Moose group caches are long to build

[jecisc]

Moose group caches are taking a lot of time to build.

For example 1 just imported a Java model and I have the impression that 2/3 of the time spend in the import is during the building of the moose groups caches.

Maybe we could review how those are done? Maybe some could be optimized? Maybe some are unused and could be removed? Maybe some are not accessed often and could be computed lazily?

[badetitou]

I can tell for BL that all my scripts to import model uses

myModel installWithCache: false

Because it is too slow. I suppose it can be more useful if you build the image with the model online

[NicolasAnquetil]

I would remove them.
Many are created and never used ...
Can we find a way to build them lazily ?

[Gabriel-Darbord]

It also seems to me that a lot of cache content already exists in the byType storage...

[NicolasAnquetil]

byType uses the cache, no?

[Gabriel-Darbord]

In models, the cache is filled using the byType storage.
The difference is that the cache will contain a MooseGroup and byType contains an OrderedCollection.

allWithType: aFamixClass ofGroupClass: aGroupClass

	aFamixClass isTrait ifTrue: [ MooseAllWithTypeOnTrait signal ].
	^ self cacheAt: aFamixClass ifAbsentPut: [
		  aGroupClass
			  withAll: (self entityStorage selectAllWithType: aFamixClass) "<- byType"
			  withDescription:
			  'All ' , aFamixClass name asLowercase asEnglishPlural ]

[NicolasAnquetil]

So that settles it to me, no?
remove the cache and keep byType
eventually computes the caches lazily