---
tags: NumPy, BIDS
---

# 2019-3-06 BIDS/NumPy Development Meeting

- Join via Zoom at [https://berkeley.zoom.us/j/400054438](https://berkeley.zoom.us/j/400054438), or dial-in through a [local phone number](https://zoom.us/u/adQDmEc1wI).
- [Trello workboard](https://trello.com/b/Azg4fYZH/numpy-at-bids)
- [Previous meetings](https://github.com/BIDS-numpy/docs/tree/master/status_meetings)
- [NumPy at BIDS Calendar](https://calendar.google.com/calendar?cid=YmVya2VsZXkuZWR1X2lla2dwaWdtMjMyamJobGRzZmIyYzJqODFjQGdyb3VwLmNhbGVuZGFyLmdvb2dsZS5jb20)

**Present:** Hameer, Ralf, Matti, Tyler, Stéfan, Oleksandr, ?Guest?

## Follow-up from last meeting/discussions

- sorting progress?
- Timsort currently appears to be superior--is `radixsort` still worth adding? Timsort replaces mergesort, but radixsort would be more intrusive to ABI. Conclusion: let's just keep the PR for now, could be useful in future!
- TODO: Document contact person and how we access the gcc build farm (and other similar resources).
- Matti, Ralf and Tyler have access to the [build farm](https://cfarm.tetaneutral.net/machines/list/) via ssh. Matti is looking at the [unpackbits pr](https://github.com/numpy/numpy/pull/12962) on a MIPS (little-endian) machine. Ralf has started looking into AIX build issues.
- TODO: give core devs access to a private repo with debug machine access info (gcc farm, etc.)
- Cannot host this on GitHub, they only allow 3 private collaborators
- This now lives at https://gitlab.com/numpy/logistics.git
- Intel distribution of NumPy---have they started looking at issues to "poach?": https://github.com/IntelPython/numpy
- should we publish that repo somewhere? In an issue template perhaps (mattip)

## Community Suggested Topics

- Outreachy: NumPy is now a mentoring organization. Deadline for project definition and adding mentors: March 12th. Sebastian + Matti have agreed to mentor.
- Hameer:
- Add mixins for easier duck-typing of arrays.
- PR: https://github.com/numpy/numpy/pull/10460
- there's an open issue here that requires more discussion: https://github.com/numpy/numpy/issues/10446
- Would be nice to turn `np.where` into a ufunc, but I don't know how yet.
- https://github.com/numpy/numpy/issues/8994
- The ufuncs are generated via a python script in `core/code_generators/generate_umath.py`, which spits out c code. Usually the loops are written in `core/umath/loops*` or so, as a set of `*.c.src` files. To add one,

## Work status

Matti:
- Turning on `-n` for sphinx-build revealed over 1000 bad references in the documentation. Example: the [SeeAlso](http://www.numpy.org/devdocs/reference/generated/numpy.polynomial.chebyshev.chebroots.html) section.
- numpydoc fix may be needed re: backtick usage (currently, `arg` is not interpreted, leading to a "missing reference" in Sphinx)
- is parameter and return value formatting documented in numpydocs? It should parse and mark these for special formatting, but other things (See Also and others) must be references.
- Unfortunately I was sick for most of the week with a bad throat/cold, and could not really function. A bit better today.

Tyler:
- trying to push [Eric's `np.clip() ufunc` PR](https://github.com/numpy/numpy/pull/12519) forward a bit by:
- adding tests
- addressing reviewer comments where possible
- some `datetime64` PRs remain challenging:
- [changes in `int->m8` casting behavior](https://github.com/numpy/numpy/pull/13061)
- [`datetime64` span chokes/ silent wrapping](https://github.com/numpy/numpy/pull/11873)
- [draft of `nansum` using `_nan_mask()`](https://github.com/numpy/numpy/pull/12801) has been a little underwhelming so far
- a variety of other efforts in NumPy & SciPy, including reviewing PRs for SciPy paper
- Connected with Carl Kleffner about SciPy Fortran compiler
- [OpenBLAS version detection PR](https://github.com/numpy/numpy/pull/12790) by Tyler is working, needs review. Downside currently is that it generates an extra DLL, is this acceptable?

---

Please remember to archive this file and commit it to github.com/bids-numpy/docs