 files changed, 24 insertions(+), 13 deletions(-)
diff --git a/src/pcresearch.c b/src/pcresearch.c
index ce65758..c41f7ef 100644
--- a/src/pcresearch.c
+++ b/src/pcresearch.c
@@ -44,6 +44,10 @@ static pcre_jit_stack *jit_stack;
string matches when that flag is used. */
static int empty_match[2];
+/* This must be at least 2; everything after that is for performance
+ in pcre_exec. */
+enum { NSUB = 300 };
+
void
Pcompile (char const *pattern, size_t size)
{
@@ -132,8 +136,10 @@ Pcompile (char const *pattern, size_t size)
# endif
free (re);
- empty_match[false] = pcre_exec (cre, extra, "", 0, 0, PCRE_NOTBOL, NULL, 0);
- empty_match[true] = pcre_exec (cre, extra, "", 0, 0, 0, NULL, 0);
+ int sub[NSUB];
+ empty_match[false] = pcre_exec (cre, extra, "", 0, 0,
+ PCRE_NOTBOL, sub, NSUB);
+ empty_match[true] = pcre_exec (cre, extra, "", 0, 0, 0, sub, NSUB);
#endif /* HAVE_LIBPCRE */
}
@@ -146,11 +152,7 @@ Pexecute (char const *buf, size_t size, size_t *match_size,
error (EXIT_TROUBLE, 0, _("internal error"));
return -1;
#else
- /* This array must have at least two elements; everything after that
- is just for performance improvement in pcre_exec. */
- enum { nsub = 300 };
- int sub[nsub];
-
+ int sub[NSUB];
char const *p = start_ptr ? start_ptr : buf;
bool bol = p[-1] == eolbyte;
char const *line_start = buf;
@@ -174,15 +176,19 @@ Pexecute (char const *buf, size_t size, size_t
*match_size,
{
int options = bol ? 0 : PCRE_NOTBOL;
int valid_bytes;
- e = pcre_exec (cre, extra, p, line_end - p, 0, options, sub, nsub);
+ e = pcre_exec (cre, extra, p, line_end - p, 0, options, sub, NSUB);
if (e != PCRE_ERROR_BADUTF8)
break;
valid_bytes = sub[0];
- e = (valid_bytes == 0
- ? empty_match[bol]
- : pcre_exec (cre, extra, p, valid_bytes, 0,
- options | PCRE_NO_UTF8_CHECK | PCRE_NOTEOL,
- sub, nsub));
+ if (valid_bytes == 0)
+ {
+ sub[1] = 0;
+ e = empty_match[bol];
+ }
+ else
+ e = pcre_exec (cre, extra, p, valid_bytes, 0,
+ options | PCRE_NO_UTF8_CHECK | PCRE_NOTEOL,
+ sub, NSUB);
if (e != PCRE_ERROR_NOMATCH)
break;
p += valid_bytes + 1;
diff --git a/tests/pcre-invalid-utf8-input b/tests/pcre-invalid-utf8-input
index 9da4b18..78bd1cf 100755
--- a/tests/pcre-invalid-utf8-input
+++ b/tests/pcre-invalid-utf8-input
@@ -21,4 +21,9 @@ test $? -eq 0 || fail=1
LC_ALL=en_US.UTF-8 grep -P 'k$' in
test $? -eq 1 || fail=1
+echo k >exp
+
+LC_ALL=en_US.UTF-8 grep -aoP 'k*' in >out || fail=1
+compare exp out || fail=1
+
Exit $fail
--
1.9.3
