Gabriel Scherer
Gabriel Scherer
I like the idea. I think that `Sys` would be a better place than `Gc`; it makes sense given that some implementations of OCaml give a different semantics to physical...
For a callsite I think that `physically_equal v1 v2` reads better than `physical_equality v1 v2`, or maybe `phys_equal`.
I also prefer to have an explicit name, `phys_equal` or `Sys.phys_equal`, over an operator. I think that physical equality is a subtle operation, used by people who care about the...
The use of eagerly-evaluated let-bindings in class definitions is a very intentional design choice that was certainly motivated by practical applications. Changing the semantics at this point sounds fairly dangerous,...
It may be that grepping for `class` and a manual review would let you isolate candidates for interesting eager-let behavior without building or running any code. (Classes with parameters are...
The following search on Github already gives some interesting results: `/^class [a-zA-Z_]* =$/ /^ *let/ lang:OCaml` [Search results](https://github.com/search?q=%2F^class+[a-zA-Z_]*+%3D%24%2F+%2F^+*let%2F+lang%3AOCaml+&type=code).
cc @garrigue who may have an opinion
In my opinion this is now part of the language specification, so we would need strong reasons to change it. I consider the current `let open` behavior to be a...
Maybe `the expected type is "Parsetree.constant"` could be `a record type is expected, "Parsetree.constant"`?
Right now the answer is "neither", #12131 has not been merged at all. I will try to merge today so that it appears in trunk (but maybe not the opam...