Pablo Winant

Actually, the reason I mentioned the outter loop as part of the problem is that sometime you want to write the whole algorithm as a function of 'f', so there...

@spencerlyon2: Julia clearly has a point in making this kind of code design much easier to implement. Hopefully, the Python side will catchup. While numba continues to improve I wonder...

I have implemented an experimental version of a class wrapper there: https://github.com/albop/numba_experiments/blob/master/Using%20a%20class%20to%20define%20functions%20with%20numba.ipynb. It is rather simple and basically creates a new class like you would do manually. _In principle_ one...

@oyamad: thank you for the heads-up. I had missed the recent updates. Actually, I remember reading another issue about jit-classes where methods starting with underscores where mentioned but I can't...

Very cool indeed ! Just when I had neglected to look at numba release notes. It's a very good idea to provide simple numba-compatible optimizers. In the short term there...

@mmcky: where does this one stands ? I just realized that I may have forgotten to make the gist public.

Good catch about the email ! That was careless of me... I honestly don't know if it's ready to merge. This was more a proof of concept and should probably...

Thank you @vchuravy ! I did what you suggest and looked at the generated code, but there is no useful information there as most of the logic seems to be...

I missed the beginnning of this discussion and haven't looked closely at the code, but regarding point 1: 1/ I would be very curious to see what can be done...

I shall add that dealing with chunks in an efficient way, is precisely what `blaze` is supposed to be able to do. Another option to consider is `xarray+dask`. Actually getting...