OpenRefine / openrefine.org

Source website for openrefine.org
https://openrefine.org
Other
133 stars 119 forks source link

Document operations which don't respect facets #308

Closed tfmorris closed 5 months ago

tfmorris commented 7 months ago

Some operations don't respect facet settings. Some are obvious like row/column transform, but the list should be documented.

For the convenience of anyone working on the documentation, I'm copying over @tfmorris's "list of operations which are NOT affected by facets," from the support forum:

Column - Move, Remove, Rename, Reorder Denormalize (which isn't exposed anywhere in the UI that I can find) MultiValuedCell - Split, Join RowReorder (ie Sort -> Reorder rows permanently) Transpose - Columns to Rows, Rows to Column, KeyValueColumnize

Originally posted by @benmiller314 in https://github.com/OpenRefine/OpenRefine/issues/5978#issuecomment-1642264430

To Reproduce

Steps to reproduce the behavior:

  1. First, do ...
  2. Then, ...
  3. Finally, ...

Current Results

There is no documentation of which operations don't respect facet settings.

Expected Behavior

The list is documented.

thadguidry commented 7 months ago

The sort/reorder has always been a weird one for me, in that it does not respect facets, and many times I've needed exactly that. (only sort/reorder THESE record rows) Oh well. Eventually we can do record reordering with facets and that will take care of the need directly.

ostephens commented 7 months ago

@thadguidry Can you say more about sorting respecting facets? In particular:

thadguidry commented 7 months ago

@ostephens permanently re-order. It happens when I create logical records where the id is created (or recreated correctly) for a set of sub-records. But in order to create the right ids that I need, I first need the records (actually nested sub records of items in an inventory location ) sorted by an audit date column. After making the correct record ids, I need the records sorted permanently by date in order to load into other tools that expect a date sorted list of inventory records to match transactional history of the inventory item records and their new ids. If I remove all the facet/filters, and since the sort was permanent previously, I expect those newly updated record row indexes from the sort, to remain intact.

Makes somewhat sense?

Of course, flattening is the best approach typically for something like this... but you know what happens when we try to flatten nested records...we get our lovely set of a million OpenRefine columns and our browsers trying to repaint them :-)

ostephens commented 7 months ago

I'm afraid I don't understand the example @thadguidry.

I can't quite understand how you can permanently sort only a subset of rows. For example if I have a data set of 5 rows, and a filter that selects rows 1, 3 and 5, and I permanently sort them by some column what happens to the rows that were 2 and 4?

tfmorris commented 7 months ago

@thadguidry perhaps you can open a feature request for the feature you want and move this discussion there.

WR-Smiley commented 7 months ago

I'm new to the project (and OS contribution generally); I'd be happy to get started by updating the relevant documentation with this info, since I don't see that someone has done so yet.