Skip to content
/ regXwild Public

⏱ Superfast ^Advanced wildcards++? | Unique algorithms that was implemented on native unmanaged C++ but easily accessible in .NET via Conari (with caching of 0x29 opcodes +optimizations) etc.

License

Notifications You must be signed in to change notification settings

3F/regXwild

Repository files navigation

⏱ Superfast ^Advanced wildcards++? *,|,?,^,$,+,#,>,++??,##??,>c in addition to slow regex engines and more.

✔ regex-like quantifiers, amazing meta symbols, and speed...

Unique algorithms that was implemented on native unmanaged C++ but easily accessible in .NET through Conari (recommended due to caching of 0x29 opcodes + related optimizations), and others such as python etc.

Build status release License NuGet package Tests

Build history

Samples regXwild filter n
number = '1271'; number = '????'; 0 - 4
year = '2020'; '##'|'####' 2 | 4
year = '20'; = '##??' 2 | 4
number = 888; number = +??; 1 - 3
Samples regXwild filter
everything is ok ^everything*ok$
systems system?
systems sys###s
A new 'X1' project ^A*'+' pro?ect
professional system pro*system
regXwild in action pro?ect$|open*source+act|^regXwild

Why regXwild ?

It was designed to be faster than just fast for features that usually go beyond the typical wildcards. Seriously, We love regex, I love, You love; 2013 far behind but regXwild still relevant for speed and powerful wildcards-like features, such as ##?? (which means 2 or 4) ...

🔍 Easy to start

Unmanaged native C++ or managed .NET project. It doesn't matter, just use it:

C++

#include <regXwild.h>
using namespace net::r_eg::regXwild;
...
EssRxW rxw;
if(rxw.match(_T("regXwild"), _T("reg?wild"))) {
    // ...
}

C# if Conari

using dynamic l = new ConariX("regXwild.dll");
...
if(l.match<bool>("regXwild", "reg?wild")) {
    // ...
}

🏄 Amazing meta symbols

ESS version (advanced EXT version)

metasymbol meaning
* {0, ~}
| str1 or str2 or ...
? {0, 1}, ??? {0, 3}, ...
^ [str... or [str1...
$ ...str] or ...str1]
+ {1, ~}, +++ {3, ~}, ...
# {1}, ## {2}, ### {3}, ...
> Legacy > (F_LEGACY_ANYSP = 0x008) as [^/]*str | [^/]*$
>c 1.4+ Modern > as [^**c**]*str | [^**c**]*$

EXT version (more simplified than ESS)

metasymbol meaning
* {0, ~}
> as [^/\\]+
| str1 or str2 or ...
? {0, 1}, ??? {0, 3}, ...

🧮 Quantifiers

1.3+ ++??; ##??

regex regXwild n
.* * 0+
.+ + 1+
.? ? 0 | 1
.{1} # 1
.{2} ## 2
.{2, } ++ 2+
.{0, 2} ?? 0 - 2
.{2, 4} ++?? 2 - 4
(?:.{2}|.{4}) ##?? 2 | 4
.{3, 4} +++? 3 - 4
(?:.{1}|.{3}) #?? 1 | 3

and similar ...

Play with our actual Unit-Tests.

🚀 Awesome speed

  • ~2000 times faster when C++.
  • For .NET (including modern .NET Core), Conari provides optional caching of 0x29 opcodes (Calli) and more to get similar to C++ result as possible.

Match result and Replacements

1.4+

EssRxW::MatchResult m;
rxw.match
(
    _T("number = '8888'; //TODO: up"),
    _T("'+'"),
    EssRxW::EngineOptions::F_MATCH_RESULT,
    &m
);
//m.start = 9
//m.end = 15
...
input.replace(m.start, m.end - m.start, _T("'9777'"));
tstring str = _T("year = 2021; dd = 17;");
...
if(rxw.replace(str, _T(" ##;"), _T(" 00;"))) {
    // year = 2021; dd = 00;
}

🍰 Open and Free

Open Source project; MIT License, Enjoy 🎉

License

The MIT License (MIT)

Copyright (c) 2013-2021  Denis Kuzmin <x-3F@outlook.com> github/3F

[ ☕ Make a donation ]

regXwild contributors: https://github.com/3F/regXwild/graphs/contributors

We're waiting for your awesome contributions!

Speed

Procedure of testing

  • Use the algo subproject as tester of the main algorithms (Release cfg - x32 & x64)
  • In general, calculation is simple and uses average as i = (t2 - t1); (sum(i) / n) where:
    • i - one iteration for searching by filter. Represents the delta of time t2 - t1
    • n - the number of repeats of the matching to get average.

e.g.:

{
    Meter meter;
    int results = 0;

    for(int total = 0; total < average; ++total)
    {
        meter.start();
        for(int i = 0; i < iterations; ++i)
        {
            if((alg.*method)(data, filter)) {
                //...
            }
        }
        results += meter.delta();
    }

    TRACE((results / average) << "ms");
}

for regex results it also prepares additional basic_regex from filter, but of course, only one for all iterations:

meter.start();

auto rfilter = tregex(
    filter,
    regex_constants::icase | regex_constants::optimize
);

results += meter.delta();
...

Please note:

  • +icase means ignore case sensitivity when matching the filter(pattern) within the searched string, i.e. ignoreCase = true. Without this, everything will be much faster of course. That is, icase always adds complexity.
  • Below, MultiByte can be faster than Unicode (for the same platform and the same way of module use) but it depends on specific architecture and can be about ~2 times faster when native C++, and about ~4 times faster when .NET + Conari and related.
  • The results below can be different on different machines. You need only look at the difference (in milliseconds) between algorithms for a specific target.
  • To calculate the data, as in the table below, you need execute algo.exe

Sample of speed for Unicode

340 Unicode Symbols and 10^4 iterations (340 x 10000); Filter: L"nime**haru*02*Magica"

algorithms (see impl. from algo) +icase [x32] +icase [x64]
Find + Find ~58ms ~44ms
Iterator + Find ~57ms ~46ms
Getline + Find ~59ms ~54ms
Iterator + Substr ~165ms ~132ms
Iterator + Iterator ~136ms ~118ms
main :: based on Iterator + Find ~53ms ~45ms
​ ​
Final algorithm - EXT version: ~50ms ~26ms
Final algorithm - ESS version: ~50ms ~27ms
​ ​
regexp-c++11(regex_search) ~59309ms ~53334ms
regexp-c++11(only as ^match$ like a '==') ~12ms ~5ms
regexp-c++11(regex_match with endings .*) ~59503ms ~53817ms

ESS vs EXT

350 Unicode Symbols and 10^4 iterations (350 x 10000);

Operation (+icase) EXT [x32] ESS [x32] EXT [x64] ESS [x64]
ANY ~54ms ~55ms ~32ms ~34ms
ANYSP ~60ms ~59ms ~37ms ~38ms
ONE ~56ms ~56ms ~33ms ~35ms
SPLIT ~92ms ~94ms ~58ms ~63ms
BEGIN --- ~38ms --- ~19ms
END --- ~39ms --- ~21ms
MORE --- ~44ms --- ~23ms
SINGLE --- ~43ms --- ~22ms

For .NET users through Conari engine:

Same test Data & Filter: 10^4 iterations

Release cfg; x32 or x64 regXwild (Unicode)

Attention: For more speed you need upgrading to Conari 1.3 or higher !

algorithms (see impl. from snet) +icase [x32] +icase [x64]
regXwild via Conari v1.2 (Lambda) - ESS ~1032ms ~1418ms x
regXwild via Conari v1.2 (DLR) - ESS ~1238ms ~1609ms x
regXwild via Conari v1.2 (Lambda) - EXT ~1117ms ~1457ms x
regXwild via Conari v1.2 (DLR) - EXT ~1246ms ~1601ms x
​ ​
regXwild via Conari v1.3 (Lambda) - ESS ~58ms ~42ms <<
regXwild via Conari v1.3 (DLR) - ESS ~218ms ~234ms
regXwild via Conari v1.3 (Lambda) - EXT ~54ms ~35ms <<
regXwild via Conari v1.3 (DLR) - EXT ~214ms ~226ms
​ ​
.NET Regex engine [Compiled] ~38310ms ~37242ms
.NET Regex engine [Compiled]{only ^match$} < 1ms ~3ms
.NET Regex engine ~31565ms ~30975ms
.NET Regex engine {only ^match$} < 1ms ~1ms

How to get regXwild

regXwild v1.1+ can also be installed through NuGet same for both unmanaged and managed projects.

For .NET it will put x32 & x64 regXwild into $(TargetDir). Use it with your .net modules through Conari and so on.

x64 + x32 Unicode + MultiByte modules;

Please note: Modern regXwild packages will no longer be distributed together with Conari. Please consider to use it separately, Conari nuget packages.