[TC] CS projections replaced with uvar #648
Conversation
|
Note that the compilation with uvar for projections in this PR only addresses instances compilation |
|
I'm a bit lost, the test passes but I don't see how the new term constructor is compiled... miracle? 😀 |
|
The precompilator transforms a term like Here I decompile tc.canonical-projection (during the elpi's rule creation): I replace the sub-term with a fresh elpi-var Am I clarifying your remark ? |
FissoreD
force-pushed
the
tc-compile-more-constructors
branch
3 times, most recently
from
3eceb49
to
78c7545
Compare
FissoreD
force-pushed
the
tc-compile-more-constructors
branch
from
b1c455e
to
ad0b8dc
Compare
FissoreD
force-pushed
the
tc-compile-more-constructors
branch
from
dc047a6
to
f348d07
Compare
FissoreD
force-pushed
the
tc-compile-more-constructors
branch
from
a015aeb
to
7a07b58
Compare
FissoreD
force-pushed
the
tc-compile-more-constructors
branch
2 times, most recently
from
e140d15
to
25de1d1
Compare
…ity constants" This reverts commit 6c63978.
This aims to solve the compilation error produced by the compilation of
```
Module foo.
Class B (i : nat).
Section s.
(* Class with coercion depending on section parameters *)
Context (A : Type).
Class C (i : A) : Set := {
x (x : A) :: B 3
}.
End s.
End foo.
```
gares
force-pushed
the
tc-compile-more-constructors
branch
from
a9d89b5
to
7986110
Compare
|
@CohenCyril this job https://github.com/LPCIC/coq-elpi/actions/runs/11143050022/job/30967371306?pr=648 |
If a subterm is a (primitive) projection, it is replaced with fresh elpi variable
This variable and the original subterm are related by the coq.unify-eq predicate, to ensure
that each subterm of a class call correctly unifies with the corresponding subterm in the instance.
Here the test for this PR