html_url,issue_url,id,node_id,user,created_at,updated_at,author_association,body,reactions,performed_via_github_app,issue
https://github.com/pydata/xarray/pull/4972#issuecomment-828924833,https://api.github.com/repos/pydata/xarray/issues/4972,828924833,MDEyOklzc3VlQ29tbWVudDgyODkyNDgzMw==,12229877,2021-04-29T04:03:39Z,2021-04-29T04:03:39Z,CONTRIBUTOR,"> is there a way to separate `sizes` into `dims` and `shape` without using `draw`? I tried `dims, shape = sizes.map(lambda s: tuple(zip(*s)))`, but since the map strategy is not iterable the final unpacking fails.

You've just got to use `draw` for this.

> I'd prefer the verbose repr of `st.builds()` when debugging strategies, but if I know the strategy is correct and understand what it generates I'd probably prefer the smaller repr of `@st.composite`.

That's a good way of thinking about it :slightly_smiling_face: ","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,818059250
https://github.com/pydata/xarray/pull/4972#issuecomment-808736180,https://api.github.com/repos/pydata/xarray/issues/4972,808736180,MDEyOklzc3VlQ29tbWVudDgwODczNjE4MA==,12229877,2021-03-27T13:51:54Z,2021-03-27T13:51:54Z,CONTRIBUTOR,"Looking at https://github.com/keewis/xarray/compare/duckarray-tests...duckarray-tests-hypothesis, for high-level feedback:

- Overall it looks pretty good; though ping me again if/when it's a PR and I'll do line-level feedback on idiom issues
- A more general test would generate [the shapes](https://hypothesis.readthedocs.io/en/latest/numpy.html#hypothesis.extra.numpy.array_shapes), and the [axes to reduce over](https://hypothesis.readthedocs.io/en/latest/numpy.html#hypothesis.extra.numpy.valid_tuple_axes) - reducing a 1D array over the first dimension is going to miss things
- You use `@st.composite` when [the `.map()` method](https://hypothesis.readthedocs.io/en/latest/data.html#adapting-strategies) and a lambda would suffice (though the perf gain is small enough that this is mostly a readability issue)
- I don't see the point of `Label`, and we advise against mixing ""a strategy or a value"".  We break this rule a few times for backwards-compatibility in our Numpy support, but wouldn't write such an API these days.

And I'm always delighted to see people using Hypothesis to test libraries that I use and love 🥰🤩","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,818059250