issue_comments: 293749896

This data as json

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/issues/1372#issuecomment-293749896 https://api.github.com/repos/pydata/xarray/issues/1372 293749896 MDEyOklzc3VlQ29tbWVudDI5Mzc0OTg5Ng== 306380 2017-04-13T01:15:19Z 2017-04-13T01:15:19Z MEMBER

Optimization has come up a bit recently. In other projects like dask-glm I've actually been thinking about just turning it off entirely. The spread of desires here is wide. I think it would be good to make it a bit more customizable so that different applications can more easily customize their optimization suite (see https://github.com/dask/dask/issues/2206)

But for this application in particular presumably all of these already tasks fuse together into a single composite task, yes? Then we would need to commute certain operations in some order (preferring to push down slicing operations). Then we would need to merge all of the slices, yes? I think it would be interesting to see how fast one could make such an optimization. If the answer is "decently fast" then we might also want to look into doing inplace operations. If the answer is "only sorta-fast" then I would hesitate before adding this as a default to dask.array but would again stress that we should work to make different projects customize optimization to suit their own needs.

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   221387277