This improves the regex evaluation back to when we still had \b in front, while (hopefully) doing as it should.
For the test referenced in #331, this change drops us back to 90ms, which is still twice as slow as originally, but there's also more features supported now, so I find this acceptable.
@metoule Please check that this is doing what you meant. If I understood your comment about \b right, you had to remove it because that also matched @, so even though @? is included in the identifier part it was never captured. If so, then this lookbehind should model the same as \b but it respects @ in front. Demo: https://regex101.com/r/WhA5GM/2
This improves the regex evaluation back to when we still had
\b
in front, while (hopefully) doing as it should. For the test referenced in #331, this change drops us back to 90ms, which is still twice as slow as originally, but there's also more features supported now, so I find this acceptable.@metoule Please check that this is doing what you meant. If I understood your comment about
\b
right, you had to remove it because that also matched@
, so even though@?
is included in the identifier part it was never captured. If so, then this lookbehind should model the same as\b
but it respects@
in front. Demo: https://regex101.com/r/WhA5GM/2Close #291