Incremental GA support starting from CSchankGA

[sarrasoussia]

Work based on #4009

[JacquesCarette]

Quite a lot of this looks reasonable.

[JacquesCarette]

I am not sure the usual make test builds website, which is where this is happening.

[balacij]

What commands are you running? If you can share logs, we can try to debug it.

[balacij]

@JacquesCarette A basic make tests the website as well as the examples.

[balacij]

Testing locally, I see errors, for example, with dblpend:

WARNING! Overwriting `cartesian` :: ConceptChunk
Symbol Table: dblpend: ERROR! Overwriting a chunk (`dblpend_multivector` :: `ConceptChunk`) with a chunk of a different type: `IdeaDict`
CallStack (from HasCallStack):
  error, called at lib/Database/Drasil/ChunkDB.hs:217:18 in drasil-database-0.1.1.0-AILWoHk0oaT2jusNLXYhMa:Database.Drasil.ChunkDB
make: *** [dblpend_gen] Error 1

This is by running make (with no arguments) in the code folder. Obviously, this is a new error which needs to be fixed.

@sarrasoussia With #4320, we changed the ChunkDB quite a bit. There is one hard rule now (immediately relevant to this error): chunks cannot share the same UID. In this case, you should be able to remove the IdeaDict in favour of keeping just the ConceptChunk version of dblpend_multivector.

[balacij]

Can you please post the results of make --version?

Do you want to try debugging this over a Discord call tomorrow?

[sarrasoussia]

Can you please post the results of make --version?

Do you want to try debugging this over a Discord call tomorrow?

Yes please, thank youu

[sarrasoussia]

Note:

Right now we had both ClifKind (Scalar, Vector, etc.) and also the idea of picking grades with [Int]. That’s redundant — the kind already tells you the grade, and it doesn’t cover mixed cases (like scalar + bivector). So better to just drop ClifKind and use a list of Ints for grades.

Grades are just numbers between 0 and n (where n is the dimension). Anything outside that doesn’t make sense, so we need to handle it in code. There are a few ways:

• ClifS Dimension [Int] Space This can allow ClifS (Fixed 3) [42] Real which is incorrect. So we need to implement a smart constructor that checks the list is in 0..n.
  mkClifS :: Dimension -> [Int] -> Space -> Maybe Space
mkClifS (Fixed n) grades base =
if all (\g -> g >= 0 && g <= n) grades
then Just (ClifS (Fixed n) grades base)
else Nothing

• Wrap Int in a Grade newtype and provide a validator. Makes it clear when something is supposed to be a grade, not just any integer.
  newtype Grade = Grade Int
deriving (Eq, Show)
mkGrade :: Dimension -> Int -> Maybe Grade
mkGrade (Fixed n) g =
if g >= 0 && g <= n then Just (Grade g) else Nothing
  so now it is ClifS Dimension [Grade] Space

[JacquesCarette]

Your 'Note' is great. I would actually implement a combination of both, i.e.

data ClifS = ClifS Dimension Grades Space

and a smart constructor for Grades the way that you outline (but it would hide a [Int] not just Int) that the smart constructor for ClifS would call.

Rather than either of those returning Maybe, I would use error where you currently use Nothing.