catala/compiler/desugared/ast.ml

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

247 lines
7.8 KiB
OCaml
Raw Normal View History

2020-06-22 17:16:55 +03:00
(* This file is part of the Catala compiler, a specification language for tax
and social benefits computation rules. Copyright (C) 2020 Inria, contributor:
Nicolas Chataing <nicolas.chataing@ens.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. *)
2020-12-14 19:00:42 +03:00
(** Abstract syntax tree of the desugared representation *)
2021-01-21 23:33:04 +03:00
open Utils
open Shared_ast
2020-11-23 14:20:38 +03:00
2020-12-14 19:00:42 +03:00
(** {1 Names, Maps and Keys} *)
module IdentMap : Map.S with type key = String.t = Map.Make (String)
module RuleName : Uid.Id with type info = Uid.MarkedString.info =
Uid.Make (Uid.MarkedString) ()
2020-12-14 19:00:42 +03:00
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)
2020-11-23 14:20:38 +03:00
(** Inside a scope, a definition can refer either to a scope def, or a subscope
def *)
module ScopeDef = struct
type t =
| Var of ScopeVar.t * StateName.t option
| SubScopeVar of SubScopeName.t * ScopeVar.t * Pos.t
(** In this case, the [ScopeVar.t] lives inside the context of the
subscope's original declaration *)
2020-11-23 14:20:38 +03:00
let compare x y =
match x, y with
| Var (x, None), Var (y, None)
| Var (x, Some _), Var (y, None)
| Var (x, None), Var (y, Some _)
| Var (x, _), SubScopeVar (_, y, _)
| SubScopeVar (_, x, _), Var (y, _) ->
ScopeVar.compare x y
| Var (x, Some sx), Var (y, Some sy) ->
let cmp = ScopeVar.compare x y in
if cmp = 0 then StateName.compare sx sy else cmp
| SubScopeVar (x', x, _), SubScopeVar (y', y, _) ->
let cmp = SubScopeName.compare x' y' in
if cmp = 0 then ScopeVar.compare x y else cmp
2020-11-23 14:20:38 +03:00
let get_position x =
match x with
| Var (x, None) -> Marked.get_mark (ScopeVar.get_info x)
| Var (_, Some sx) -> Marked.get_mark (StateName.get_info sx)
| SubScopeVar (_, _, pos) -> pos
let format_t fmt x =
2020-11-23 14:20:38 +03:00
match x with
| Var (v, None) -> ScopeVar.format_t fmt v
| Var (v, Some sv) ->
Format.fprintf fmt "%a.%a" ScopeVar.format_t v StateName.format_t sv
| SubScopeVar (s, v, _) ->
Format.fprintf fmt "%a.%a" SubScopeName.format_t s ScopeVar.format_t v
2020-11-23 14:20:38 +03:00
2020-11-23 18:12:45 +03:00
let hash x =
match x with
| Var (v, None) -> ScopeVar.hash v
| Var (v, Some sv) -> Int.logxor (ScopeVar.hash v) (StateName.hash sv)
| SubScopeVar (w, v, _) ->
Int.logxor (SubScopeName.hash w) (ScopeVar.hash v)
2020-11-23 14:20:38 +03:00
end
2020-12-14 19:00:42 +03:00
module ScopeDefMap : Map.S with type key = ScopeDef.t = Map.Make (ScopeDef)
module ScopeDefSet : Set.S with type elt = ScopeDef.t = Set.Make (ScopeDef)
(** {1 AST} *)
2020-06-22 17:16:55 +03:00
2022-08-25 17:08:08 +03:00
type location = desugared glocation
module LocationSet : Set.S with type elt = location Marked.pos =
Set.Make (struct
type t = location Marked.pos
2022-08-25 17:08:08 +03:00
let compare = Expr.compare_location
end)
type expr = (desugared, untyped mark) gexpr
2022-08-25 17:08:08 +03:00
module ExprMap = Map.Make (struct
type t = expr
2022-02-28 20:34:32 +03:00
2022-08-25 17:08:08 +03:00
let compare = Expr.compare
end)
2022-02-28 20:34:32 +03:00
type exception_situation =
| BaseCase
| ExceptionToLabel of LabelName.t Marked.pos
| ExceptionToRule of RuleName.t Marked.pos
type label_situation = ExplicitlyLabeled of LabelName.t Marked.pos | Unlabeled
2020-11-25 12:10:27 +03:00
type rule = {
rule_id : RuleName.t;
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
rule_just : expr boxed;
rule_cons : expr boxed;
rule_parameter : (expr Var.t * typ) option;
rule_exception : exception_situation;
rule_label : label_situation;
2020-11-25 12:10:27 +03:00
}
2020-06-22 17:16:55 +03:00
module Rule = struct
type t = rule
(** Structural equality (otherwise, you should just compare the [rule_id]
fields) *)
let compare r1 r2 =
match r1.rule_parameter, r2.rule_parameter with
| None, None -> (
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
let j1 = Expr.unbox r1.rule_just in
let j2 = Expr.unbox r2.rule_just in
match Expr.compare j1 j2 with
| 0 ->
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
let c1 = Expr.unbox r1.rule_cons in
let c2 = Expr.unbox r2.rule_cons in
Expr.compare c1 c2
| n -> n)
| Some (v1, t1), Some (v2, t2) -> (
match Shared_ast.Expr.compare_typ t1 t2 with
| 0 -> (
let open Bindlib in
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
let b1 = unbox (bind_var v1 (Expr.Box.inj r1.rule_just)) in
let b2 = unbox (bind_var v2 (Expr.Box.inj r2.rule_just)) in
2022-08-25 17:08:08 +03:00
let _, j1, j2 = unbind2 b1 b2 in
match Expr.compare j1 j2 with
| 0 ->
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
let b1 = unbox (bind_var v1 (Expr.Box.inj r1.rule_cons)) in
let b2 = unbox (bind_var v2 (Expr.Box.inj r2.rule_cons)) in
2022-08-25 17:08:08 +03:00
let _, c1, c2 = unbind2 b1 b2 in
Expr.compare c1 c2
| n -> n)
| n -> n)
| None, Some _ -> -1
| Some _, None -> 1
end
let empty_rule (pos : Pos.t) (have_parameter : typ option) : rule =
2020-11-25 12:10:27 +03:00
{
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
rule_just = Expr.box (ELit (LBool false), Untyped { pos });
rule_cons = Expr.box (ELit LEmptyError, Untyped { pos });
rule_parameter =
2022-02-28 20:34:32 +03:00
(match have_parameter with
| Some typ -> Some (Var.make "dummy", typ)
2022-02-28 20:34:32 +03:00
| None -> None);
rule_exception = BaseCase;
rule_id = RuleName.fresh ("empty", pos);
rule_label = Unlabeled;
2020-11-25 12:10:27 +03:00
}
2020-08-10 00:01:42 +03:00
let always_false_rule (pos : Pos.t) (have_parameter : typ option) : rule =
{
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
rule_just = Expr.box (ELit (LBool true), Untyped { pos });
rule_cons = Expr.box (ELit (LBool false), Untyped { pos });
rule_parameter =
2022-02-28 20:34:32 +03:00
(match have_parameter with
| Some typ -> Some (Var.make "dummy", typ)
2022-02-28 20:34:32 +03:00
| None -> None);
rule_exception = BaseCase;
rule_id = RuleName.fresh ("always_false", pos);
rule_label = Unlabeled;
}
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 assertion = expr boxed
2020-06-22 17:16:55 +03:00
type variation_typ = Increasing | Decreasing
type reference_typ = Decree | Law
type meta_assertion =
| FixedBy of reference_typ Marked.pos
| VariesWith of unit * variation_typ Marked.pos option
2020-06-22 17:16:55 +03:00
type scope_def = {
scope_def_rules : rule RuleMap.t;
scope_def_typ : typ;
scope_def_is_condition : bool;
scope_def_io : Scopelang.Ast.io;
}
type var_or_states = WholeVar | States of StateName.t list
2020-06-22 17:16:55 +03:00
type scope = {
scope_vars : var_or_states ScopeVarMap.t;
scope_sub_scopes : ScopeName.t SubScopeMap.t;
scope_uid : ScopeName.t;
scope_defs : scope_def ScopeDefMap.t;
2020-06-22 17:16:55 +03:00
scope_assertions : assertion list;
2020-09-13 01:33:56 +03:00
scope_meta_assertions : meta_assertion list;
2020-06-22 17:16:55 +03:00
}
type program = { program_scopes : scope ScopeMap.t; program_ctx : decl_ctx }
2022-10-10 16:15:36 +03:00
let rec locations_used e : LocationSet.t =
match e with
| ELocation l, m -> LocationSet.singleton (l, Expr.mark_pos m)
| EAbs (binder, _), _ ->
let _, body = Bindlib.unmbind binder in
locations_used body
2022-10-10 16:15:36 +03:00
| e ->
Expr.shallow_fold
(fun e -> LocationSet.union (locations_used e))
e LocationSet.empty
2020-11-27 13:37:21 +03:00
let free_variables (def : rule RuleMap.t) : Pos.t ScopeDefMap.t =
let add_locs (acc : Pos.t ScopeDefMap.t) (locs : LocationSet.t) :
Pos.t ScopeDefMap.t =
LocationSet.fold
(fun (loc, loc_pos) acc ->
2020-11-27 13:37:21 +03:00
ScopeDefMap.add
(match loc with
2022-08-25 17:08:08 +03:00
| DesugaredScopeVar (v, st) -> ScopeDef.Var (Marked.unmark v, st)
| SubScopeVar (_, sub_index, sub_var) ->
ScopeDef.SubScopeVar
( Marked.unmark sub_index,
Marked.unmark sub_var,
Marked.get_mark sub_index ))
2020-11-27 13:37:21 +03:00
loc_pos acc)
locs acc
in
RuleMap.fold
(fun _ rule acc ->
2020-11-27 18:27:10 +03:00
let locs =
LocationSet.union
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
(locations_used (Expr.unbox rule.rule_just))
(locations_used (Expr.unbox rule.rule_cons))
2020-11-27 18:27:10 +03:00
in
add_locs acc locs)
2020-11-27 13:37:21 +03:00
def ScopeDefMap.empty