Bugfixes/fix provider gpus UI

[Redm4x]

Display gpu models from feature discovery instead of capabilities attributes

• Return gpu models from feature discovery in providers endpoints
• Use those model instead of capabilities in provider list/detail pages
• Added missing indexes to improve performance

CREATE INDEX IF NOT EXISTS provider_snapshot_node_snapshot_id
    ON public."providerSnapshotNode" USING btree
    ("snapshotId" ASC NULLS LAST)
    WITH (deduplicate_items=True)
    TABLESPACE pg_default;
	
CREATE INDEX IF NOT EXISTS provider_snapshot_node_cpu_snapshot_node_id
    ON public."providerSnapshotNodeCPU" USING btree
    ("snapshotNodeId" ASC NULLS LAST)
    WITH (deduplicate_items=True)
    TABLESPACE pg_default;
	
CREATE INDEX IF NOT EXISTS provider_snapshot_node_gpu_snapshot_node_id
    ON public."providerSnapshotNodeGPU" USING btree
    ("snapshotNodeId" ASC NULLS LAST)
    WITH (deduplicate_items=True)
    TABLESPACE pg_default;

[ygrishajev]

imho this kind of comments don't help much. It's already obvious that providers is a providers list. If it's essential to emphasise it's got attributes and auditors (is it?) it can be named accordingly. e.g. providersWithAttrsAndAuditors. It would also help understanding code better further if it's important.

[ygrishajev]

this one is hard to understand, ideally I'd imagine this should be extracted into a named function with some vars that would better explain the conditions.

[ygrishajev]

Could you pls explain what this means?
I understand that findIndex matching same vendor, model, etc. Why should it also be of the same index?

[ygrishajev]

ok, I see, it's to remove duplicates

[ygrishajev]

It's really hard to understand this. I'd suggest if we could add a dedicated function for this.

type Matcher<T> = (a: T, b: T) => boolean;

function createFilterUnique<T>(matcher: Matcher<T> = (a, b) => a === b): (value: T, index: number, array: T[]) => boolean {
  return (value, index, array) => {
    return array.findIndex((other) => matcher(value, other)) === index;
  };
}

in this case the code above would be:

const distinctGpuModels = gpuModels.filter(
    createFilterUnique(x, o) => x.vendor === o.vendor && x.model === o.model && x.ram === o.ram && x.interface === o.interface))
);

and example below would be

.filter(createFilterUnique())

[ygrishajev]

can be tested here

[Redm4x]

It's to keep only the first occurence of each gpus since there are duplicates in the array. If the current index i matches findIndex that means it's the first occurence so it's kept, otherwise it is filtered out as duplicate.

[ygrishajev]

yea, I figured after asking the question :) check out a suggestion pls

[Redm4x]

mb your other replies weren't showing when I answered. For sure that would be easier to understand, I'll add it! 👍

[Redm4x]

I added the createFilterUnique helper with relevant unit tests

[ygrishajev]

This could also utilise that new filter function. What do you think?

Suggested change

	const gpuModels = provider.gpuModels.map(x => x.model).filter((x, i, arr) => arr.indexOf(x) === i);
	const gpuModels = provider.gpuModels.map(x => x.model).filter(createFilterUnique());

[Redm4x]

Yeah, I added it to the deploy-web project too. No unit tests for now, but we could add them once the big refactor is done (rebrand).

[ygrishajev]

Suggested change

	.filter((x, i, arr) => arr.indexOf(x) === i)
	.filter(createFilterUnique())