Joey

> As for "an exact test case", I'm running the code at the start of this thread. Oh yes of course sorry, I forgot 😅

@elliott5 Isn't a new copy required each time for append because append copies the slice?

@elliott5 I see the issue now thank you! Should only at most have 1 allocation per append, right now it's possible to have 2 and the copy is premature at...

Once utf8 passes this issue should be closed as the init code for utf8 test relies on this benchmark being fast enough to not timeout the test runner.

@elliott5 yes a new issue for the capacity difference of bytes buffer would be great.

Nice! I'll close this issue as it's now in the performance optimization realm not bug fixing.

Hey Sonygod, thanks so much for showing interest! It's currently still very expirimental and unable to do full library conversions, in most part because go's low level standard library is...

> One idea to consider, if go2hx is not bound by Go's memory layout (as TardisGo was, described above), is that Go uintptr perhaps does not have to be implemented...

The approach above was taken and this issue is resolved.

Thank you for the great advice @elliott5 ! Plugging into the go docs and auto generating the standard library would be superb for working with the generated code, even if...