Jukka Lehtosalo
Jukka Lehtosalo
Right now mypyc uses lots of things from CPython (e.g. the stdlib, built-in types, and the cyclic GC). Not relying on CPython would require writing a new language runtime and...
> Maybe mypyc could just bundle the CPython parts required for running the compilation product? This could be doable. Maybe something similar to Cython's `--embed` option could do the trick...
The are no plans right now. If somebody would contribute a simple enough implementation, we might accept it.
`Dynamic[X, Y]` sounds similar to "unsafe unions" that have been discussed previously. `X` would be compatible with `Dynamic[X, Y]`, and vice versa. `Dynamic[X]` could potentially be represented as an unsafe...
If the item type is erased (e.g., a non-primitive Python class or a callable type), or too complex (possibly unions), a packed array can still be used, but the runtime...
A few updates: * I think I'd want to support negative indexing after all (e.g. `a[-1]`), since it's such a common thing in Python and shouldn't have a high runtime...
One more thing: Maybe packed arrays should be hashable even though they are mutable. Otherwise we'd probably need a separate frozen packed array type, and to use arrays as hash...
> I am pretty sketched out by that. Python is pretty consistent about that rule. That's fair. Variable-length tuples as dict keys aren't very common, so initially we could survive...
Increasing priority since this is a pretty basic Python feature that we don't support.
Thanks for the detailed bug report! I can confirm that this fails to compile with gcc on Linux.