For our new PCRE2CodeMBS class, we add a new Matches method. It returns an iterator, so you can use it with a for-each loop in Xojo:
One of the key elements ot reach best performance is to reuse the PCRE2CodeMBS objects and avoid compilation as much as you can. Second, you avoid creating PCRE2MatchDataMBS objects and reuse them. The iterator does that by default for you.
Let us show you the numbers from our benchmark:
|RegEx Xojo||RegEx MBS||PCRE2||PCRE2+JIT||PCRE2 Iterator||PCRE2 Iterator+JIT|
|Match 500 in 1 MB||171317||3478||3521||266||3231||277|
|Match 500 in 10 MB||1711241||30627||30742||928||30542||1063|
|Match 5000 in 1 MB||1642035||3234||3776||434||3660||478|
|Match 5000 in 10 MB||16834198||30906||31200||1363||31157||1456|
Tests made with MBS Xojo Plugins 22.2pr and Xojo 2022r2.1 on a current MacBook Pro. The Xojo value is a 10 times average while the MBS calls are averaged over running 100 times. But at the quick turn around times of the compiled code, it gets hard to measure!
If you haven't tried yet, please give the PCRE2 classes with just-in-time compiler a try. They are often 10 times faster compared to our already quick RegExMBS class.