---
tags: NumPy, BIDS
---

# 2019-3-20 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:** Stefan, Matti, Tyler, Chuck, Hameer

## Follow-up from last meeting/discussions

- Report on the [random number generator refactor call](https://github.com/BIDS-numpy/docs/blob/master/meetings/2019-03-19_random-number-generator-refactor.md)
- Not clear if it will be ready for 1.17 - Chuck is beginning to think about 1.17 release.
- Reminder of developer meeting, 10/11 May
- The number of open PRs is down near 240. Nice!!
- Stéfan's one humble open PR (`constant` as default mode for padding) [got merged](https://github.com/numpy/numpy/pull/4808); a straightforward follow-up to review would be [#11358](https://github.com/numpy/numpy/pull/11358)
- UArray: For overriding and duck-typing: https://github.com/Quansight-Labs/uarray

## Community Suggested Topics

- Do we want a release calendar for NumPy?
- We could add aims for bigger scope features we'd like to have included in each.
- S: Perhaps can live with the roadmap?
- Precision error on wheels as of late
- Traced back to merging of FFT refactor
- Shows up only on 32-bit, not on Windows [link](https://travis-ci.org/MacPython/numpy-wheels)
- Manylinux2010

- Matthew's thought on timeline:

> Maybe build the wheels when the image comes out, along with the manylinux1 wheels, and switch to building 2010 wheels only, in a year or so?

- Images are not out yet; when they come out, 32-bit will initially not be supported

## Work status

Matti:
- Outreachy applicants are starting to appear
- Documentation is almost down to no warnings, due to a one-line [numpydocs fix](https://github.com/numpy/numpydoc/pull/201) and a [fix](https://github.com/numpy/numpy/pull/13119) for repeat c function definitions. The next step is to turn on `-WT --keep-going` to make sure commits that add new warnings fail.
- Started work on [merging randomgen](https://github.com/numpy/numpy/pull/13163), learned lots about git and [merging two unrelated repos](https://github.com/numpy/numpy/pull/13163/files#diff-e84dba42a7326dfd8f7cc3e8f2957551). Now getting the cython code to build.
-

Tyler:
- reducing my open PRs: SciPy (1) & NumPy (7) based on feedback from last meeting; that said, I'm not a big fan of this---this may have already reduced visibility of potentially useful work for which reviewer bandwidth simply isn't there; one of my closures was already re-opened
- DevOps efforts
- SciPy
- ([merged](https://github.com/scipy/scipy/pull/9940)) removed extraneous NumPy distutils copies from SciPy Azure CI
- ([merged](https://github.com/scipy/scipy/pull/9774)) Azure CI OpenBLAS versions now **finally** match scipy-wheels. The trick was to match mingw64 versions more closely with the wheels workflow
- NumPy
- ([merged](https://github.com/numpy/numpy/pull/12457)) NumPy now uses wheels-matched OpenBLAS backend in MacOS CI testing (previously, **no** linear algebra backend was enabled)
- ([open](https://github.com/numpy/numpy/pull/12790)) Thanks to the above addition of MacOS OpenBLAS usage, all 3 platforms under test on Azure CI now have compile-time & post-mortem OpenBLAS version availability in PR 12790
- Fixing Bugs
- NumPy
- ([open](https://github.com/numpy/numpy/pull/13061)) Working to address reviewer comments to prohibit certain undesirable integer casting operations that run through `timedelta64` casts, but this will probably require some decisions about issuing warnings and what will effect on i.e., pandas be?

## PR count

Top 10 PR contributors:

1. eric-wieser: 32
2. jaimefrio: 10
3. juliantaylor: 6
4. madphysicist: 6
5. mattip: 5
6. bashtage: 4
7. tylerjereddy: 4
8. zerothi: 4
9. pitrou: 3
10. hameerabbasi: 3

PRs for eric-wieser:

- #12519: [ENH/DEP: Use a ufunc under the hood for ndarray.clip](https://github.com/numpy/numpy/pull/12519)
- #12473: [ ENH: Make the contents of dtype.fields namedtuple-like](https://github.com/numpy/numpy/pull/12473)
- #12467: [WIP: Make type(np.generic) != type](https://github.com/numpy/numpy/pull/12467)
- #12205: [MAINT: avoid relying on `np.generic.__name__` in `np.dtype.name`](https://github.com/numpy/numpy/pull/12205)
- #12178: [BUG: Prevent segfaults if python runs out of memory concatenating strings in ufunc error messages](https://github.com/numpy/numpy/pull/12178)
- #11955: [MAINT: Move _do_append and _do_prepend out of the functions to generate padding values](https://github.com/numpy/numpy/pull/11955)
- #11924: [WIP/NEP: Making details of ``np.core`` private](https://github.com/numpy/numpy/pull/11924)
- #11829: [MAINT: Use textwrap.dedent in f2py tests](https://github.com/numpy/numpy/pull/11829)
- #11397: [DEP: Deprecate the normed argument to histogramdd to match histogram](https://github.com/numpy/numpy/pull/11397)
- #10765: [BUG: Disable hex(np.floating) and oct(np.floating) on python 2](https://github.com/numpy/numpy/pull/10765)
- #10741: [ENH: Implement `np.floating.as_integer_ratio`](https://github.com/numpy/numpy/pull/10741)
- #10723: [BUG: longdouble(int) does not work](https://github.com/numpy/numpy/pull/10723)
- #10417: [ENH: Allow dtype objects to be indexed with multiple fields at once](https://github.com/numpy/numpy/pull/10417)
- #10308: [API: Make MaskedArray.mask return a view, rather than the underlying mask](https://github.com/numpy/numpy/pull/10308)
- #10151: [BUG: Numpy scalar types sometimes have the same name](https://github.com/numpy/numpy/pull/10151)
- #10132: [BUG: np.matrix does not return an array when given a full index ending in …](https://github.com/numpy/numpy/pull/10132)
- #10025: [ENH: Add the ndmin argument from np.array to np.as[any]array](https://github.com/numpy/numpy/pull/10025)
- #9963: [ENH: Add the C99 remainder function, as `np.remainder_ieee`](https://github.com/numpy/numpy/pull/9963)
- #9951: [ENH: Make dtype iterable](https://github.com/numpy/numpy/pull/9951)
- #9877: [WIP: Make str → bool conversion consistent](https://github.com/numpy/numpy/pull/9877)
- #9705: [BUG: Fix some (not all) compiler warnings](https://github.com/numpy/numpy/pull/9705)
- #9690: [BUG: Fix broken out argument in np.ma ufunc_likes](https://github.com/numpy/numpy/pull/9690)
- #9459: [MAINT: Remove unhelpful docstrings, improve helpful ones](https://github.com/numpy/numpy/pull/9459)
- #9428: [ENH: add a neigborwise function](https://github.com/numpy/numpy/pull/9428)
- #9353: [BUG: Fix np.ma.take on un-indexable scalars](https://github.com/numpy/numpy/pull/9353)
- #8970: [ENH/API: Change flexible types to indicate resizability with elsize == -1](https://github.com/numpy/numpy/pull/8970)
- #8910: [ENH: avoid allocations in getmaskarray](https://github.com/numpy/numpy/pull/8910)
- #8868: [MAINT: Split ufunc into two subtypes, and remove methods where innapropriate](https://github.com/numpy/numpy/pull/8868)
- #8853: [BUG: Prevent power(uint64, int64) casting to float](https://github.com/numpy/numpy/pull/8853)
- #8662: [ENH: preserve subclasses in ufunc.outer](https://github.com/numpy/numpy/pull/8662)
- #8641: [BUG: Preserve types of empty arrays in ix_ when known](https://github.com/numpy/numpy/pull/8641)
- #8640: [ENH: generalize `ix_` to work for arbitrary dimensionality of input](https://github.com/numpy/numpy/pull/8640)
- #8615: [ENH: Add support for coercing structured arrays into dicts](https://github.com/numpy/numpy/pull/8615)
- #8514: [ENH: Make it possible to call .view on object arrays](https://github.com/numpy/numpy/pull/8514)
- #8511: [MAINT: make np.ma.apply_along_axis consistent with np.apply_along_axis](https://github.com/numpy/numpy/pull/8511)

PRs for jaimefrio:

- #9424: [ENH: Adds an out argument to bincount.](https://github.com/numpy/numpy/pull/9424)
- #7473: [BUG: shift operator cycles, fixes #2449](https://github.com/numpy/numpy/pull/7473)
- #7464: [ENH: Make bincount output same type as weights, fixes #6854](https://github.com/numpy/numpy/pull/7464)
- #7450: [BUG, MAINT: concatenate of empty sequences, fixes #1586](https://github.com/numpy/numpy/pull/7450)
- #7441: [MANT: refactor PyArrayMultiIterObject constructors](https://github.com/numpy/numpy/pull/7441)
- #6219: [MAINT: refactor unravel_index for code repetition](https://github.com/numpy/numpy/pull/6219)
- #5890: [BUG: Don't use longlong for PyLong if it isn't larger than long](https://github.com/numpy/numpy/pull/5890)
- #5740: [MAINT: Refactor .flags attribute of ndarray](https://github.com/numpy/numpy/pull/5740)
- #5508: [WIP: make views of different type checks less strict](https://github.com/numpy/numpy/pull/5508)
- #5487: [WIP: handling of nan in unique, fixes #2111](https://github.com/numpy/numpy/pull/5487)
- #4366: [BUG: Allow any integer type in bincount (fixes #823)](https://github.com/numpy/numpy/pull/4366)
- #4246: [BUG: `take` casting logic with an `out=` argument](https://github.com/numpy/numpy/pull/4246)

PRs for juliantaylor:

- #11113: [AVX for floats?](https://github.com/numpy/numpy/pull/11113)
- #9163: [BUG: use data malloc for transfer function data](https://github.com/numpy/numpy/pull/9163)
- #8924: [ENH: add scalar special cases for boolean logical loops](https://github.com/numpy/numpy/pull/8924)
- #8856: [MAINT: remove duplicated integer code](https://github.com/numpy/numpy/pull/8856)
- #8783: [ENH: cache large memory allocations on systems with MADV_FREE](https://github.com/numpy/numpy/pull/8783)
- #7865: [WIP: use openmp vector function ABI for sin, cos, exp and log](https://github.com/numpy/numpy/pull/7865)
- #4904: [more small-array performance improvements](https://github.com/numpy/numpy/pull/4904)

PRs for madphysicist:

- #12829: [ENH: Added option to suppress stout/err capture in tests](https://github.com/numpy/numpy/pull/12829)
- #12828: [DOC: Updates to nditer usage instructions](https://github.com/numpy/numpy/pull/12828)
- #10820: [ENH: Added support for arrays with `dtype=object` to `np.isinf`, `np.isnan`, `np.isfinite`](https://github.com/numpy/numpy/pull/10820)
- #9514: [ENH: Adding np.neighborwise function](https://github.com/numpy/numpy/pull/9514)
- #7804: [ENH: Added atleast_nd](https://github.com/numpy/numpy/pull/7804)
- #7322: [ENH: Added optional `scale` parameter to `sinc` to enable non-normali…](https://github.com/numpy/numpy/pull/7322)

PRs for mattip:

- #13163: [WIP, ENH: randomgen](https://github.com/numpy/numpy/pull/13163)
- #13005: [MAINT: remove warning for pre-1.9 numpy](https://github.com/numpy/numpy/pull/13005)
- #12962: [ENH: Add 'order' keyword to packbits, unpackbits](https://github.com/numpy/numpy/pull/12962)
- #12714: [WIP, ENH: extend np.vectorize for flexible NEP20 signatures](https://github.com/numpy/numpy/pull/12714)
- #12660: [WIP, NEP: add dtype design NEP - DtypeMeta, Dtype, and hierarchy](https://github.com/numpy/numpy/pull/12660)
- #12630: [WIP, NEP: add dtype design NEP](https://github.com/numpy/numpy/pull/12630)
- #12619: [WIP, NEP: draft dtype ufunc NEP](https://github.com/numpy/numpy/pull/12619)
- #12609: [DEP: issue warning when creating 0-stride broadcasted arrays](https://github.com/numpy/numpy/pull/12609)
- #12594: [WIP, DEV: add pypy3 to azure CI](https://github.com/numpy/numpy/pull/12594)
- #12585: [WIP: refactor dtype to be a type subclass](https://github.com/numpy/numpy/pull/12585)
- #12428: [ENH: always use zip64, upgrade pickle protocol to 3](https://github.com/numpy/numpy/pull/12428)
- #12284: [WIP: ENH: supply our version of numpy.pxd, requires cython>=0.29](https://github.com/numpy/numpy/pull/12284)
- #11684: [BUG: Raise when unravel_index, ravel_multi_index are given empty input](https://github.com/numpy/numpy/pull/11684)

---

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