Mark Dufour

basic test added. leaving this open, since the 'format specifiers' (such as !r or :.04f) are not yet supported. if anyone might be interested, feel free to implement them!

just adding some interesting articles I'm finding on the topic. https://www.gingerbill.org/series/memory-allocation-strategies/ perhaps a simple escape analysis in combination with an existing stack allocator (as in part 3 of this article)...

on the C++ side, something like this may actually be sufficient (just the rough idea): ``` class StackFrame { StackFrame() { this->SP = SP; } ~StackFrame() { SP = this->SP;...

so we maintain a separate memory stack on which to allocate objects. if a function wants to do stack allocation, we just add 'StackFrame __sss'. this stores the SP on...

pushed one commit to handle ast.AnnAssign at least, solving some cases. next would be to look at attribute assignment (leading to the error about 'ctx' probably).

done: ignoring 'typing' import, pure annotation-assignment..

argh, it appears I was measuring this incorrectly somehow.. :S commit reverted. keeping the issue open to have another look at the 3x slowness caused by (unused!!) exception handling in...

I was unable to get around the slowness, too bad.

for 0.9.9 at least update the docs with missing functionality. so first we need to get a precise overview.

okay the docs seem uptodate: only accumulate, pairwise and starmap are (still) missing.