issue_comments

Table actions

5 rows where issue = 662982199 sorted by updated_at descending

✎ View and edit SQL

This data as json, CSV (advanced)

Suggested facets: reactions, created_at (date), updated_at (date)

author_association 2

issue 1

  • Parallel tasks on subsets of a dask array wrapped in an xarray Dataset · 5 ✖
id html_url issue_url node_id user created_at updated_at ▲ author_association body reactions performed_via_github_app issue
662563406 https://github.com/pydata/xarray/issues/4241#issuecomment-662563406 https://api.github.com/repos/pydata/xarray/issues/4241 MDEyOklzc3VlQ29tbWVudDY2MjU2MzQwNg== maximemorariu 41797673 2020-07-22T16:45:42Z 2020-07-22T16:45:42Z NONE

This is a fundamental problem that is rather hard to solve without creating a copy of the data.

We just released the rechunker package, which makes it easy to create a copy of your data with a different chunking scheme (e.g contiguous in time, chunked in space). If you have enough disk space to store a copy, this might be a good solution.

Thanks for confirming and pointing me to rechunker, that looks nice.

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   Parallel tasks on subsets of a dask array wrapped in an xarray Dataset 662982199
662517426 https://github.com/pydata/xarray/issues/4241#issuecomment-662517426 https://api.github.com/repos/pydata/xarray/issues/4241 MDEyOklzc3VlQ29tbWVudDY2MjUxNzQyNg== rabernat 1197350 2020-07-22T15:22:51Z 2020-07-22T15:22:51Z MEMBER

The reason is that my function here must be applied along the time dimension (e.g., a rolling median in time), but my data is chunked across the time dimension

This is a fundamental problem that is rather hard to solve without creating a copy of the data.

We just released the rechunker package, which makes it easy to create a copy of your data with a different chunking scheme (e.g contiguous in time, chunked in space). If you have enough disk space to store a copy, this might be a good solution.

 
{
    "total_count": 2,
    "+1": 2,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   Parallel tasks on subsets of a dask array wrapped in an xarray Dataset 662982199
662512964 https://github.com/pydata/xarray/issues/4241#issuecomment-662512964 https://api.github.com/repos/pydata/xarray/issues/4241 MDEyOklzc3VlQ29tbWVudDY2MjUxMjk2NA== dcherian 2448579 2020-07-22T15:14:53Z 2020-07-22T15:14:53Z MEMBER

You could try dask's map_overlap to share "halo" or Ghost points between chunks. Also see https://image.dask.org/en/latest/dask_image.ndfilters.html#dask_image.ndfilters.median_filter

 
{
    "total_count": 1,
    "+1": 1,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   Parallel tasks on subsets of a dask array wrapped in an xarray Dataset 662982199
662509778 https://github.com/pydata/xarray/issues/4241#issuecomment-662509778 https://api.github.com/repos/pydata/xarray/issues/4241 MDEyOklzc3VlQ29tbWVudDY2MjUwOTc3OA== maximemorariu 41797673 2020-07-22T15:09:24Z 2020-07-22T15:09:24Z NONE

Thanks for your answer. Yes I looked at apply_ufunc and map_blocks and cannot use these here. The reason is that my function here must be applied along the time dimension (e.g., a rolling median in time), but my data is chunked across the time dimension. I could of course re-chunk the data (create a copy where there are no chunks along the time dimension), but I would like to know if this can be avoided.

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   Parallel tasks on subsets of a dask array wrapped in an xarray Dataset 662982199
661847133 https://github.com/pydata/xarray/issues/4241#issuecomment-661847133 https://api.github.com/repos/pydata/xarray/issues/4241 MDEyOklzc3VlQ29tbWVudDY2MTg0NzEzMw== keewis 14808389 2020-07-21T13:02:20Z 2020-07-21T13:03:52Z MEMBER

cannot be done by just using numpy-like functions

did you look at apply_ufunc (examples) and map_blocks? Functions applied with apply_ufunc will receive whatever was wrapped by dask while map_blocks allows you to work with xarray objects. See also the docs.

 
{
    "total_count": 0,
    "+1": 0,
    "-1": 0,
    "laugh": 0,
    "hooray": 0,
    "confused": 0,
    "heart": 0,
    "rocket": 0,
    "eyes": 0
}
   Parallel tasks on subsets of a dask array wrapped in an xarray Dataset 662982199

Advanced export

JSON shape: default, array, newline-delimited, object

CSV options: 

CREATE TABLE [issue_comments] (
   [html_url] TEXT,
   [issue_url] TEXT,
   [id] INTEGER PRIMARY KEY,
   [node_id] TEXT,
   [user] INTEGER REFERENCES [users]([id]),
   [created_at] TEXT,
   [updated_at] TEXT,
   [author_association] TEXT,
   [body] TEXT,
   [reactions] TEXT,
   [performed_via_github_app] TEXT,
   [issue] INTEGER REFERENCES [issues]([id])
);
CREATE INDEX [idx_issue_comments_issue]
    ON [issue_comments] ([issue]);
CREATE INDEX [idx_issue_comments_user]
    ON [issue_comments] ([user]);