scoder

> is there any reason why you wouldn't do a `3.1.x` release? It's not ready yet.

Thanks for the report. This is really two issues in one. It's actually a feature request, but until we support that, we should forbid it since we currently generate invalid...

Hi. That seems a nice offer, thanks. I looked through the (remote) PR and I can imagine where the code is coming from. However, the generated unit tests look very...

> It would be nice if we could have the long and long long versions too, but unfortunately that results in "ambiguous overloaded method" on any call of float, even...

> otherwise things [like] `1.0 + round(x)` would use Python conversions because type analysis would have already been done assuming the output was a PyObject. We still apply the optimisation...

I integrated a feature into 3.1 that will make this implementation easier. See https://github.com/cython/cython/commit/99e9df17d2cabf5714eec9e844144720aea81a9d You can now generate a dedicated function corresponding to the call arguments when encountering a specific...

I second the proposal to extract the optional part into a separate module. You mentioned that you can use NumPy and SciPy instead, so there's apparently some functionality that you...

I think you can just slice the array to provide an explicit length when iterating over it.

I'd say it's a gotcha. We should do something about it, if possible. It might be worth a warning.

The crashes apparently happen while running Cython as pure Python code that translates a Python module in Cython itself. That suggests a bug in CPython rather than Cython. Recent work...