The following document provides performance comparison for the following regular expression libraries:
Linux localhost.localdomain 2.6.30.9-99.fc11.x86_64 #1 SMP Tue Nov 17 21:30:38 EST 2009 x86_64 x86_64 x86_64 GNU/Linux
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 30
model name : Intel(R) Core(TM) i5 CPU 750 @ 2.67GHz
stepping : 5
cpu MHz : 1197.000
cache size : 8192 KB
physical id : 0
siblings : 4
core id : 0
cpu cores : 4
apicid : 0
initial apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 11
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology tsc_reliable nonstop_tsc pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 sse4_2 lahf_lm ida tpr_shadow vnmi flexpriority ept vpid
bogomips : 5467.63
clflush size : 64
cache_alignment : 64
address sizes : 36 bits physical, 48 bits virtual
power management:
The following tests evaluate the time taken to match the expression to the input string.
For each result, the top number has been normalized relative to the fastest time, so 1 is as good as it gets.
| astl | boost::regex | boost::xpressive | greta | pcre | Text | Expression |
| 2.61111 | 2.66667 | 1.41667 | 1 | 1.13889 | 100- this is a line of ftp response which contains a message string | ^([0-9]+)(\-| |$)(.*)$ |
| 1 | 3.19231 | 2.15385 | 2 | 1.57692 | 1234-5678-1234-456 | ([[:digit:]]{4}[- ]){3}[[:digit:]]{3,4} |
| 1 | 3.575 | 3.125 | 2.425 | 2.675 | john_maddock@compuserve.com | ^([a-zA-Z0-9_\-\.]+)@((\[[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.)|(([a-zA-Z0-9\-]+\.)+))([a-zA-Z]{2,4}|[0-9]{1,3})(\]?)$ |
| 1 | 6.6 | 5.8 | 4.2 | 4.9 | foo12@foo.edu | ^([a-zA-Z0-9_\-\.]+)@((\[[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.)|(([a-zA-Z0-9\-]+\.)+))([a-zA-Z]{2,4}|[0-9]{1,3})(\]?)$ |
| 1 | 5.48 | 4.64 | 3.36 | 3.68 | bob.smith@foo.tv | ^([a-zA-Z0-9_\-\.]+)@((\[[0-9]{1,3}\.[0-9]{1,3}\.[0-9]{1,3}\.)|(([a-zA-Z0-9\-]+\.)+))([a-zA-Z]{2,4}|[0-9]{1,3})(\]?)$ |
| 1 | 4.73333 | 1.86667 | 1.53333 | 2.2 | EH10 2QQ | ^[a-zA-Z]{1,2}[0-9][0-9A-Za-z]{0,1} {0,1}[0-9][A-Za-z]{2}$ |
| 1 | 5.08333 | 2.33333 | 1.66667 | 2.75 | G1 1AA | ^[a-zA-Z]{1,2}[0-9][0-9A-Za-z]{0,1} {0,1}[0-9][A-Za-z]{2}$ |
| 1 | 4.61538 | 2.15385 | 1.84615 | 2.53846 | SW1 1ZZ | ^[a-zA-Z]{1,2}[0-9][0-9A-Za-z]{0,1} {0,1}[0-9][A-Za-z]{2}$ |
| 1 | 3.66667 | 1.8 | 1.53333 | 1.8 | 4/1/2001 | ^[[:digit:]]{1,2}/[[:digit:]]{1,2}/[[:digit:]]{4}$ |
| 1 | 3.29412 | 1.64706 | 1.35294 | 1.64706 | 12/12/2001 | ^[[:digit:]]{1,2}/[[:digit:]]{1,2}/[[:digit:]]{4}$ |
| 1 | 7.5 | 3.25 | 2.125 | 3.5 | 123 | ^[-+]?[[:digit:]]*\.?[[:digit:]]*$ |
| 1 | 5.06667 | 1.8 | 1.33333 | 2 | +3.14159 | ^[-+]?[[:digit:]]*\.?[[:digit:]]*$ |
| 1 | 5.06667 | 1.8 | 1.33333 | 1.93333 | -3.14159 | ^[-+]?[[:digit:]]*\.?[[:digit:]]*$ |
These are average values: the perfect library would get 1, which means best performance for all tests.
| astl | 1.12393 |
| greta | 1.97762 |
| pcre | 2.48767 |
| boost::xpressive | 2.59896 |
| boost::regex | 4.65693 |
The next test measures the time to find all matches in a long English text. The text is the complete works of Mark Twain, from Project Gutenberg. The text is 19Mb long.
For each result, the top number has been normalized relative to the fastest time, so 1 is as good as it gets.
| astl | boost::regex | boost::xpressive | greta | pcre | Text | Expression |
| 1 | 1.21429 | 1 | 1 | 1.21429 | complete works of Mark Twain | Twain |
| 1 | 1.30769 | 1.07692 | 1.07692 | 1.23077 | complete works of Mark Twain | Huck[[:alpha:]]+ |
| 1 | 2.40566 | 2.0566 | 7.46226 | 11.1887 | complete works of Mark Twain | [[:alpha:]]+ing |
| 2.76316 | 1.36842 | 1 | 4.44737 | 1.97368 | complete works of Mark Twain | ^[^ ]*?Twain |
| 2.21739 | 1 | 1.13043 | 4.73913 | 1.3913 | complete works of Mark Twain | Tom|Sawyer|Huckleberry|Finn |
| 1.2 | 1.10909 | 1 | 2.4 | 2.36364 | complete works of Mark Twain | (Tom|Sawyer|Huckleberry|Finn).{0,30}river|river.{0,30}(Tom|Sawyer|Huckleberry|Finn) |
These are average values: the perfect library would get 1, which means best performance for all tests.
| boost::xpressive | 1.21066 |
| boost::regex | 1.40086 |
| astl | 1.53009 |
| pcre | 3.22706 |
| greta | 3.52095 |
For each of the following regular expressions the time taken to find all occurrences of the expression within the boost C++ source file crc.hpp was measured.
For each result, the top number has been normalized relative to the fastest time, so 1 is as good as it gets.
| astl | boost::regex | boost::xpressive | greta | pcre | Text | Expression |
| 3.38462 | 1 | 1.23077 | 5.07692 | 2.46154 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/crc.hpp | ^(template[[:space:]]*<[^;:{]+>[[:space:]]*)?(class|struct)[[:space:]]*(\<\w+\>([ ]*\([^)]*\))?[[:space:]]*)*(\<\w*\>)[[:space:]]*(<[^;:{]+>[[:space:]]*)?(\{|:[^;\{()]*\{) |
| 1 | 11.9535 | 10.7907 | 5.76744 | 24.3953 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/crc.hpp | (^[ ]+#(?:[^\\\n]|\\[^\n_[:punct:][:alnum:]]*[\n[:punct:][:alnum:]_])*)|(//[^\n]*|/\*.*?\*/)|\<([+-]?(?:(?:0x[[:xdigit:]]+)|(?:(?:[[:digit:]]*\.)?[[:digit:]]+(?:[eE][+-]?[[:digit:]]+)?))u?(?:(?:int(?:8|16|32|64))|L)?)\>|('(?:[^\\']|\\.)*'|"(?:[^\\"]|\\.)*")|\<(__asm|__cdecl|__declspec|__export|__far16|__fastcall|__fortran|__import|__pascal|__rtti|__stdcall|_asm|_cdecl|__except|_export|_far16|_fastcall|__finally|_fortran|_import|_pascal|_stdcall|__thread|__try|asm|auto|bool|break|case|catch|cdecl|char|class|const|const_cast|continue|default|delete|do|double|dynamic_cast|else|enum|explicit|extern|false|float|for|friend|goto|if|inline|int|long|mutable|namespace|new|operator|pascal|private|protected|public|register|reinterpret_cast|return|short|signed|sizeof|static|static_cast|struct|switch|template|this|throw|true|try|typedef|typeid|typename|union|unsigned|using|virtual|void|volatile|wchar_t|while)\> |
| 2.26316 | 1.05263 | 1 | 3.68421 | 1.42105 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/crc.hpp | ^[ ]*#[ ]*include[ ]+("[^"]+"|<[^>]+>) |
| 2.26316 | 1.05263 | 1 | 3.68421 | 1.42105 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/crc.hpp | ^[ ]*#[ ]*include[ ]+("boost/[^"]+"| |
These are average values: the perfect library would get 1, which means best performance for all tests.
| astl | 2.22773 |
| boost::xpressive | 3.50537 |
| boost::regex | 3.76469 |
| greta | 4.5532 |
| pcre | 7.42475 |
For each of the following regular expressions the time taken to find all occurrences of the expression within the boost html documentation file libs/libraries.htm was measured.
For each result, the top number has been normalized relative to the fastest time, so 1 is as good as it gets.
| astl | boost::regex | boost::xpressive | greta | pcre | Text | Expression |
| 1.04348 | 1.17391 | 1 | 2.6087 | 1.6087 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/libraries.htm | beman|john|dave |
| 1 | 1.9 | 1.2 | 1 | 2.1 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/libraries.htm | <p>.*?</p> |
| 1 | 1.5 | 1.5 | 1.82143 | 1.92857 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/libraries.htm | <a[^>]+href=("[^"]*"|[^[:space:]]+)[^>]*> |
| 1 | 1.11765 | 1.05882 | 1 | 1.35294 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/libraries.htm | <h[12345678][^>]*>.*?</h[12345678]> |
| 1 | 2.85714 | 1.14286 | 1.28571 | 2.85714 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/libraries.htm | <img[^>]+src=("[^"]*"|[^[:space:]]+)[^>]*> |
| 1 | 2.71429 | 1.14286 | 1 | 2.85714 | /home/vintz/src/sourceforge/svn/astl/trunk/bench/libraries.htm | <font[^>]+face=("[^"]*"|[^[:space:]]+)[^>]*>.*?</font> |
These are average values: the perfect library would get 1, which means best performance for all tests.
| astl | 1.00725 |
| boost::xpressive | 1.17409 |
| greta | 1.45264 |
| boost::regex | 1.87716 |
| pcre | 2.11742 |
These are average values: the perfect library would get 1, which means best performance for all tests.
| astl | 1.33607 |
| boost::xpressive | 2.14194 |
| greta | 2.54357 |
| pcre | 3.24502 |
| boost::regex | 3.28507 |
The next test evaluates the behavior of the libraries when the complexity of the expression increases: starting from a simple expression, at each step a new expression is concatenated as an alternative and the time to find all matches 10 times is measured.
Step 1: [[:upper:]]\. +[[:upper:]]\. +
Step 2: [[:upper:]]\. +[[:upper:]]\. +|Mr\.? [[:upper:]][[:lower:]]+ ([[:upper:]][[:lower:]] )*
Step 3: [[:upper:]]\. +[[:upper:]]\. +|Mr\.? [[:upper:]][[:lower:]]+ ([[:upper:]][[:lower:]] )*|days?
...
Step 20: [[:upper:]]\. +[[:upper:]]\. +|Mr\.? [[:upper:]][[:lower:]]+ ([[:upper:]][[:lower:]] )*|days?|P\. *S\.--|[[:upper:]][[:alpha:]]+ +[Ii]sles?|[$£]\d+([.,]\d+)*|"[[:alnum:]]+"|Mrs\.? [[:upper:]][[:lower:]]+ ([[:upper:]][[:lower:]] )*|(\d|1[0-2]) ?[aApP]\.?[Mm]\.?|([aA]\.?[Dd]|[bB]\.?[Cc])\.? \d{3,4}|[[:upper:]][[:lower:]]+[!?]|[[:upper:]]+( [[:upper:]]+)*|[hH]orses?|[oO]nce upon a time|Twain|(?i:CHAPTER [IVXLDC]+)|good terms?|Section \d+|New York|[Dd][Ee][Aa][Rr]
The text is the complete works of Mark Twain, from Project Gutenberg. The text is 19Mb long.