Merge pull request #38 from BIDS-numpy/2019-06-27-notes

add notes from 2019-06-27

Author: mattip

status_meetings/status-2019-06-27.md

@@ -0,0 +1,49 @@
+---
+tags: NumPy
+---
+
+# 2019-06-27 NumPy Development Meeting
+
+- 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:** Sebastian, Matti, Ralf, Chuck, Hameer
+
+## Follow-up from last meeting / discussions
+- Talk for SciPy2019
+  - We need data to show how the grant has affected the project
+    - Nelle responded to a private mail but has not yet provided data
+  - Talk so far is in [the repo](https://gitlab.com/numpy/scipy2019-presentation-insidenumpy)---note that this is a private repo
+  - there are some links in the repo to scripts that generate plots
+
+- 1.17 release upcoming: review PRs/Issues marked with [1.17 milestone](https://github.com/numpy/numpy/milestone/62). At last count there are 9 open issues and 5 open PRs with the milestone, maybe just start bumping issues at this point
+  - [PR #13764](https://github.com/numpy/numpy/issues/13764) waiting for approval, moves windll code to a `_distributor_init.py` file when using OpenBlas on Windows
+  - 4 of the PRs have to do with random, see below.
+  - Memory leak needs to be fixed: https://github.com/numpy/numpy/issues/13853
+  - Document `__array_function__` https://github.com/numpy/numpy/issues/13844
+
+- Random generator work
+    - How many BitGenerators to include: the general concensus is to include a minimum rather than a maximum, and to use a new BitGenerator as the default. Now we are left with 3, one more will be added in (a chaotic one) 
+      - Created numpy/bitgenerator repo
+      - [PR to package](https://github.com/mattip/numpy/pull/40) the required support files to subclass BitGenerator in an external repo. Where should the new header file go?
+      - move into somewhere inside the return value from `numpy.get_include()``
+  - PR for SeedSequence as a way to seed the BitGenerators should be merged by the time this meeting happens.
+
+    
+# Topics
+
+- Towncrier usage?
+  - need a PR to set up the infrastructure. No strong opinions, we should try it out with some reorganization of the file.
+  -  It will not be a complete automated solution, there will be more mork needed to polish the final results.
+
+## Additional Details
+
+- Matti
+  - made the numpy/bitgenerator repo function
+  - more work around SeedSeqeunce. BitGenerators
+- Sebastian
+  - dtypes: prototype seems to be the correct direction. The dtypes heirarchy seems to make sense, but what to do with multiple inheritance ( unit-meters and int32 together?)
+      - Hameer: Attribute on `dtype` plus a type factory. :smile:
+          - Background: There are languages that completely reject the multiple-inheritance approach, e.g. C#, because it can lead to a mess of a design. Much nicer is to have something like a template. The closest thing to a template in Python is `__class_getitem__`, but that is not recommended beyond static type-checking purposes. The other option is a type-factory that sets class attributes.
+