This PR changes the 4 constraints on rank and shape for deferred constants to merge the two rank constraints into one, and merge the two shape constraints into one. I did that because when I first read the 4 constraints, I was very confused how they were different, because I was still a bit confused about explicit and ultimate specifications. I initially thought the pairs of constraints were duplicates and one of each could be removed. I think by combining them, it becomes more clear that there is no duplicate parts. It's also less text!
(Also, I made this other variation of the constraints which I think is effectively the same but structured differently. Just another option to consider if you don't like what's in the main PR edit.
Constraint: If any ultimate specification of a deferred constant has
explicit shape S, then an explicit specification of that
deferred constant shall have explicit shape S, and all other
ultimate specifications of that deferred constant shall
have deferred rank, implied shape with the same rank as
S, or explicit shape S.
Constraint: If any ultimate specification of a deferred constant has
implied shape with rank R, then an explicit specification
of that deferred constant shall have explicit or implied
shape of rank R, and all other ultimate specifications
of that deferred constant shall have deferred rank or
implied shape of rank R.
)
However, while I made that change, I realized the normative text was insufficient to handle the case where all ultimate specifications are deferred rank. For reference, here was the previous normative text:
If any ultimate specification of a deferred constant has a non-deferred rank, R, then that deferred constant has rank R. Otherwise it has deferred rank.
If any ultimate specification of a deferred constant has an explicit shape S, then that deferred constant has shape S. Otherwise the it has implied shape.
If all ultimate specifications are deferred rank, the 1st paragraph says the deferred constant is deferred rank, but the 2nd paragraph says it has implied shape. Both of those cannot be true at once. To fix that, I combined all three cases (explicit shape, implied shape, deferred rank) into one paragraph and sorted them by precedence:
If any ultimate specification of a deferred constant has an explicit shape S, then that deferred constant has shape S. Otherwise, if any ultimate specification of that deferred constant has implied shape, then it has implied shape with the same rank. Otherwise it has deferred rank.
It's a similar structure to the variations of the constraints I wrote above in the PR description, which is why I thought they might be nice, to match the structure of this new normative text.
This PR changes the 4 constraints on rank and shape for deferred constants to merge the two rank constraints into one, and merge the two shape constraints into one. I did that because when I first read the 4 constraints, I was very confused how they were different, because I was still a bit confused about explicit and ultimate specifications. I initially thought the pairs of constraints were duplicates and one of each could be removed. I think by combining them, it becomes more clear that there is no duplicate parts. It's also less text!
(Also, I made this other variation of the constraints which I think is effectively the same but structured differently. Just another option to consider if you don't like what's in the main PR edit.
)
However, while I made that change, I realized the normative text was insufficient to handle the case where all ultimate specifications are deferred rank. For reference, here was the previous normative text:
If all ultimate specifications are deferred rank, the 1st paragraph says the deferred constant is deferred rank, but the 2nd paragraph says it has implied shape. Both of those cannot be true at once. To fix that, I combined all three cases (explicit shape, implied shape, deferred rank) into one paragraph and sorted them by precedence:
It's a similar structure to the variations of the constraints I wrote above in the PR description, which is why I thought they might be nice, to match the structure of this new normative text.