GenConstr can intern directly to constr - #22106
Conversation
Using rocq-prover/rocq#22106 Implicit argument handling is not ideal and only works for maximal implicits. To fix that I think we would need to parse applied `lib:qid args` instead of parsing just the `lib:qid`, or change the upstream Rocq code somehow. TODO optcomp
Using rocq-prover/rocq#22106 Implicit argument handling may be bugged TODO optcomp
Using rocq-prover/rocq#22106 Implicit argument handling may be bugged TODO optcomp
Using rocq-prover/rocq#22106 TODO optcomp
|
|
||
| type ('raw, 'glb) constr_intern_fun = ?loc:Loc.t -> glob_sign -> 'raw -> 'glb | ||
|
|
||
| type constr_intern_info = { passthrough_impls : bool } |
There was a problem hiding this comment.
instead of telling the interner to look at the produced term we could return the data produced by find appl_head_data (as used in
Line 2317 in 9fa45ea
That would allow genargs to have implicits even when interning to GGenarg
not sure if worth doing
|
This version seems to work. We need the ability to control whether implicits of the produced term are used to support both |
|
@SkySkimmer please ping me when you have time to write the optcomp layer for elpi. |
|
In principle this PR can be merged before the elpi side is ready (it modifies the |
It is not needed and was just there to prototype the API.
|
I pushed a tiny commit that removed the elpi overlay to double-check everything is fine. If the CI goes through I'll merge right after. |
|
@coqbot merge now |
For
lib:qidsyntax LPCIC/coq-elpi#1035