Hi Leonard,
I tried the 4.13 branch out on one of our content processing pipelines at my day job. It processes about 27k HTML files with <h1>/<h2>/etc. hierarchy and extracts about 80k HTML fragments from them.
The runtime results are:
4.12.2
586 seconds (average of three runs)
4.13
574 seconds (average of three runs)
Impressive - even with the extra type checking, 4.13 is a bit faster! I did follow-up testing to confirm.
The output HTML fragments generated by the pipeline are identical for both versions.
I hope this is helpful,
- Chris