pcre2/doc/html/pcre2perform.html

281 lines
12 KiB
HTML
Raw Permalink Normal View History

2014-10-20 18:48:14 +02:00
<html>
<head>
<title>pcre2perform specification</title>
</head>
<body bgcolor="#FFFFFF" text="#00005A" link="#0066FF" alink="#3399FF" vlink="#2222BB">
<h1>pcre2perform man page</h1>
<p>
Return to the <a href="index.html">PCRE2 index page</a>.
</p>
<p>
This page is part of the PCRE2 HTML documentation. It was generated
automatically from the original man page. If there is any nonsense in it,
please consult the man page, in case the conversion went wrong.
<br>
2015-10-17 15:50:56 +02:00
<ul>
<li><a name="TOC1" href="#SEC1">PCRE2 PERFORMANCE</a>
<li><a name="TOC2" href="#SEC2">COMPILED PATTERN MEMORY USAGE</a>
2017-03-31 18:49:33 +02:00
<li><a name="TOC3" href="#SEC3">STACK AND HEAP USAGE AT RUN TIME</a>
2015-10-17 15:50:56 +02:00
<li><a name="TOC4" href="#SEC4">PROCESSING TIME</a>
<li><a name="TOC5" href="#SEC5">AUTHOR</a>
<li><a name="TOC6" href="#SEC6">REVISION</a>
</ul>
<br><a name="SEC1" href="#TOC1">PCRE2 PERFORMANCE</a><br>
2014-10-20 18:48:14 +02:00
<P>
Two aspects of performance are discussed below: memory usage and processing
time. The way you express your pattern as a regular expression can affect both
of them.
</P>
2015-10-17 15:50:56 +02:00
<br><a name="SEC2" href="#TOC1">COMPILED PATTERN MEMORY USAGE</a><br>
2014-10-20 18:48:14 +02:00
<P>
Patterns are compiled by PCRE2 into a reasonably efficient interpretive code,
so that most simple patterns do not use much memory for storing the compiled
2017-03-31 18:49:33 +02:00
version. However, there is one case where the memory usage of a compiled
pattern can be unexpectedly large. If a parenthesized group has a quantifier
with a minimum greater than 1 and/or a limited maximum, the whole group is
repeated in the compiled code. For example, the pattern
2014-10-20 18:48:14 +02:00
<pre>
(abc|def){2,4}
</pre>
is compiled as if it were
<pre>
(abc|def)(abc|def)((abc|def)(abc|def)?)?
</pre>
(Technical aside: It is done this way so that backtrack points within each of
the repetitions can be independently maintained.)
</P>
<P>
For regular expressions whose quantifiers use only small numbers, this is not
usually a problem. However, if the numbers are large, and particularly if such
repetitions are nested, the memory usage can become an embarrassment. For
example, the very simple pattern
<pre>
((ab){1,1000}c){1,3}
</pre>
uses over 50KiB when compiled using the 8-bit library. When PCRE2 is
2017-03-31 18:49:33 +02:00
compiled with its default internal pointer size of two bytes, the size limit on
a compiled pattern is 65535 code units in the 8-bit and 16-bit libraries, and
2017-03-31 18:49:33 +02:00
this is reached with the above pattern if the outer repetition is increased
from 3 to 4. PCRE2 can be compiled to use larger internal pointers and thus
handle larger compiled patterns, but it is better to try to rewrite your
pattern to use less memory if you can.
2014-10-20 18:48:14 +02:00
</P>
<P>
One way of reducing the memory usage for such patterns is to make use of
PCRE2's
<a href="pcre2pattern.html#subpatternsassubroutines">"subroutine"</a>
facility. Re-writing the above pattern as
<pre>
((ab)(?2){0,999}c)(?1){0,2}
</pre>
reduces the memory requirements to around 16KiB, and indeed it remains under
20KiB even with the outer repetition increased to 100. However, this kind of
pattern is not always exactly equivalent, because any captures within
subroutine calls are lost when the subroutine completes. If this is not a
problem, this kind of rewriting will allow you to process patterns that PCRE2
cannot otherwise handle. The matching performance of the two different versions
of the pattern are roughly the same. (This applies from release 10.30 - things
were different in earlier releases.)
2017-03-31 18:49:33 +02:00
</P>
<br><a name="SEC3" href="#TOC1">STACK AND HEAP USAGE AT RUN TIME</a><br>
<P>
From release 10.30, the interpretive (non-JIT) version of <b>pcre2_match()</b>
uses very little system stack at run time. In earlier releases recursive
function calls could use a great deal of stack, and this could cause problems,
but this usage has been eliminated. Backtracking positions are now explicitly
remembered in memory frames controlled by the code.
</P>
<P>
The size of each frame depends on the size of pointer variables and the number
of capturing parenthesized groups in the pattern being matched. On a 64-bit
system the frame size for a pattern with no captures is 128 bytes. For each
capturing group the size increases by 16 bytes.
</P>
<P>
Until release 10.41, an initial 20KiB frames vector was allocated on the system
stack, but this still caused some issues for multi-thread applications where
each thread has a very small stack. From release 10.41 backtracking memory
frames are always held in heap memory. An initial heap allocation is obtained
the first time any match data block is passed to <b>pcre2_match()</b>. This is
remembered with the match data block and re-used if that block is used for
another match. It is freed when the match data block itself is freed.
</P>
<P>
The size of the initial block is the larger of 20KiB or ten times the pattern's
frame size, unless the heap limit is less than this, in which case the heap
limit is used. If the initial block proves to be too small during matching, it
is replaced by a larger block, subject to the heap limit. The heap limit is
checked only when a new block is to be allocated. Reducing the heap limit
between calls to <b>pcre2_match()</b> with the same match data block does not
affect the saved block.
2017-03-31 18:49:33 +02:00
</P>
<P>
In contrast to <b>pcre2_match()</b>, <b>pcre2_dfa_match()</b> does use recursive
2017-03-31 18:49:33 +02:00
function calls, but only for processing atomic groups, lookaround assertions,
and recursion within the pattern. The original version of the code used to
allocate quite large internal workspace vectors on the stack, which caused some
problems for some patterns in environments with small stacks. From release
10.32 the code for <b>pcre2_dfa_match()</b> has been re-factored to use heap
memory when necessary for internal workspace when recursing, though recursive
function calls are still used.
</P>
<P>
The "match depth" parameter can be used to limit the depth of function
recursion, and the "match heap" parameter to limit heap memory in
<b>pcre2_dfa_match()</b>.
2014-10-20 18:48:14 +02:00
</P>
2015-10-17 15:50:56 +02:00
<br><a name="SEC4" href="#TOC1">PROCESSING TIME</a><br>
2014-10-20 18:48:14 +02:00
<P>
Certain items in regular expression patterns are processed more efficiently
than others. It is more efficient to use a character class like [aeiou] than a
set of single-character alternatives such as (a|e|i|o|u). In general, the
simplest construction that provides the required behaviour is usually the most
efficient. Jeffrey Friedl's book contains a lot of useful general discussion
about optimizing regular expressions for efficient performance. This document
contains a few observations about PCRE2.
</P>
<P>
Using Unicode character properties (the \p, \P, and \X escapes) is slow,
because PCRE2 has to use a multi-stage table lookup whenever it needs a
character's property. If you can find an alternative pattern that does not use
character properties, it will probably be faster.
</P>
<P>
By default, the escape sequences \b, \d, \s, and \w, and the POSIX
character classes such as [:alpha:] do not use Unicode properties, partly for
backwards compatibility, and partly for performance reasons. However, you can
set the PCRE2_UCP option or start the pattern with (*UCP) if you want Unicode
character properties to be used. This can double the matching time for items
such as \d, when matched with <b>pcre2_match()</b>; the performance loss is
less with a DFA matching function, and in both cases there is not much
difference for \b.
</P>
<P>
When a pattern begins with .* not in atomic parentheses, nor in parentheses
that are the subject of a backreference, and the PCRE2_DOTALL option is set,
the pattern is implicitly anchored by PCRE2, since it can match only at the
start of a subject string. If the pattern has multiple top-level branches, they
must all be anchorable. The optimization can be disabled by the
PCRE2_NO_DOTSTAR_ANCHOR option, and is automatically disabled if the pattern
contains (*PRUNE) or (*SKIP).
</P>
<P>
If PCRE2_DOTALL is not set, PCRE2 cannot make this optimization, because the
dot metacharacter does not then match a newline, and if the subject string
contains newlines, the pattern may match from the character immediately
following one of them instead of from the very start. For example, the pattern
2014-10-20 18:48:14 +02:00
<pre>
.*second
</pre>
matches the subject "first\nand second" (where \n stands for a newline
character), with the match starting at the seventh character. In order to do
this, PCRE2 has to retry the match starting after every newline in the subject.
</P>
<P>
If you are using such a pattern with subject strings that do not contain
newlines, the best performance is obtained by setting PCRE2_DOTALL, or starting
the pattern with ^.* or ^.*? to indicate explicit anchoring. That saves PCRE2
from having to scan along the subject looking for a newline to restart at.
</P>
<P>
Beware of patterns that contain nested indefinite repeats. These can take a
long time to run when applied to a string that does not match. Consider the
pattern fragment
<pre>
^(a+)*
</pre>
This can match "aaaa" in 16 different ways, and this number increases very
rapidly as the string gets longer. (The * repeat can match 0, 1, 2, 3, or 4
times, and for each of those cases other than 0 or 4, the + repeats can match
different numbers of times.) When the remainder of the pattern is such that the
entire match is going to fail, PCRE2 has in principle to try every possible
variation, and this can take an extremely long time, even for relatively short
strings.
</P>
<P>
An optimization catches some of the more simple cases such as
<pre>
(a+)*b
</pre>
where a literal character follows. Before embarking on the standard matching
procedure, PCRE2 checks that there is a "b" later in the subject string, and if
there is not, it fails the match immediately. However, when there is no
following literal this optimization cannot be used. You can see the difference
by comparing the behaviour of
<pre>
(a+)*\d
</pre>
with the pattern above. The former gives a failure almost instantly when
applied to a whole line of "a" characters, whereas the latter takes an
appreciable time with strings longer than about 20 characters.
</P>
<P>
In many cases, the solution to this kind of performance issue is to use an
atomic group or a possessive quantifier. This can often reduce memory
2017-03-31 18:49:33 +02:00
requirements as well. As another example, consider this pattern:
<pre>
([^&#60;]|&#60;(?!inet))+
</pre>
It matches from wherever it starts until it encounters "&#60;inet" or the end of
the data, and is the kind of pattern that might be used when processing an XML
file. Each iteration of the outer parentheses matches either one character that
is not "&#60;" or a "&#60;" that is not followed by "inet". However, each time a
parenthesis is processed, a backtracking position is passed, so this
formulation uses a memory frame for each matched character. For a long string,
a lot of memory is required. Consider now this rewritten pattern, which matches
exactly the same strings:
<pre>
([^&#60;]++|&#60;(?!inet))+
</pre>
This runs much faster, because sequences of characters that do not contain "&#60;"
are "swallowed" in one item inside the parentheses, and a possessive quantifier
is used to stop any backtracking into the runs of non-"&#60;" characters. This
version also uses a lot less memory because entry to a new set of parentheses
happens only when a "&#60;" character that is not followed by "inet" is encountered
(and we assume this is relatively rare).
2017-03-31 18:49:33 +02:00
</P>
<P>
This example shows that one way of optimizing performance when matching long
subject strings is to write repeated parenthesized subpatterns to match more
than one character whenever possible.
</P>
<br><b>
SETTING RESOURCE LIMITS
</b><br>
<P>
You can set limits on the amount of processing that takes place when matching,
2017-03-31 18:49:33 +02:00
and on the amount of heap memory that is used. The default values of the limits
are very large, and unlikely ever to operate. They can be changed when PCRE2 is
built, and they can also be set when <b>pcre2_match()</b> or
2017-03-31 18:49:33 +02:00
<b>pcre2_dfa_match()</b> is called. For details of these interfaces, see the
<a href="pcre2build.html"><b>pcre2build</b></a>
documentation and the section entitled
<a href="pcre2api.html#matchcontext">"The match context"</a>
in the
<a href="pcre2api.html"><b>pcre2api</b></a>
documentation.
</P>
<P>
The <b>pcre2test</b> test program has a modifier called "find_limits" which, if
applied to a subject line, causes it to find the smallest limits that allow a
pattern to match. This is done by repeatedly matching with different limits.
2014-10-20 18:48:14 +02:00
</P>
2015-10-17 15:50:56 +02:00
<br><a name="SEC5" href="#TOC1">AUTHOR</a><br>
2014-10-20 18:48:14 +02:00
<P>
Philip Hazel
<br>
Retired from University Computing Service
2014-10-20 18:48:14 +02:00
<br>
2014-11-21 17:45:06 +01:00
Cambridge, England.
2014-10-20 18:48:14 +02:00
<br>
</P>
2015-10-17 15:50:56 +02:00
<br><a name="SEC6" href="#TOC1">REVISION</a><br>
2014-10-20 18:48:14 +02:00
<P>
Last updated: 27 July 2022
2014-10-20 18:48:14 +02:00
<br>
Copyright &copy; 1997-2022 University of Cambridge.
2014-10-20 18:48:14 +02:00
<br>
<p>
Return to the <a href="index.html">PCRE2 index page</a>.
</p>