Tree search is not scoped properly

[grantfitzsimmons]

Describe the bug
When searching in a tree, the results are not scoped.

To Reproduce
Steps to reproduce the behavior:

1. Go to the hujinnhc_2024_07_01 database on the Test Panel
2. Log into any collection
3. Go to the Geography tree
4. Search 'Israel'
5. See that there are a number of results while only one record exists in this tree

Screenshots

https://hujinnhc20240701-edge.test.specifysystems.org/specify/tree/geography/?conformation=~~84501~84503---

Encountering this issue on edge, v7.9.6, v7.9.5, and v7.9.4. Seems to have broken in the v7.9.4 release.

Expected behavior
Only nodes from the current tree should appear!

Works correctly in v7.9.3.

System Information
Specify 7 System Information - 2024-07-01T16_54_20.967Z.txt

Reported By
The Hebrew University of Jerusalem

Lately (I think since the last update) every time we look for a place in the geography tree we get multiple instances of the same result:

Only one of the options is the "real" result (i.e. the one in the current tree). The other options belong to different disciplines but somehow also appear here.

Additional context
This has been encountered internally and an issue may already exist.

[melton-jason]

As of #3304, Specify tries to be smart about scoping and prevent Hiearchy Exceptions thrown by the backend (as seen in #3564 and #4989) when trying to fetch a collection of resources.

If the collection fetch is supposed to scoped (by specifying domainfilter: true in the fetchCollection options) but the frontend determines the table can not be scoped, then the domainfilter is removed to the final request sent to the backend.

specify7/specifyweb/frontend/js_src/lib/components/DataModel/collection.ts

Lines 105 to 111 in 080b32c

	else if (key === 'domainFilter') {
	// GetScope() returns undefined for tables with only collectionmemberid.
	const scopingField = tables[tableName].getScope();
	return value === true &&
	(tableName === 'Attachment' || typeof scopingField === 'object')
	? 'true'
	: undefined;

This results in behavior like in this issue (and an identical issue: #4542 (review)), where we are indicating the request should be scoped, but the frontend is determining the tree table to be 'unscopable', so it omits the domainfilter.

specify7/specifyweb/frontend/js_src/lib/components/TreeView/Search.tsx

Lines 71 to 96 in 080b32c

	fetchCollection(
	// eslint-disable-next-line @typescript-eslint/no-unnecessary-type-assertion
	tableName as AnyTree['tableName'],
	{
	limit: DEFAULT_FETCH_LIMIT,
	orderBy: 'name',
	domainFilter: true,
	},
	{
	[resolvedSearchField]: searchCaseSensitive
	? value
	: value.toLowerCase(),
	}
	).then(({ records }) =>
	records.map((node) => {
	const rankDefinition = treeDefinitionItems.find(
	({ rankId }) => rankId === node.rankId
	);
	const rankName = rankDefinition?.title || rankDefinition?.name;
	return {
	label: node.fullName ?? node.name,
	subLabel: rankName,
	data: node as SerializedResource<SCHEMA>,
	};
	})
	)

This assumption that trees are 'unscopable' is fairly reasonable: their scope can not be easily determined dynamically, but domainfilter is allowed in the request because the scope is hardcoded and determinable on the backend:

specify7/specifyweb/specify/filter_by_col.py

Lines 27 to 34 in 080b32c

	if queryset.model in (Geography, Geologictimeperiod, Lithostrat):
	return queryset.filter(definition__disciplines=collection.discipline)
	if queryset.model is Taxon:
	return queryset.filter(definition__discipline=collection.discipline)
	if queryset.model is Storage:
	return queryset.filter(definition__institutions=collection.discipline.division.institution.id)

---

The preferable solution here would be to handle scoping on the backend and always pass a domainfilter=true if it is specified by the caller of the endpoint on the frontend.

As mentioned in: #5044 (review)

[realVinayak]

I somehow think that inverting is better. If we know we can't scope, frontend shouldn't scope If we can't reliably know whether we can scope or not, then frontend shouldn't decide.The ideal implementation in that PR would have done this: check if we absolutely can't scope (or can scope), then only pass in relevant flag. Otherwise, don't act smart. Yes, that would have involved some hard coding, but worth it. Should just move to backend though.

[emenslin]

Can recreate in edge (7.9.6)

[grantfitzsimmons]

This manifests in worse behavior than just search–

Pycreus is a named taxon in both the 'Vascular Plants' and 'Lichens' collections in this database.

There is only one in each tree:

Yet in the WorkBench I need to disambiguate it between the two:

From the disambiguation dialog, I see the following:

• Taxon ID 8078 (current discipline, VP)
• Taxon ID 15920 (other discipline, Lichens)

I can select the node from the other discipline

[CarolineDenis]

@grantfitzsimmons, I cannot reproduce this on Edge