PyScript performance

[almarklein]

I did some pystone tests some time ago, and then PyScript was much faster than cPython. Did these benchmarks again, and now its about the same. What feature that we added makes things so slow? Is it something significant, or just something that the test emphasizes on?

See also #288

[almarklein]

One option that I've always had in mind for this, and that's also used by Transcrypt, is to allow a kind of pragma, probably in the form of a with statement to create blocks of "pure JS", where e.g. operator overloading is not used.

Another approach can be to allow type annotations, and to infer types from literals, and thereby be able to take some shortcuts. Possibly use mypy to get more type information via flow analysis. The advantage of this is that its easier to "apply" from the user point of view, without the risk of a user going into "pure JS" mode but not knowing JS enough to oversee the pitfalls.

[almarklein]

Note to self: Transform.map in CG project would be a nice use-case for PyScript to produce efficient code without manual intervention.

[almarklein]

Issue moved to flexxui/pscript #11 via ZenHub