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/4740#issuecomment-832044344,https://api.github.com/repos/pydata/xarray/issues/4740,832044344,MDEyOklzc3VlQ29tbWVudDgzMjA0NDM0NA==,2448579,2021-05-04T15:45:50Z,2021-05-04T15:46:07Z,MEMBER,OK should be good to merge. We can set the `meta` in a future version (It did not work on 2.15.0 which is the dask version we can bump to right now),"{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664 https://github.com/pydata/xarray/pull/4740#issuecomment-831944002,https://api.github.com/repos/pydata/xarray/issues/4740,831944002,MDEyOklzc3VlQ29tbWVudDgzMTk0NDAwMg==,2448579,2021-05-04T13:31:52Z,2021-05-04T13:31:52Z,MEMBER,Test failure is real and looks like some bug with `dtype` vs `meta` in our min dask version. I've bumped dask to see if it fixes it.,"{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664 https://github.com/pydata/xarray/pull/4740#issuecomment-778853034,https://api.github.com/repos/pydata/xarray/issues/4740,778853034,MDEyOklzc3VlQ29tbWVudDc3ODg1MzAzNA==,14371165,2021-02-14T22:33:39Z,2021-02-14T22:33:39Z,MEMBER,@max-sixty It's the profiler [Spyder](https://github.com/spyder-ide/spyder) has. It's been quite useful for hunting down various bottlenecks but it has its quirks as well. ,"{""total_count"": 1, ""+1"": 1, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664 https://github.com/pydata/xarray/pull/4740#issuecomment-778841404,https://api.github.com/repos/pydata/xarray/issues/4740,778841404,MDEyOklzc3VlQ29tbWVudDc3ODg0MTQwNA==,5635139,2021-02-14T21:03:02Z,2021-02-14T21:03:02Z,MEMBER,"Thanks for pinging @Illviljan . Looks good from my side too. @Illviljan what's that profiling tool you're using? The hierarchy of functions looks useful!","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664 https://github.com/pydata/xarray/pull/4740#issuecomment-767770962,https://api.github.com/repos/pydata/xarray/issues/4740,767770962,MDEyOklzc3VlQ29tbWVudDc2Nzc3MDk2Mg==,14371165,2021-01-26T19:22:33Z,2021-01-26T19:22:33Z,MEMBER,"Well, I think this is good enough. It's quite an improvement already and I think the error will disappear if the CI is rerun. ","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664 https://github.com/pydata/xarray/pull/4740#issuecomment-752690563,https://api.github.com/repos/pydata/xarray/issues/4740,752690563,MDEyOklzc3VlQ29tbWVudDc1MjY5MDU2Mw==,2448579,2020-12-30T17:00:25Z,2020-12-30T17:00:25Z,MEMBER,"> Another more difficult idea is to improve the dask arrays shape-calculation. Perhaps saving the value and only do the cached_cumsum call when the chunks actually has changed? Thanks for opening an issue on the dask side. On the xarray side, we want to cache some of these properties, see #3514","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664 https://github.com/pydata/xarray/pull/4740#issuecomment-752286931,https://api.github.com/repos/pydata/xarray/issues/4740,752286931,MDEyOklzc3VlQ29tbWVudDc1MjI4NjkzMQ==,14371165,2020-12-30T00:36:24Z,2020-12-30T00:36:24Z,MEMBER,"Looking good! One thing I notice now with these changes is that `ndim` is now creeping up as the 2nd worst time consumer: ![bild](https://user-images.githubusercontent.com/14371165/103321860-f3dffa00-4a3b-11eb-8b4e-941f1f9e027b.png) An easy improvement to this is adding `ndim = var.ndim` and replacing all the `var.ndim` around `interp_func`. That reduces the number of calls from 68000 to 58000. Another more difficult idea is to improve the dask arrays shape-calculation. Perhaps saving the value and only do the cached_cumsum call when the chunks actually has changed?","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,776042664