Thoughts for the next survey

[seberg]

Just an issue to start on thoughts for a next survey. I know we already mentioned more in many other places, but I can't think of them right now, so hopefully we will just add things when we remember:

• "What sizes of arrays do you usually work with". We get a lot of "performance!" comments, but IMO, those are tricky, since while everyone can see that NumPy doesn't use multiple cores, there is no point in asking us to do that for mid-sized arrays.
• Rather than rank "importance" on issues such as Performance and Reliability (which is good), I think we should probably ask more directly if they think there is need to improve. And e.g. on the tradeoff between consistency/cleanup/new features vs. "stability".

[rgommers]

And e.g. on the tradeof between consistency/cleanup/new features vs. "stability".

This one I particularly like as a new question.

[seberg]

So the current phrasing for the priority was: "should be given the highest priority for the NumPy project in the next 12 months".

Another thought might be to phrase it in terms of additional resources (or modification of the priority allotment): "Given additional (e.g. a grant) resources, what do you think this should be used for." (The point is that reliability is always the highest priority, similarly decent performance). If users would ask us to spend grant money on better testing, then we would know they perceive an issue. Right now the answers feel to me like a complicated mix of "what is important", "what should be improved", and even "what can be improved".)

That is probably orthogonal to the "stability" question, I expect those just need to be two distinct questions.