catala/compiler/shared_ast/definitions.ml

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

404 lines
13 KiB
OCaml
Raw Normal View History

(* This file is part of the Catala compiler, a specification language for tax
and social benefits computation rules. Copyright (C) 2020-2022 Inria,
contributor: Denis Merigoux <denis.merigoux@inria.fr>, Alain Delaët-Tixeuil
<alain.delaet--tixeuil@inria.fr>, Louis Gesbert <louis.gesbert@inria.fr>
Licensed under the Apache License, Version 2.0 (the "License"); you may not
use this file except in compliance with the License. You may obtain a copy of
the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
License for the specific language governing permissions and limitations under
the License. *)
(** This module defines generic types for types, literals and expressions shared
through several of the different ASTs. *)
(* Doesn't define values, so OK to have without an mli *)
open Utils
module Runtime = Runtime_ocaml.Runtime
module ScopeName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
Make scopes directly callable Quite a few changes are included here, some of which have some extra implications visible in the language: - adds the `Scope of { -- input_v: value; ... }` construct in the language - handle it down the pipeline: * `ScopeCall` in the surface AST * `EScopeCall` in desugared and scopelang * expressions are now traversed to detect dependencies between scopes * transformed into a normal function call in dcalc - defining a scope now implicitely defines a structure with the same name, with the output variables of the scope defined as fields. This allows us to type the return value from a scope call and access its fields easily. * the implications are mostly in surface/name_resolution.ml code-wise * the `Scope_out` struct that was defined in scope_to_dcalc is no longer needed/used and the fields are no longer renamed (changes some outputs; the explicit suffix for variables with multiple states is ignored as well) * one benefit is that disambiguation works just like for structures when there are conflicts on field names * however, it's now a conflict if a scope and a structure have the same name (side-note: issues with conflicting enum / struct names or scope variables / subscope names were silent and are now properly reported) - you can consequently use scope names as types for variables as well. Writing literals is not allowed though, they can only be obtained by calling the scope. Remaining TODOs: - context variables are not handled properly at the moment - error handling on invalid calls - tests show a small error message regression; lots of examples will need tweaking to avoid scope/struct name or struct fields / output variable conflicts - add a `->` syntax to make struct field access distinct from scope output var access, enforced with typing. This is expected to reduce confusion of users and add a little typing precision. - document the new syntax & implications (tutorial, cheat-sheet) - a consequence of the changes is that subscope variables also can now be typed. A possible future evolution / simplification would be to rewrite subscopes as explicit scope calls early in the pipeline. That could also allow to manipulate them as expressions (bind them in let-ins, return them...)
2022-10-21 16:47:17 +03:00
module ScopeSet : Set.S with type elt = ScopeName.t = Set.Make (ScopeName)
module ScopeMap : Map.S with type key = ScopeName.t = Map.Make (ScopeName)
module StructName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module StructFieldName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module StructMap : Map.S with type key = StructName.t = Map.Make (StructName)
module EnumName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module EnumConstructor : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module EnumMap : Map.S with type key = EnumName.t = Map.Make (EnumName)
(** Only used by surface *)
module RuleName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module RuleMap : Map.S with type key = RuleName.t = Map.Make (RuleName)
module RuleSet : Set.S with type elt = RuleName.t = Set.Make (RuleName)
module LabelName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module LabelMap : Map.S with type key = LabelName.t = Map.Make (LabelName)
module LabelSet : Set.S with type elt = LabelName.t = Set.Make (LabelName)
(** Only used by desugared/scopelang *)
module ScopeVar : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
2022-08-25 17:08:08 +03:00
module ScopeVarSet : Set.S with type elt = ScopeVar.t = Set.Make (ScopeVar)
module ScopeVarMap : Map.S with type key = ScopeVar.t = Map.Make (ScopeVar)
module SubScopeName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
module SubScopeNameSet : Set.S with type elt = SubScopeName.t =
Set.Make (SubScopeName)
module SubScopeMap : Map.S with type key = SubScopeName.t =
Map.Make (SubScopeName)
module StructFieldMap : Map.S with type key = StructFieldName.t =
Map.Make (StructFieldName)
module EnumConstructorMap : Map.S with type key = EnumConstructor.t =
Map.Make (EnumConstructor)
module StateName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
(** {1 Abstract syntax tree} *)
(** {2 Types} *)
type typ_lit = TBool | TUnit | TInt | TRat | TMoney | TDate | TDuration
type typ = naked_typ Marked.pos
and naked_typ =
| TLit of typ_lit
| TTuple of typ list
| TStruct of StructName.t
| TEnum of EnumName.t
| TOption of typ
| TArrow of typ * typ
| TArray of typ
| TAny
(** {2 Constants and operators} *)
type date = Runtime.date
type duration = Runtime.duration
type op_kind =
| KInt
| KRat
| KMoney
| KDate
| KDuration (** All ops don't have a KDate and KDuration. *)
type ternop = Fold
type binop =
| And
| Or
| Xor
| Add of op_kind
| Sub of op_kind
| Mult of op_kind
| Div of op_kind
| Lt of op_kind
| Lte of op_kind
| Gt of op_kind
| Gte of op_kind
| Eq
| Neq
| Map
| Concat
| Filter
type log_entry =
| VarDef of naked_typ
(** During code generation, we need to know the type of the variable being
logged for embedding *)
| BeginCall
| EndCall
| PosRecordIfTrueBool
type unop =
| Not
| Minus of op_kind
| Log of log_entry * Uid.MarkedString.info list
| Length
| IntToRat
| MoneyToRat
| RatToMoney
| GetDay
| GetMonth
| GetYear
| FirstDayOfMonth
| LastDayOfMonth
| RoundMoney
| RoundDecimal
type operator = Ternop of ternop | Binop of binop | Unop of unop
type except = ConflictError | EmptyError | NoValueProvided | Crash
(** {2 Generic expressions} *)
(** Define a common base type for the expressions in most passes of the compiler *)
type desugared = [ `Desugared ]
type scopelang = [ `Scopelang ]
type dcalc = [ `Dcalc ]
type lcalc = [ `Lcalc ]
type 'a any = [< desugared | scopelang | dcalc | lcalc ] as 'a
(** Literals are the same throughout compilation except for the [LEmptyError]
case which is eliminated midway through. *)
2022-07-27 13:18:14 +03:00
type 'a glit =
| LBool : bool -> 'a glit
| LEmptyError : [< desugared | scopelang | dcalc ] glit
| LInt : Runtime.integer -> 'a glit
| LRat : Runtime.decimal -> 'a glit
| LMoney : Runtime.money -> 'a glit
| LUnit : 'a glit
| LDate : date -> 'a glit
| LDuration : duration -> 'a glit
2022-07-27 13:18:14 +03:00
(** Locations are handled differently in [desugared] and [scopelang] *)
type 'a glocation =
| DesugaredScopeVar :
ScopeVar.t Marked.pos * StateName.t option
-> desugared glocation
| ScopelangScopeVar : ScopeVar.t Marked.pos -> scopelang glocation
| SubScopeVar :
ScopeName.t * SubScopeName.t Marked.pos * ScopeVar.t Marked.pos
-> [< desugared | scopelang ] glocation
type ('a, 't) gexpr = (('a, 't) naked_gexpr, 't) Marked.t
(** General expressions: groups all expression cases of the different ASTs, and
uses a GADT to eliminate irrelevant cases for each one. The ['t] annotations
are also totally unconstrained at this point. The dcalc exprs, for example,
are then defined with [type naked_expr = dcalc naked_gexpr] plus the
annotations.
A few tips on using this GADT:
- To write a function that handles cases from different ASTs, explicit the
type variables: [fun (type a) (x: a naked_gexpr) -> ...]
- For recursive functions, you may need to additionally explicit the
generalisation of the variable: [let rec f: type a . a naked_gexpr -> ...] *)
and ('a, 't) naked_gexpr =
(* Constructors common to all ASTs *)
| ELit : 'a glit -> ('a any, 't) naked_gexpr
| EApp : {
f : ('a, 't) gexpr;
args : ('a, 't) gexpr list;
}
-> ('a any, 't) naked_gexpr
| EOp : operator -> ('a any, 't) naked_gexpr
| EArray : ('a, 't) gexpr list -> ('a any, 't) naked_gexpr
| EVar : ('a, 't) naked_gexpr Bindlib.var -> ('a any, 't) naked_gexpr
| EAbs : {
binder : (('a, 't) naked_gexpr, ('a, 't) gexpr) Bindlib.mbinder;
tys : typ list;
}
-> ('a any, 't) naked_gexpr
| EIfThenElse : {
cond : ('a, 't) gexpr;
etrue : ('a, 't) gexpr;
efalse : ('a, 't) gexpr;
}
-> ('a any, 't) naked_gexpr
| EStruct : {
name : StructName.t;
fields : ('a, 't) gexpr StructFieldMap.t;
}
-> ('a any, 't) naked_gexpr
| EStructAccess : {
name : StructName.t;
e : ('a, 't) gexpr;
field : StructFieldName.t;
}
-> ('a any, 't) naked_gexpr
| EInj : {
name : EnumName.t;
e : ('a, 't) gexpr;
cons : EnumConstructor.t;
}
-> ('a any, 't) naked_gexpr
| EMatch : {
name : EnumName.t;
e : ('a, 't) gexpr;
cases : ('a, 't) gexpr EnumConstructorMap.t;
}
-> ('a any, 't) naked_gexpr
(* Early stages *)
| ELocation :
'a glocation
-> (([< desugared | scopelang ] as 'a), 't) naked_gexpr
| EScopeCall : {
scope : ScopeName.t;
args : ('a, 't) gexpr ScopeVarMap.t;
}
Make scopes directly callable Quite a few changes are included here, some of which have some extra implications visible in the language: - adds the `Scope of { -- input_v: value; ... }` construct in the language - handle it down the pipeline: * `ScopeCall` in the surface AST * `EScopeCall` in desugared and scopelang * expressions are now traversed to detect dependencies between scopes * transformed into a normal function call in dcalc - defining a scope now implicitely defines a structure with the same name, with the output variables of the scope defined as fields. This allows us to type the return value from a scope call and access its fields easily. * the implications are mostly in surface/name_resolution.ml code-wise * the `Scope_out` struct that was defined in scope_to_dcalc is no longer needed/used and the fields are no longer renamed (changes some outputs; the explicit suffix for variables with multiple states is ignored as well) * one benefit is that disambiguation works just like for structures when there are conflicts on field names * however, it's now a conflict if a scope and a structure have the same name (side-note: issues with conflicting enum / struct names or scope variables / subscope names were silent and are now properly reported) - you can consequently use scope names as types for variables as well. Writing literals is not allowed though, they can only be obtained by calling the scope. Remaining TODOs: - context variables are not handled properly at the moment - error handling on invalid calls - tests show a small error message regression; lots of examples will need tweaking to avoid scope/struct name or struct fields / output variable conflicts - add a `->` syntax to make struct field access distinct from scope output var access, enforced with typing. This is expected to reduce confusion of users and add a little typing precision. - document the new syntax & implications (tutorial, cheat-sheet) - a consequence of the changes is that subscope variables also can now be typed. A possible future evolution / simplification would be to rewrite subscopes as explicit scope calls early in the pipeline. That could also allow to manipulate them as expressions (bind them in let-ins, return them...)
2022-10-21 16:47:17 +03:00
-> (([< desugared | scopelang ] as 'a), 't) naked_gexpr
(* Lambda-like *)
| EAssert : ('a, 't) gexpr -> (([< dcalc | lcalc ] as 'a), 't) naked_gexpr
(* Default terms *)
| EDefault : {
excepts : ('a, 't) gexpr list;
just : ('a, 't) gexpr;
cons : ('a, 't) gexpr;
}
-> (([< desugared | scopelang | dcalc ] as 'a), 't) naked_gexpr
| EErrorOnEmpty :
('a, 't) gexpr
-> (([< desugared | scopelang | dcalc ] as 'a), 't) naked_gexpr
(* Lambda calculus with exceptions *)
| ETuple : ('a, 't) gexpr list -> ((lcalc as 'a), 't) naked_gexpr
| ETupleAccess : {
e : ('a, 't) gexpr;
index : int;
size : int;
}
-> ((lcalc as 'a), 't) naked_gexpr
| ERaise : except -> ((lcalc as 'a), 't) naked_gexpr
| ECatch : {
body : ('a, 't) gexpr;
exn : except;
handler : ('a, 't) gexpr;
}
-> ((lcalc as 'a), 't) naked_gexpr
Swap boxing and annotations in expressions This was the only reasonable solution I found to the issue raised [here](https://github.com/CatalaLang/catala/pull/334#discussion_r987175884). This was a pretty tedious rewrite, but it should now ensure we are doing things correctly. As a bonus, the "smart" expression constructors are now used everywhere to build expressions (so another refactoring like this one should be much easier) and this makes the code overall feel more straightforward (`Bindlib.box_apply` or `let+` no longer need to be visible!) --- Basically, we were using values of type `gexpr box = naked_gexpr marked box` throughout when (re-)building expressions. This was done 99% of the time by using `Bindlib.box_apply add_mark naked_e` right after building `naked_e`. In lots of places, we needed to recover the annotation of this expression later on, typically to build its parent term (to inherit the position, or build the type). Since it wasn't always possible to wrap these uses within `box_apply` (esp. as bindlib boxes aren't a monad), here and there we had to call `Bindlib.unbox`, just to recover the position or type. This had the very unpleasant effect of forcing the resolution of the whole box (including applying any stored closures) to reach the top-level annotation which isn't even dependant on specific variable bindings. Then, generally, throwing away the result. Therefore, the change proposed here transforms - `naked_gexpr marked Bindlib.box` into - `naked_gexpr Bindlib.box marked` (aliased to `boxed_gexpr` or `gexpr boxed` for convenience) This means only 1. not fitting the mark into the box right away when building, and 2. accessing the top-level mark directly without unboxing The functions for building terms from module `Shared_ast.Expr` could be changed easily. But then they needed to be consistently used throughout, without manually building terms through `Bindlib.apply_box` -- which covers most of the changes in this patch. `Expr.Box.inj` is provided to swap back to a box, before binding for example. Additionally, this gives a 40% speedup on `make -C examples pass_all_tests`, which hints at the amount of unnecessary work we were doing --'
2022-10-06 20:13:45 +03:00
type ('a, 't) boxed_gexpr = (('a, 't) naked_gexpr Bindlib.box, 't) Marked.t
(** The annotation is lifted outside of the box for expressions *)
type 'e boxed = ('a, 't) boxed_gexpr constraint 'e = ('a, 't) gexpr
(** [('a, 't) gexpr boxed] is [('a, 't) boxed_gexpr]. The difference with
[('a, 't) gexpr Bindlib.box] is that the annotations is outside of the box,
and can therefore be accessed without the need to resolve the box *)
type ('e, 'b) binder = (('a, 't) naked_gexpr, 'b) Bindlib.binder
constraint 'e = ('a, 't) gexpr
(** The expressions use the {{:https://lepigre.fr/ocaml-bindlib/} Bindlib}
library, based on higher-order abstract syntax *)
type ('e, 'b) mbinder = (('a, 't) naked_gexpr, 'b) Bindlib.mbinder
constraint 'e = ('a, 't) gexpr
(** {2 Markings} *)
type untyped = { pos : Pos.t } [@@ocaml.unboxed]
type typed = { pos : Pos.t; ty : typ }
(** The generic type of AST markings. Using a GADT allows functions to be
polymorphic in the marking, but still do transformations on types when
appropriate. Expected to fill the ['t] parameter of [gexpr] and [gexpr] (a
['t] annotation different from this type is used in the middle of the typing
processing, but all visible ASTs should otherwise use this. *)
type _ mark = Untyped : untyped -> untyped mark | Typed : typed -> typed mark
(** Useful for errors and printing, for example *)
type any_expr = AnyExpr : (_, _ mark) gexpr -> any_expr
(** {2 Higher-level program structure} *)
(** Constructs scopes and programs on top of expressions. The ['e] type
parameter throughout is expected to match instances of the [gexpr] type
defined above. Markings are constrained to the [mark] GADT defined above.
Note that this structure is at the moment only relevant for [dcalc] and
[lcalc], as [scopelang] has its own scope structure, as the name implies. *)
(** This kind annotation signals that the let-binding respects a structural
invariant. These invariants concern the shape of the expression in the
let-binding, and are documented below. *)
type scope_let_kind =
| DestructuringInputStruct (** [let x = input.field]*)
| ScopeVarDefinition (** [let x = error_on_empty e]*)
| SubScopeVarDefinition
(** [let s.x = fun _ -> e] or [let s.x = error_on_empty e] for input-only
subscope variables. *)
| CallingSubScope (** [let result = s ({ x = s.x; y = s.x; ...}) ]*)
| DestructuringSubScopeResults (** [let s.x = result.x ]**)
| Assertion (** [let _ = assert e]*)
2022-08-16 17:54:42 +03:00
type 'e scope_let = {
scope_let_kind : scope_let_kind;
scope_let_typ : typ;
scope_let_expr : 'e;
scope_let_next : ('e, 'e scope_body_expr) binder;
scope_let_pos : Pos.t;
}
constraint 'e = (_ any, _ mark) gexpr
(** This type is parametrized by the expression type so it can be reused in
later intermediate representations. *)
(** A scope let-binding has all the information necessary to make a proper
let-binding expression, plus an annotation for the kind of the let-binding
that comes from the compilation of a {!module: Scopelang.Ast} statement. *)
2022-08-16 17:54:42 +03:00
and 'e scope_body_expr =
| Result of 'e
2022-08-16 17:54:42 +03:00
| ScopeLet of 'e scope_let
constraint 'e = (_ any, _ mark) gexpr
2022-08-16 17:54:42 +03:00
type 'e scope_body = {
scope_body_input_struct : StructName.t;
scope_body_output_struct : StructName.t;
scope_body_expr : ('e, 'e scope_body_expr) binder;
}
constraint 'e = (_ any, _ mark) gexpr
(** Instead of being a single expression, we give a little more ad-hoc structure
to the scope body by decomposing it in an ordered list of let-bindings, and
a result expression that uses the let-binded variables. The first binder is
the argument of type [scope_body_input_struct]. *)
2022-08-16 17:54:42 +03:00
type 'e scope_def = {
scope_name : ScopeName.t;
2022-08-16 17:54:42 +03:00
scope_body : 'e scope_body;
scope_next : ('e, 'e scopes) binder;
}
constraint 'e = (_ any, _ mark) gexpr
(** Finally, we do the same transformation for the whole program for the kinded
lets. This permit us to use bindlib variables for scopes names. *)
2022-08-16 17:54:42 +03:00
and 'e scopes =
| Nil
| ScopeDef of 'e scope_def
constraint 'e = (_ any, _ mark) gexpr
type struct_ctx = typ StructFieldMap.t StructMap.t
type enum_ctx = typ EnumConstructorMap.t EnumMap.t
type scope_out_struct = {
out_struct_name : StructName.t;
out_struct_fields : StructFieldName.t ScopeVarMap.t;
}
Make scopes directly callable Quite a few changes are included here, some of which have some extra implications visible in the language: - adds the `Scope of { -- input_v: value; ... }` construct in the language - handle it down the pipeline: * `ScopeCall` in the surface AST * `EScopeCall` in desugared and scopelang * expressions are now traversed to detect dependencies between scopes * transformed into a normal function call in dcalc - defining a scope now implicitely defines a structure with the same name, with the output variables of the scope defined as fields. This allows us to type the return value from a scope call and access its fields easily. * the implications are mostly in surface/name_resolution.ml code-wise * the `Scope_out` struct that was defined in scope_to_dcalc is no longer needed/used and the fields are no longer renamed (changes some outputs; the explicit suffix for variables with multiple states is ignored as well) * one benefit is that disambiguation works just like for structures when there are conflicts on field names * however, it's now a conflict if a scope and a structure have the same name (side-note: issues with conflicting enum / struct names or scope variables / subscope names were silent and are now properly reported) - you can consequently use scope names as types for variables as well. Writing literals is not allowed though, they can only be obtained by calling the scope. Remaining TODOs: - context variables are not handled properly at the moment - error handling on invalid calls - tests show a small error message regression; lots of examples will need tweaking to avoid scope/struct name or struct fields / output variable conflicts - add a `->` syntax to make struct field access distinct from scope output var access, enforced with typing. This is expected to reduce confusion of users and add a little typing precision. - document the new syntax & implications (tutorial, cheat-sheet) - a consequence of the changes is that subscope variables also can now be typed. A possible future evolution / simplification would be to rewrite subscopes as explicit scope calls early in the pipeline. That could also allow to manipulate them as expressions (bind them in let-ins, return them...)
2022-10-21 16:47:17 +03:00
type decl_ctx = {
ctx_enums : enum_ctx;
ctx_structs : struct_ctx;
ctx_scopes : scope_out_struct ScopeMap.t;
Make scopes directly callable Quite a few changes are included here, some of which have some extra implications visible in the language: - adds the `Scope of { -- input_v: value; ... }` construct in the language - handle it down the pipeline: * `ScopeCall` in the surface AST * `EScopeCall` in desugared and scopelang * expressions are now traversed to detect dependencies between scopes * transformed into a normal function call in dcalc - defining a scope now implicitely defines a structure with the same name, with the output variables of the scope defined as fields. This allows us to type the return value from a scope call and access its fields easily. * the implications are mostly in surface/name_resolution.ml code-wise * the `Scope_out` struct that was defined in scope_to_dcalc is no longer needed/used and the fields are no longer renamed (changes some outputs; the explicit suffix for variables with multiple states is ignored as well) * one benefit is that disambiguation works just like for structures when there are conflicts on field names * however, it's now a conflict if a scope and a structure have the same name (side-note: issues with conflicting enum / struct names or scope variables / subscope names were silent and are now properly reported) - you can consequently use scope names as types for variables as well. Writing literals is not allowed though, they can only be obtained by calling the scope. Remaining TODOs: - context variables are not handled properly at the moment - error handling on invalid calls - tests show a small error message regression; lots of examples will need tweaking to avoid scope/struct name or struct fields / output variable conflicts - add a `->` syntax to make struct field access distinct from scope output var access, enforced with typing. This is expected to reduce confusion of users and add a little typing precision. - document the new syntax & implications (tutorial, cheat-sheet) - a consequence of the changes is that subscope variables also can now be typed. A possible future evolution / simplification would be to rewrite subscopes as explicit scope calls early in the pipeline. That could also allow to manipulate them as expressions (bind them in let-ins, return them...)
2022-10-21 16:47:17 +03:00
}
2022-08-16 17:54:42 +03:00
type 'e program = { decl_ctx : decl_ctx; scopes : 'e scopes }