[Lens] Clear out all attribute properties before updating

[flash1293]

Fixes #72855

Lens uses an unindexed object type to store its attributes. This has some advantages, but becomes a problem when updating, because when doing an update, Elasticsearch is merging objects instead of replacing them (https://discuss.elastic.co/t/updating-an-object-field/110735). This means if there are objects used as maps with unknown keys, they will stay in the saved object even though they were deleted from the update document.

E.g. if the following document is stored:

{
  "attributes": {
    "myMap": {
      "a": 1,
      "b": 2,
    }
  }
}

and

savedObjectsClient.update('mydoc', '1234', {
  myMap: {
    b: 5
  }
})

then the resulting document will look like this:

{
  "attributes": {
    "myMap": {
      "a": 1,
      "b": 5,
    }
  }
}

This PR fixes the problem for Lens by doing a two-phased update setting all keys to null before updating them with the actual new attributes using the bulk update API. The order of these updates is guaranteed (https://discuss.elastic.co/t/order-of--bulk-request-operations/98124).

This PR also removes some local types which were in place because the core service wasn't typed when the code was written.

[wylieconlon]

Instead of using bulkUpdate, is it more correct to use index? I'm basing this off the docs which say that "update" will perform a merge, but "index" rewrites the whole thing. https://www.elastic.co/guide/en/elasticsearch/reference/current/docs-update.html

It also appears that index is a method supported in the repository_es_client.ts but not necessarily exported by the types

[flash1293]

That makes a lot of sense, thanks. I would like to get a fix out without waiting for changed core apis, do you think this approach is fine as a stopgap measure while working with the platform team to get a better solution?

[wylieconlon]

Tested in safari, and this change definitely fixes the behavior. This is a good fix.

[flash1293]

@elasticmachine merge upstream

[flash1293]

@elasticmachine merge upstream

[flash1293]

@elasticmachine merge upstream

[kibanamachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request
• Commit: 0e6ed88

Build metrics

page load bundle size

id	value	diff	baseline
lens	843.9KB	+717.0B	843.2KB

History

• 💔 Build #67167 failed deb02cb
• 💚 Build #66876 succeeded 8e182aa
• 💚 Build #66839 succeeded 94293c2
• 💔 Build #66833 failed 3ad79cd

To update your PR or re-run it, just comment with:
@elasticmachine merge upstream