Mlsem_types.TVOpval all_vars : kind -> Mlsem_types__.Tvar.MVarSet.tEvery type and row variable of that kind created since the program started (cf. Var.all_vars). all_vars KNoInfer is the set of rigid variables passed as monomorphic to tally.
val vars : Ty.t -> Mlsem_types__.Tvar.MVarSet.tAll the variables of a type. This is a syntactic notion inherited from sstt: a variable that does not affect the denotation of the type may or may not be reported, depending on the simplifications already applied to it.
val vars' : Ty.t list -> Mlsem_types__.Tvar.MVarSet.tval top_vars : Ty.t -> Mlsem_types__.Tvar.MVarSet.tOnly the variables occurring at top level, i.e. not under a type constructor.
val strict_vars : Ty.t -> Mlsem_types__.Tvar.MVarSet.tOnly the variables occurring strictly under a type constructor.
`Pos (resp. `Neg) means the type is monotonic (resp. anti-monotonic) in the variable, `Both that it is neither, and `None that the variable does not affect the denotation at all — which can happen because vars is syntactic, so a reported variable may be vacuous.
The primed variants below combine the polarities over a list of types, as if the variable occurred in each of them. Variables of polarity `None are omitted from the results.
val is_ground_typ : Ty.t -> boolval refresh :
?preserve_names:bool ->
kind:kind ->
Mlsem_types__.Tvar.MVarSet.t ->
Mlsem_types__.Tvar.Subst.tA substitution renaming each given variable to a fresh one of the given kind. With ~preserve_names:true the fresh variables keep the display name of the originals, which makes them indistinguishable when printed with TVar.pp.
val shorten_names :
?kind:kind ->
Mlsem_types__.Tvar.MVarSet.t ->
Mlsem_types__.Tvar.Subst.tA substitution renaming the given variables to fresh ones named a, b, … — for display only.
val pp_typ_short : Stdlib.Format.formatter -> Ty.t -> unitPrints with variables renamed by shorten_names.
val pp_typ_uniq : Stdlib.Format.formatter -> Ty.t -> unitPrints with variables renamed so that distinct variables print differently, even when they share a display name.
val pp_typ_subst :
Mlsem_types__.Tvar.Subst.t ->
Stdlib.Format.formatter ->
Ty.t ->
unitPrints the type after applying the given substitution.
val clean :
pos1:Ty.t ->
neg1:Ty.t ->
pos2:Mlsem_types__.Tvar.Row.t ->
neg2:Mlsem_types__.Tvar.Row.t ->
Mlsem_types__.Tvar.MVarSet.t ->
Ty.t ->
Ty.tclean p n mono t substitutes in t all variables not in mono and only occurring positively by p, and all variables not in mono and only occurring negatively by n
The smallest instance of the type, obtained by cleaning the variables outside mono with empty positively and any negatively.
Dual of bot_instance: the largest instance.
Same as bot_instance and top_instance, but the polarities of the variables are computed over all the types at once, so that the same substitution is applied to each of them. In particular, this preserves subtyping between them (e.g. between the bounds of a gradual type).
val tally :
?record:bool ->
Mlsem_types__.Tvar.MVarSet.t ->
(Ty.t * Ty.t) list ->
Mlsem_types__.Tvar.Subst.t listtally mono cs returns substitutions — of the variables not in mono — that make every constraint (s,t) of cs satisfy s ≤ t. The empty list means the constraints are unsatisfiable. Set ~record:false to keep the instance out of Recording's log; this matters for calls made while comparing candidate solutions, which are not part of the derivation being built.
val tally_const_rows :
?record:bool ->
Mlsem_types__.Tvar.MVarSet.t ->
(Ty.t * Ty.t) list ->
Mlsem_types__.Tvar.Subst.t listLike tally, but row variables are not instantiated with rows that introduce new fields.
val decompose :
Mlsem_types__.Tvar.MVarSet.t ->
Mlsem_types__.Tvar.Subst.t ->
Mlsem_types__.Tvar.Subst.t ->
Mlsem_types__.Tvar.Subst.t listdecompose mono s1 s2 returns the substitutions s whose domains is disjoint from mono, and such that the composition of s and s1 yields s2. A non-empty result means that s1 is more general than s2, which is used to compare candidate solutions of tallying instances.
factorize (pvs,nvs) t splits t into a part that can be factorized by pvs and nvs, and a remainder: it returns (t1,t2) such that t is equivalent to (/\pvs /\ ~\/nvs /\ t1) \/ t2. The factorized part gathers the DNF lines of t in which every variable of pvs occurs positively and every variable of nvs occurs negatively, with those occurrences removed.