[ML] DF Analytics Regression exploration: replace table with data grid #63650
Conversation
alvarezmelissa87
added
release_note:enhancement
:ml
v8.0.0
Feature:Data Frame Analytics
alvarezmelissa87
requested review from
walterra,
peteharverson and
jgowdyelastic
|
Pinging @elastic/ml-ui (:ml) |
|
Any idea why
@peteharverson - added 'boolean' schema for 'boolean' type fields in e825faa |
|
Not needed here (as I think it used to show an error before too), but is there a way we can not show an error in this case, when filtering for training/testing data, and the doc count is zero as expected:
@peteharverson - good question. This particular error can be returned for reasons other than no documents being returned from the given query to run the evaluate analysis on so I defaulted to always showing it. One thing we could do is only show the error if the docCount is greater than 0. Happy to add it in a follow up as I'll need to do it for both regression/classification views. |
|
When the query bar query changes, we need to check the selected page is still valid. Here I was on the last page (125), edit the filter, and it looks like the grid is trying to reload page 125 even though there are only 24 pages for the new query value:
|
|
Can we prevent the user trying to sort on the object type
@peteharverson - good catch! Fixed in e825faa |
| featureImportanceFields.push({ | ||
| id: `${resultsField}.feature_importance`, | ||
| name: `${resultsField}.feature_importance`, | ||
| type: KBN_FIELD_TYPES.NUMBER, | ||
| }); |
There was a problem hiding this comment.
I'm not sure I understand this change, can you explain why having the individual fields is no longer necessary?
There was a problem hiding this comment.
The feature_importance field is an array of objects so the <results_field>.feature_importance.<id> field name wasn't actually mapping to anything - that column was just empty.
Setting it as just feature_importance lets the whole array show up in the popup.
|
Regarding this comment by @peteharverson:
In the parallel PR I'm working on to consolidate data grid code, I fixed it the following way: whenever the query changes, the paging will reset to the first page. https://github.com/elastic/kibana/pull/63447/files#diff-3a8487254a622f633414d662c4e7a84dR45 |
There was a problem hiding this comment.
Tested latest edits and LGTM.
I think it would be nice to hide that 'error' shown in #63650 (comment) when filtering for testing / training, as long as we don't hide genuine errors which are causing the doc count to be zero.
Happy for the paging issue when the query changes the total number of pages to be done in a follow-up.
💚 Build SucceededHistory
To update your PR or re-run it, just comment with: |
elastic#63650) * add feature_importance column correctly * wip: switch regression table to datagrid * add search bar to regression view * ensure feature importance fields show up correctly * wip: filter by training/testing * remove separate testing/training filter * make error more clear * handle lucene queries * remove unnecessary comment * ensure boolean shows up correctly.no sorting by feature importance * remove unused translations
#63650) (#63739) * add feature_importance column correctly * wip: switch regression table to datagrid * add search bar to regression view * ensure feature importance fields show up correctly * wip: filter by training/testing * remove separate testing/training filter * make error more clear * handle lucene queries * remove unnecessary comment * ensure boolean shows up correctly.no sorting by feature importance * remove unused translations
Summary
Related meta issue: #51288
EuiInMemoryTablewithEuiDataGridcomponentQueryStringInputcomponentNOTE
Paging issue @peteharverson mentioned will be fixed in a follow up once @walterra 's fix has been merged in order to use the same solution.
Checklist
Delete any items that are not applicable to this PR.