Merge pull request #47 from BIDS-numpy/status-2019-09-18

Save status meeting Note 2019-09-18

Author: seberg

status_meetings/status-2019-09-18.md

@@ -0,0 +1,79 @@
+---
+tags: NumPy
+---
+
+# 2019-09-18 NumPy Development Meeting
+
+- Time: 11am Pacific Time
+- Join via Zoom at https://berkeley.zoom.us/j/762261535
+- [Trello workboard](https://trello.com/b/Azg4fYZH/numpy-at-bids)
+- [Previous meetings](https://github.com/BIDS-numpy/docs/tree/master/status_meetings)
+
+**Present:** Inessa, Hameer, Stéfan, Sebastian, Matti, Warren, Ralf, DSM
+
+## Follow-up from last meeting / discussions
+
+- In discussing build warnings, @rgommers raised the following ideas:
+  - added a `--debug-config` option to show this output [PR 14518](https://github.com/numpy/numpy/pull/14518)
+  - introduce a CI run that's clean with the latest GCC (existing build matrix entry could be updated). See [PR 14527](https://github.com/numpy/numpy/pull/14527) which adds -Werror to the build phase as an option to runtests
+  - introduce a CI run that's clean on macOS with Clang (lots of warnings right now). [PR ](https://github.com/numpy/numpy/pull/14534) made a start, but clang 10 is pretty aggresive
+    - clang on linux/windows too?
+
+# Topics
+
+- Using AVX in non-ufunc code [PR 14497](https://github.com/numpy/numpy/pull/14497)
+    - We cannot just drop support for older CPUs
+    - Much increased complexity
+    - [Better PR for SIMD support](https://github.com/numpy/numpy/issues/13393) after IBM bounty
+
+- Supporting low-level cython `pxd` routines from `random` as a public API [issue 14517](https://github.com/numpy/numpy/issues/14517)
+  - There is no plan/NEP/documented API
+  - Get in touch with Robert to find out whether this was an intentionally supported use-case
+
+- Moving/copying all meeting notes and presentations/proposals to the numpy github organization. We should check the permissions needed for the hackmd.io integration before enabling it. Mailing list discussion.
+  - archive repo
+
+- *NEP 32 - Remove financial functions.* There have been no more responses on the mailing list in the last week.  Unless there are new objections, after the meeting Warren will send an email to the mailing list with the proposal to accept NEP 32, in accordance with ["How a NEP becomes accepted"](https://github.com/numpy/numpy/blob/master/doc/neps/nep-0000.rst#how-a-nep-becomes-accepted).
+    - Send email to formally accept the NEP.
+
+- [Storing dtype metadata issue](https://github.com/numpy/numpy/issues/14142):
+    - h5py new uses metadata on the dtype
+    - Arrays saved as `npy` which have such dtypes cannot be loaded again.
+    - Should we try to allow it? Raise an error? (what if `safe_eval` cannot load the metadata).
+
+- Reverting changes to legacy code in random [gh-14531](https://github.com/numpy/numpy/pull/14531)
+  - we would prefer to merge Warren's PR at this point, if we get more problems we will reconsider.
+
+- [NumPy User Stories](https://mail.python.org/pipermail/numpy-discussion/2019-September/080066.html) for https://numpy.org
+
+- NumPy logo refresh
+  - The issue about logo redesign: [numpy.org/37](https://github.com/numpy/numpy.org/issues/37)
+
+  - (Matti requested that NumPy appear on it somehow)
+
+- The NumPy name is going through a trademark review and with the help of NumFocus may be trademarked.
+
+- Last year's report from the team at BIDS [is available on GitHub](https://github.com/BIDS-numpy/docs/tree/master/reports). Next report is due October.
+
+## Additional Details
+
+- Matti
+  - adding options to distutils commands to quiet down the build_src phase (all those configure tests) and allow -Werror in the actual build
+  - first cut at the annual report document
+  - scipy numpy api update [PR 10840](https://github.com/scipy/scipy/pull/10840)
+ 
+- Sebastian
+    - [Extensible Datatypes for NumPy](https://hackmd.io/kxuh15QGSjueEKft5SaMug) 
+    - Started (prototyping): https://github.com/seberg/numpy/pull/6
+        - Dtype discovery is largely working, a bit slow (still faster that current one for integers).
+    - TODO: Add some pseudocode examples for example for how promotion will work or how dtype discovery works.
+
+- Warren
+  - Repo for my experiments with custom dtypes using the current API is https://github.com/WarrenWeckesser/numtypes
+  - [Repo for notes about NumPy](https://github.com/WarrenWeckesser/numpy-notes)
+    These notes will focus on using the NumPy C API.
+
+---
+
+Please remember to archive this file and commit it to github.com:bids-numpy/docs
+