Introduce PureModelContextCollection type and use it in WorkspaceSDLCLoader

[abhishoya-gs]

What type of PR is this?

Improvement

What does this PR do / why is it needed ?

As of today, WorkspaceSDLCLoader calls ModelLoader for each dependency. This introduces as many numbers of HTTP requests and doesn't account for dependency override.

Solution

• Introduce PureModelContextCollection type
• Modify WorkspaceSDLCLoader to call ModelLoader with a PureModelContextCollection
• Support fetching PureModelContextCollection using AlloySDLCLoader

Additionally, PureModelContextCollection can be used further to send smaller payloads from clients like Studio.

Which issue(s) this PR fixes:

Fixes #

Other notes for reviewers:

For testing, I rely on change in test setup and manual testing.

Does this PR introduce a user-facing change?

No

[github-actions]

Test Results

     761 files  ±0       761 suites  ±0   1h 6m 29s ⏱️ + 1m 25s
12 552 tests ±0  12 319 ✔️ ±0  233 💤 ±0  0 ❌ ±0 
15 663 runs  ±0  15 412 ✔️ ±0  251 💤 ±0  0 ❌ ±0 

Results for commit f68b9ae. ± Comparison against base commit 6a0065a.

♻️ This comment has been updated with latest results.

[MauricioUyaguari]

@abhishoya-gs
can we not change the model loader and add another type of PureModelContext.
I think @rafaelbey started something. see : #2446

[abhishoya-gs]

@abhishoya-gs can we not change the model loader and add another type of PureModelContext. I think @rafaelbey started something. see : #2446

Okay sure, seems like a better approach

[rafaelbey]

This constrains defeats one of the main purposes of the collection I wanted to introduce - improve studio flow. For example - studio can manage a collection, where an entry is a PMCD of the current project, and the other entries are pointer to the project dependencies.

[abhishoya-gs]

I will remove this constraint so it will help improve studio flow. I will also change the constraint added on shouldCache but since we cache compiled model only for PureModelContextPointer, we won't gain much there in case of studio.

Also, I have placed another check in SDLCFetcher see here. It's required as my implementation only uses alloy loaders and thus only AlloySDLC. If there's a use case to support sourcing from multiple loaders the logic would have to be changed accordingly.