(** Define base utilities for the extraction *) open Pure open TranslateCore module C = Contexts module RegionVarId = T.RegionVarId module F = Format (** The local logger *) let log = L.pure_to_extract_log type region_group_info = { id : RegionGroupId.id; (** The id of the region group. Note that a simple way of generating unique names for backward functions is to use the region group ids. *) region_names : string option list; (** The names of the region variables included in this group. Note that names are not always available... *) } module StringSet = Collections.MakeSet (Collections.OrderedString) module StringMap = Collections.MakeMap (Collections.OrderedString) type name = Names.name type type_name = Names.type_name type global_name = Names.global_name type fun_name = Names.fun_name (** Characterizes a declaration. Is in particular useful to derive the proper keywords to introduce the declarations/definitions. *) type decl_kind = | SingleNonRec (** A single, non-recursive definition. F*: [let x = ...] Coq: [Definition x := ...] *) | SingleRec (** A single, recursive definition. F*: [let rec x = ...] Coq: [Fixpoint x := ...] *) | MutRecFirst (** The first definition of a group of mutually-recursive definitions. F*: [type x0 = ... and x1 = ...] Coq: [Fixpoint x0 := ... with x1 := ...] *) | MutRecInner (** An inner definition in a group of mutually-recursive definitions. *) | MutRecLast (** The last definition in a group of mutually-recursive definitions. We need this because in some theorem provers like Coq, we need to delimit group of mutually recursive definitions (in particular, we need to insert an end delimiter). *) | Assumed (** An assumed definition. F*: [assume val x] Coq: [Axiom x : Type.] *) | Declared (** Declare a type in an interface or a module signature. Rem.: for now, in Coq, we don't declare module signatures: we thus assume the corresponding declarations. F*: [val x : Type0] Coq: [Axiom x : Type.] *) (** Return [true] if the declaration is the last from its group of declarations. We need this because in some provers (e.g., Coq), we need to delimit the end of a (group of) definition(s) (in Coq: with a "."). *) let decl_is_last_from_group (kind : decl_kind) : bool = match kind with | SingleNonRec | SingleRec | MutRecLast | Assumed | Declared -> true | MutRecFirst | MutRecInner -> false let decl_is_from_rec_group (kind : decl_kind) : bool = match kind with | SingleNonRec | Assumed | Declared -> false | SingleRec | MutRecFirst | MutRecInner | MutRecLast -> true let decl_is_from_mut_rec_group (kind : decl_kind) : bool = match kind with | SingleNonRec | SingleRec | Assumed | Declared -> false | MutRecFirst | MutRecInner | MutRecLast -> true let decl_is_first_from_group (kind : decl_kind) : bool = match kind with | SingleNonRec | SingleRec | MutRecFirst | Assumed | Declared -> true | MutRecLast | MutRecInner -> false (** Return [true] if the declaration is not the last from its group of declarations. We need this because in some provers (e.g., HOL4), we need to delimit the inner declarations (with `/\` for instance). *) let decl_is_not_last_from_group (kind : decl_kind) : bool = not (decl_is_last_from_group kind) type type_decl_kind = Enum | Struct (* TODO: this should be a module we give to a functor! *) (** A formatter's role is twofold: 1. Come up with name suggestions. For instance, provided some information about a function (its basename, information about the region group, etc.) it should come up with an appropriate name for the forward/backward function. It can of course apply many transformations, like changing to camel case/ snake case, adding prefixes/suffixes, etc. 2. Format some specific terms, like constants. TODO: unclear that this is useful now that all the backends are so much entangled in Extract.ml *) type formatter = { bool_name : string; char_name : string; int_name : integer_type -> string; str_name : string; type_decl_kind_to_qualif : decl_kind -> type_decl_kind option -> string option; (** Compute the qualified for a type definition/declaration. For instance: "type", "and", etc. Remark: can return [None] for some backends like HOL4. *) fun_decl_kind_to_qualif : decl_kind -> string option; (** Compute the qualified for a function definition/declaration. For instance: "let", "let rec", "and", etc. Remark: can return [None] for some backends like HOL4. *) field_name : name -> FieldId.id -> string option -> string; (** Inputs: - type name - field id - field name Note that fields don't always have names, but we still need to generate some names if we want to extract the structures to records... We might want to extract such structures to tuples, later, but field access then causes trouble because not all provers accept syntax like [x.3] where [x] is a tuple. *) variant_name : name -> string -> string; (** Inputs: - type name - variant name *) struct_constructor : name -> string; (** Structure constructors are used when constructing structure values. For instance, in F*: {[ type pair = { x : nat; y : nat } let p : pair = Mkpair 0 1 ]} Inputs: - type name *) type_name : type_name -> string; (** Provided a basename, compute a type name. *) global_name : global_name -> string; (** Provided a basename, compute a global name. *) fun_name : fun_name -> int -> LoopId.id option -> int -> region_group_info option -> bool * int -> string; (** Compute the name of a regular (non-assumed) function. Inputs: - function basename (TODO: shouldn't appear for assumed functions?...) - number of loops in the function (useful to check if we need to use indices to derive unique names for the loops for instance - if there is exactly one loop, we don't need to use indices) - loop id (if pertinent) - number of region groups - region group information in case of a backward function ([None] if forward function) - pair: - do we generate the forward function (it may have been filtered)? - the number of *extracted backward functions* (same comment as for the number of loops) The number of extracted backward functions if not necessarily equal to the number of region groups, because we may have filtered some of them. TODO: use the fun id for the assumed functions. *) termination_measure_name : A.FunDeclId.id -> fun_name -> int -> LoopId.id option -> string; (** Generates the name of the termination measure used to prove/reason about termination. The generated code uses this clause where needed, but its body must be defined by the user. F* and Lean only. Inputs: - function id: this is especially useful to identify whether the function is an assumed function or a local function - function basename - the number of loops in the parent function. This is used for the same purpose as in {!field:fun_name}. - loop identifier, if this is for a loop *) decreases_proof_name : A.FunDeclId.id -> fun_name -> int -> LoopId.id option -> string; (** Generates the name of the proof used to prove/reason about termination. The generated code uses this clause where needed, but its body must be defined by the user. Lean only. Inputs: - function id: this is especially useful to identify whether the function is an assumed function or a local function - function basename - the number of loops in the parent function. This is used for the same purpose as in {!field:fun_name}. - loop identifier, if this is for a loop *) trait_decl_name : trait_decl -> string; trait_impl_name : trait_impl -> string; trait_parent_clause_name : trait_decl -> trait_clause -> string; trait_const_name : trait_decl -> string -> string; trait_type_name : trait_decl -> string -> string; trait_method_name : trait_decl -> string -> string; trait_type_clause_name : trait_decl -> string -> trait_clause -> string; opaque_pre : unit -> string; (** TODO: obsolete, remove. The prefix to use for opaque definitions. We need this because for some backends like Lean and Coq, we group opaque definitions in module signatures, meaning that using those definitions requires to prefix them with a module parameter name (such as "opaque_defs."). For instance, if we have an opaque function [f : int -> int], which is used by the non-opaque function [g], we would generate (in Coq): {[ (* The module signature declaring the opaque definitions *) module type OpaqueDefs = { f_fwd : int -> int ... (* Other definitions *) } (* The definitions generated for the non-opaque definitions *) module Funs (opaque: OpaqueDefs) = { let g ... = ... opaque_defs.f_fwd ... } ]} Upon using [f] in [g], we don't directly use the the name "f_fwd", but prefix it with the "opaque_defs." identifier. *) var_basename : StringSet.t -> string option -> ty -> string; (** Generates a variable basename. Inputs: - the set of names used in the context so far - the basename we got from the symbolic execution, if we have one - the type of the variable (can be useful for heuristics, in order not to always use "x" for instance, whenever naming anonymous variables) Note that once the formatter generated a basename, we add an index if necessary to prevent name clashes: the burden of name clashes checks is thus on the caller's side. *) type_var_basename : StringSet.t -> string -> string; (** Generates a type variable basename. *) const_generic_var_basename : StringSet.t -> string -> string; (** Generates a const generic variable basename. *) trait_self_clause_basename : string; trait_clause_basename : StringSet.t -> trait_clause -> string; (** Return a base name for a trait clause. We might add a suffix to prevent collisions. In the traduction we explicitely manipulate the trait clause instances, that is we introduce one input variable for each trait clause. *) append_index : string -> int -> string; (** Appends an index to a name - we use this to generate unique names: when doing so, the role of the formatter is just to concatenate indices to names, the responsability of finding a proper index is delegated to helper functions. *) extract_literal : F.formatter -> bool -> literal -> unit; (** Format a constant value. Inputs: - formatter - [inside]: if [true], the value should be wrapped in parentheses if it is made of an application (ex.: [U32 3]) - the constant value *) extract_unop : (bool -> texpression -> unit) -> F.formatter -> bool -> unop -> texpression -> unit; (** Format a unary operation Inputs: - a formatter for expressions (called on the argument of the unop) - extraction context (see below) - formatter - expression formatter - [inside] - unop - argument *) extract_binop : (bool -> texpression -> unit) -> F.formatter -> bool -> E.binop -> integer_type -> texpression -> texpression -> unit; (** Format a binary operation Inputs: - a formatter for expressions (called on the arguments of the binop) - extraction context (see below) - formatter - expression formatter - [inside] - binop - argument 0 - argument 1 *) } (** We use identifiers to look for name clashes *) type id = | GlobalId of A.GlobalDeclId.id | FunId of fun_id | TerminationMeasureId of (A.fun_id * LoopId.id option) (** The definition which provides the decreases/termination measure. We insert calls to this clause to prove/reason about termination: the body of those clauses must be defined by the user, in the proper files. More specifically: - in F*, this is the content of the [decreases] clause. Example: ======== {[ let rec sum (ls : list nat) : Tot nat (decreases ls) = ... ]} - in Lean, this is the content of the [termination_by] clause. *) | DecreasesProofId of (A.fun_id * LoopId.id option) (** The definition which provides the decreases/termination proof. We insert calls to this clause to prove/reason about termination: the body of those clauses must be defined by the user, in the proper files. More specifically: - F* doesn't use this. - in Lean, this is the tactic used by the [decreases_by] annotations. *) | TypeId of type_id | StructId of type_id (** We use this when we manipulate the names of the structure constructors. For instance, in F*: {[ type pair = { x: nat; y : nat } let p : pair = Mkpair 0 1 ]} *) | VariantId of type_id * VariantId.id (** If often happens that variant names must be unique (it is the case in F* ) which is why we register them here. *) | FieldId of type_id * FieldId.id (** If often happens that in the case of structures, the field names must be unique (it is the case in F* ) which is why we register them here. *) | TypeVarId of TypeVarId.id | ConstGenericVarId of ConstGenericVarId.id | VarId of VarId.id | TraitDeclId of TraitDeclId.id | TraitImplId of TraitImplId.id | LocalTraitClauseId of TraitClauseId.id | TraitItemId of TraitDeclId.id * string (** A trait associated item *) | TraitParentClauseId of TraitDeclId.id * TraitClauseId.id | TraitItemClauseId of TraitDeclId.id * string * TraitClauseId.id | TraitSelfClauseId (** Specifically for the clause: [Self : Trait]. For now, we forbid provided methods (methods in trait declarations with a default implementation) from being overriden in trait implementations. We extract trait provided methods such that they take an instance of the trait as input: this instance is given by the trait self clause. For instance: {[ // // Rust // trait ToU64 { fn to_u64(&self) -> u64; // Provided method fn is_pos(&self) -> bool { self.to_u64() > 0 } } // // Generated code // struct ToU64 (T : Type) { to_u64 : T -> u64; } // The trait self clause // vvvvvvvvvvvvvvvvvvvvvv let is_pos (T : Type) (trait_self : ToU64 T) (self : T) : bool = trait_self.to_u64 self > 0 ]} *) | UnknownId (** Used for stored various strings like keywords, definitions which should always be in context, etc. and which can't be linked to one of the above. *) [@@deriving show, ord] module IdOrderedType = struct type t = id let compare = compare_id let to_string = show_id let pp_t = pp_id let show_t = show_id end module IdMap = Collections.MakeMap (IdOrderedType) module IdSet = Collections.MakeSet (IdOrderedType) (** The names map stores the mappings from names to identifiers and vice-versa. We use it for lookups (during the translation) and to check for name clashes. [id_to_name] is for debugging. *) type names_map = { id_to_name : string IdMap.t; name_to_id : id StringMap.t; (** The name to id map is used to look for name clashes, and generate nice debugging messages: if there is a name clash, it is useful to know precisely which identifiers are mapped to the same name... *) names_set : StringSet.t; opaque_ids : IdSet.t; (** TODO: this is obsolete. Remove. The set of opaque definitions. See {!formatter.opaque_pre} for detailed explanations about why we need to know which definitions are opaque to compute names. Also note that the opaque ids don't contain the ids of the assumed definitions. In practice, assumed definitions are opaque_defs. However, they are not grouped in the opaque module, meaning we never need to prefix them (with, say, "opaque_defs."): we thus consider them as non-opaque with regards to the names map. *) } (** Small helper to report name collision *) let report_name_collision (id_to_string : id -> string) (id1 : id) (id2 : id) (name : string) : unit = let id1 = "\n- " ^ id_to_string id1 in let id2 = "\n- " ^ id_to_string id2 in let err = "Name clash detected: the following identifiers are bound to the same name \ \"" ^ name ^ "\":" ^ id1 ^ id2 ^ "\nYou may want to rename some of your definitions, or report an issue." in log#serror err; raise (Failure err) let names_map_get_id_from_name (name : string) (nm : names_map) : id option = StringMap.find_opt name nm.name_to_id let names_map_check_collision (id_to_string : id -> string) (id : id) (name : string) (nm : names_map) : unit = match names_map_get_id_from_name name nm with | None -> () (* Ok *) | Some clash -> (* There is a clash: print a nice debugging message for the user *) report_name_collision id_to_string clash id name let names_map_add (id_to_string : id -> string) (is_opaque : bool) (id : id) (name : string) (nm : names_map) : names_map = (* Check if there is a clash *) names_map_check_collision id_to_string id name nm; (* Sanity check *) assert (not (StringSet.mem name nm.names_set)); (* Insert *) let id_to_name = IdMap.add id name nm.id_to_name in let name_to_id = StringMap.add name id nm.name_to_id in let names_set = StringSet.add name nm.names_set in let opaque_ids = if is_opaque then IdSet.add id nm.opaque_ids else nm.opaque_ids in { id_to_name; name_to_id; names_set; opaque_ids } let names_map_add_assumed_type (id_to_string : id -> string) (id : assumed_ty) (name : string) (nm : names_map) : names_map = let is_opaque = false in names_map_add id_to_string is_opaque (TypeId (Assumed id)) name nm let names_map_add_assumed_struct (id_to_string : id -> string) (id : assumed_ty) (name : string) (nm : names_map) : names_map = let is_opaque = false in names_map_add id_to_string is_opaque (StructId (Assumed id)) name nm let names_map_add_assumed_variant (id_to_string : id -> string) (id : assumed_ty) (variant_id : VariantId.id) (name : string) (nm : names_map) : names_map = let is_opaque = false in names_map_add id_to_string is_opaque (VariantId (Assumed id, variant_id)) name nm let names_map_add_function (id_to_string : id -> string) (is_opaque : bool) (fid : fun_id) (name : string) (nm : names_map) : names_map = names_map_add id_to_string is_opaque (FunId fid) name nm (** The unsafe names map stores mappings from identifiers to names which might collide. For some backends and some names, it might be acceptable to have collisions. For instance, in Lean, different records can have fields with the same name because Lean uses the typing information to resolve the ambiguities. This map complements the {!names_map}, which checks for collisions. *) type unsafe_names_map = { id_to_name : string IdMap.t } let unsafe_names_map_add (id : id) (name : string) (nm : unsafe_names_map) : unsafe_names_map = { id_to_name = IdMap.add id name nm.id_to_name } (** Make a (variable) basename unique (by adding an index). We do this in an inefficient manner (by testing all indices starting from 0) but it shouldn't be a bottleneck. Also note that at some point, we thought about trying to reuse names of variables which are not used anymore, like here: {[ let x = ... in ... let x0 = ... in // We could use the name "x" if [x] is not used below ... ]} However it is a good idea to keep things as they are for F*: as F* is designed for extrinsic proofs, a proof about a function follows this function's structure. The consequence is that we often end up copy-pasting function bodies. As in the proofs (in assertions and when calling lemmas) we often need to talk about the "past" (i.e., previous values), it is very useful to generate code where all variable names are assigned at most once. [append]: function to append an index to a string *) let basename_to_unique (names_set : StringSet.t) (append : string -> int -> string) (basename : string) : string = let rec gen (i : int) : string = let s = append basename i in if StringSet.mem s names_set then gen (i + 1) else s in if StringSet.mem basename names_set then gen 0 else basename type fun_name_info = { keep_fwd : bool; num_backs : int } (** Extraction context. Note that the extraction context contains information coming from the LLBC AST (not only the pure AST). This is useful for naming, for instance: we use the region information to generate the names of the backward functions, etc. *) type extraction_ctx = { crate : A.crate; trans_ctx : trans_ctx; names_map : names_map; (** The map for id to names, where we forbid name collisions (ex.: we always forbid function name collisions). *) unsafe_names_map : unsafe_names_map; (** The map for id to names, where we allow name collisions (ex.: we might allow record field name collisions). *) fmt : formatter; indent_incr : int; (** The indent increment we insert whenever we need to indent more *) use_opaque_pre : bool; (** Do we use the "opaque_defs." prefix for the opaque definitions? Opaque function definitions might refer opaque types: if we are in the opaque module, we musn't use the "opaque_defs." prefix, otherwise we use it. Also see {!names_map.opaque_ids}. *) use_dep_ite : bool; (** For Lean: do we use dependent-if then else expressions? Example: {[ if h: b then ... else ... -- ^^ -- makes the if then else dependent ]} *) fun_name_info : fun_name_info PureUtils.RegularFunIdMap.t; (** Information used to filter and name functions - we use it to print comments in the generated code, to help link the generated code to the original code (information such as: "this function is the backward function of ...", or "this function is the merged forward/backward function of ..." in case a Rust function only has one backward translation and we filter the forward function because it returns unit. *) trait_decl_id : trait_decl_id option; (** If we are extracting a trait declaration, identifies it *) is_provided_method : bool; trans_types : Pure.type_decl Pure.TypeDeclId.Map.t; trans_funs : pure_fun_translation A.FunDeclId.Map.t; functions_with_decreases_clause : PureUtils.FunLoopIdSet.t; trans_trait_decls : Pure.trait_decl Pure.TraitDeclId.Map.t; trans_trait_impls : Pure.trait_impl Pure.TraitImplId.Map.t; } (** Debugging function, used when communicating name collisions to the user, and also to print ids for internal debugging (in case of lookup miss for instance). *) let id_to_string (id : id) (ctx : extraction_ctx) : string = let global_decls = ctx.trans_ctx.global_context.global_decls in let fun_decls = ctx.trans_ctx.fun_context.fun_decls in let type_decls = ctx.trans_ctx.type_context.type_decls in (* TODO: factorize the pretty-printing with what is in PrintPure *) let get_type_name (id : type_id) : string = match id with | AdtId id -> let def = TypeDeclId.Map.find id type_decls in Print.name_to_string def.name | Assumed aty -> show_assumed_ty aty | Tuple -> raise (Failure "Unreachable") in match id with | GlobalId gid -> let name = (A.GlobalDeclId.Map.find gid global_decls).name in "global name: " ^ Print.global_name_to_string name | FunId fid -> ( match fid with | FromLlbc (fid, lp_id, rg_id) -> let fun_name = match fid with | Regular fid -> Print.fun_name_to_string (A.FunDeclId.Map.find fid fun_decls).name | Assumed aid -> A.show_assumed_fun_id aid in let lp_kind = match lp_id with | None -> "" | Some lp_id -> "loop " ^ LoopId.to_string lp_id ^ ", " in let fwd_back_kind = match rg_id with | None -> "forward" | Some rg_id -> "backward " ^ RegionGroupId.to_string rg_id in "fun name (" ^ lp_kind ^ fwd_back_kind ^ "): " ^ fun_name | Pure fid -> PrintPure.pure_assumed_fun_id_to_string fid) | DecreasesProofId (fid, lid) -> let fun_name = match fid with | Regular fid -> Print.fun_name_to_string (A.FunDeclId.Map.find fid fun_decls).name | Assumed aid -> A.show_assumed_fun_id aid in let loop = match lid with | None -> "" | Some lid -> ", loop: " ^ LoopId.to_string lid in "decreases proof for function: " ^ fun_name ^ loop | TerminationMeasureId (fid, lid) -> let fun_name = match fid with | Regular fid -> Print.fun_name_to_string (A.FunDeclId.Map.find fid fun_decls).name | Assumed aid -> A.show_assumed_fun_id aid in let loop = match lid with | None -> "" | Some lid -> ", loop: " ^ LoopId.to_string lid in "termination measure for function: " ^ fun_name ^ loop | TypeId id -> "type name: " ^ get_type_name id | StructId id -> "struct constructor of: " ^ get_type_name id | VariantId (id, variant_id) -> let variant_name = match id with | Tuple -> raise (Failure "Unreachable") | Assumed Result -> if variant_id = result_return_id then "@result::Return" else if variant_id = result_fail_id then "@result::Fail" else raise (Failure "Unreachable") | Assumed Error -> if variant_id = error_failure_id then "@error::Failure" else if variant_id = error_out_of_fuel_id then "@error::OutOfFuel" else raise (Failure "Unreachable") | Assumed Option -> if variant_id = option_some_id then "@option::Some" else if variant_id = option_none_id then "@option::None" else raise (Failure "Unreachable") | Assumed (State | Vec | Fuel | Array | Slice | Str | Range) -> raise (Failure "Unreachable") | AdtId id -> ( let def = TypeDeclId.Map.find id type_decls in match def.kind with | Struct _ | Opaque -> raise (Failure "Unreachable") | Enum variants -> let variant = VariantId.nth variants variant_id in Print.name_to_string def.name ^ "::" ^ variant.variant_name) in "variant name: " ^ variant_name | FieldId (id, field_id) -> let field_name = match id with | Tuple -> raise (Failure "Unreachable") | Assumed ( State | Result | Error | Fuel | Option | Vec | Array | Slice | Str | Range ) -> (* We can't directly have access to the fields of those types *) raise (Failure "Unreachable") | AdtId id -> ( let def = TypeDeclId.Map.find id type_decls in match def.kind with | Enum _ | Opaque -> raise (Failure "Unreachable") | Struct fields -> let field = FieldId.nth fields field_id in let field_name = match field.field_name with | None -> FieldId.to_string field_id | Some name -> name in Print.name_to_string def.name ^ "." ^ field_name) in "field name: " ^ field_name | UnknownId -> "keyword" | TypeVarId id -> "type_var_id: " ^ TypeVarId.to_string id | ConstGenericVarId id -> "const_generic_var_id: " ^ ConstGenericVarId.to_string id | VarId id -> "var_id: " ^ VarId.to_string id | TraitDeclId id -> "trait_decl_id: " ^ TraitDeclId.to_string id | TraitImplId id -> "trait_impl_id: " ^ TraitImplId.to_string id | LocalTraitClauseId id -> "local_trait_clause_id: " ^ TraitClauseId.to_string id | TraitParentClauseId (id, clause_id) -> "trait_parent_clause_id: decl_id:" ^ TraitDeclId.to_string id ^ ", clause_id: " ^ TraitClauseId.to_string clause_id | TraitItemClauseId (id, item_name, clause_id) -> "trait_item_clause_id: decl_id:" ^ TraitDeclId.to_string id ^ ", item name: " ^ item_name ^ ", clause_id: " ^ TraitClauseId.to_string clause_id | TraitItemId (id, name) -> "trait_item_id: decl_id:" ^ TraitDeclId.to_string id ^ ", type name: " ^ name | TraitSelfClauseId -> "trait_self_clause" (** We might not check for collisions for some specific ids (ex.: field names) *) let allow_collisions (id : id) : bool = match id with | FieldId _ | TraitItemClauseId _ | TraitParentClauseId _ | TraitItemId _ -> !Config.record_fields_short_names | _ -> false let ctx_add (is_opaque : bool) (id : id) (name : string) (ctx : extraction_ctx) : extraction_ctx = (* The id_to_string function to print nice debugging messages if there are * collisions *) let id_to_string (id : id) : string = id_to_string id ctx in (* We do not use the same name map if we allow/disallow collisions. We notably use it for field names: some backends like Lean can use the type information to disambiguate field projections. Remark: what we do is actually subtle. Taking the example of fields: - we allow fields from different ADT definitions to collide - we do *not* allow field names to collide with other names For instance, we don't allow naming a field "let". We enforce this by not checking collision between ids for which we permit collisions (ex.: between fields), but still checking collisions between those ids and the others (ex.: fields and keywords). *) if allow_collisions id then ( assert (not is_opaque); (* Check with the other ids *) names_map_check_collision id_to_string id name ctx.names_map; { ctx with unsafe_names_map = unsafe_names_map_add id name ctx.unsafe_names_map; }) else (* Remark: we don't check that there are no collisions with the unsafe ids. Importantly, we don't want some safe ids like keywords to clash with unsafe ids like fields names. For this, we leverage the fact that we register keywords *first*, then unsafe ids (meaning the clash will be detected with the check in the other branch of the if ... then ... else ..., and we do have to check for all possible collisions, which may be slightly too restrictive). TODO: this is a bit hacky, we might want to improve the way we detect clashes by being more precise. Overall, there is only an issue with field names which are allowed to clash with each other. *) let names_map = names_map_add id_to_string is_opaque id name ctx.names_map in { ctx with names_map } (** [with_opaque_pre]: if [true] and the definition is opaque, add the opaque prefix *) let ctx_get (with_opaque_pre : bool) (id : id) (ctx : extraction_ctx) : string = (* We do not use the same name map if we allow/disallow collisions *) if allow_collisions id then IdMap.find id ctx.unsafe_names_map.id_to_name else match IdMap.find_opt id ctx.names_map.id_to_name with | Some s -> let is_opaque = IdSet.mem id ctx.names_map.opaque_ids in if with_opaque_pre && is_opaque then ctx.fmt.opaque_pre () ^ s else s | None -> log#serror ("Could not find: " ^ id_to_string id ctx); raise Not_found let ctx_get_global (with_opaque_pre : bool) (id : A.GlobalDeclId.id) (ctx : extraction_ctx) : string = ctx_get with_opaque_pre (GlobalId id) ctx let ctx_get_function (with_opaque_pre : bool) (id : fun_id) (ctx : extraction_ctx) : string = ctx_get with_opaque_pre (FunId id) ctx let ctx_get_local_function (with_opaque_pre : bool) (id : A.FunDeclId.id) (lp : LoopId.id option) (rg : RegionGroupId.id option) (ctx : extraction_ctx) : string = ctx_get_function with_opaque_pre (FromLlbc (Regular id, lp, rg)) ctx let ctx_get_type (with_opaque_pre : bool) (id : type_id) (ctx : extraction_ctx) : string = assert (id <> Tuple); ctx_get with_opaque_pre (TypeId id) ctx let ctx_get_local_type (with_opaque_pre : bool) (id : TypeDeclId.id) (ctx : extraction_ctx) : string = ctx_get_type with_opaque_pre (AdtId id) ctx let ctx_get_assumed_type (id : assumed_ty) (ctx : extraction_ctx) : string = (* In practice, the assumed types are opaque. However, assumed types are never grouped in the opaque module, meaning we never need to prefix them: we thus consider them as non-opaque with regards to the names map. *) let is_opaque = false in ctx_get_type is_opaque (Assumed id) ctx let ctx_get_trait_self_clause (ctx : extraction_ctx) : string = let with_opaque_pre = false in ctx_get with_opaque_pre TraitSelfClauseId ctx let ctx_get_trait_decl (with_opaque_pre : bool) (id : trait_decl_id) (ctx : extraction_ctx) : string = ctx_get with_opaque_pre (TraitDeclId id) ctx let ctx_get_trait_impl (with_opaque_pre : bool) (id : trait_impl_id) (ctx : extraction_ctx) : string = ctx_get with_opaque_pre (TraitImplId id) ctx let ctx_get_trait_item (id : trait_decl_id) (item_name : string) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (TraitItemId (id, item_name)) ctx let ctx_get_trait_const (id : trait_decl_id) (item_name : string) (ctx : extraction_ctx) : string = ctx_get_trait_item id item_name ctx let ctx_get_trait_type (id : trait_decl_id) (item_name : string) (ctx : extraction_ctx) : string = ctx_get_trait_item id item_name ctx let ctx_get_trait_method (id : trait_decl_id) (item_name : string) (ctx : extraction_ctx) : string = ctx_get_trait_item id item_name ctx let ctx_get_trait_parent_clause (id : trait_decl_id) (clause : trait_clause_id) (ctx : extraction_ctx) : string = let with_opaque_pre = false in ctx_get with_opaque_pre (TraitParentClauseId (id, clause)) ctx let ctx_get_trait_item_clause (id : trait_decl_id) (item : string) (clause : trait_clause_id) (ctx : extraction_ctx) : string = let with_opaque_pre = false in ctx_get with_opaque_pre (TraitItemClauseId (id, item, clause)) ctx let ctx_get_var (id : VarId.id) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (VarId id) ctx let ctx_get_type_var (id : TypeVarId.id) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (TypeVarId id) ctx let ctx_get_const_generic_var (id : ConstGenericVarId.id) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (ConstGenericVarId id) ctx let ctx_get_local_trait_clause (id : TraitClauseId.id) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (LocalTraitClauseId id) ctx let ctx_get_field (type_id : type_id) (field_id : FieldId.id) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (FieldId (type_id, field_id)) ctx let ctx_get_struct (with_opaque_pre : bool) (def_id : type_id) (ctx : extraction_ctx) : string = ctx_get with_opaque_pre (StructId def_id) ctx let ctx_get_variant (def_id : type_id) (variant_id : VariantId.id) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (VariantId (def_id, variant_id)) ctx let ctx_get_decreases_proof (def_id : A.FunDeclId.id) (loop_id : LoopId.id option) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (DecreasesProofId (Regular def_id, loop_id)) ctx let ctx_get_termination_measure (def_id : A.FunDeclId.id) (loop_id : LoopId.id option) (ctx : extraction_ctx) : string = let is_opaque = false in ctx_get is_opaque (TerminationMeasureId (Regular def_id, loop_id)) ctx (** Generate a unique type variable name and add it to the context *) let ctx_add_type_var (basename : string) (id : TypeVarId.id) (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let name = ctx.fmt.type_var_basename ctx.names_map.names_set basename in let name = basename_to_unique ctx.names_map.names_set ctx.fmt.append_index name in let ctx = ctx_add is_opaque (TypeVarId id) name ctx in (ctx, name) (** Generate a unique const generic variable name and add it to the context *) let ctx_add_const_generic_var (basename : string) (id : ConstGenericVarId.id) (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let name = ctx.fmt.const_generic_var_basename ctx.names_map.names_set basename in let name = basename_to_unique ctx.names_map.names_set ctx.fmt.append_index name in let ctx = ctx_add is_opaque (ConstGenericVarId id) name ctx in (ctx, name) (** See {!ctx_add_type_var} *) let ctx_add_type_vars (vars : (string * TypeVarId.id) list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (name, id) -> ctx_add_type_var name id ctx) ctx vars (** Generate a unique variable name and add it to the context *) let ctx_add_var (basename : string) (id : VarId.id) (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let name = basename_to_unique ctx.names_map.names_set ctx.fmt.append_index basename in let ctx = ctx_add is_opaque (VarId id) name ctx in (ctx, name) (** Generate a unique variable name for the trait self clause and add it to the context *) let ctx_add_trait_self_clause (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let basename = ctx.fmt.trait_self_clause_basename in let name = basename_to_unique ctx.names_map.names_set ctx.fmt.append_index basename in let ctx = ctx_add is_opaque TraitSelfClauseId name ctx in (ctx, name) (** Generate a unique trait clause name and add it to the context *) let ctx_add_local_trait_clause (basename : string) (id : TraitClauseId.id) (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let name = basename_to_unique ctx.names_map.names_set ctx.fmt.append_index basename in let ctx = ctx_add is_opaque (LocalTraitClauseId id) name ctx in (ctx, name) (** See {!ctx_add_var} *) let ctx_add_vars (vars : var list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (v : var) -> let name = ctx.fmt.var_basename ctx.names_map.names_set v.basename v.ty in ctx_add_var name v.id ctx) ctx vars let ctx_add_type_params (vars : type_var list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (var : type_var) -> ctx_add_type_var var.name var.index ctx) ctx vars let ctx_add_const_generic_params (vars : const_generic_var list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (var : const_generic_var) -> ctx_add_const_generic_var var.name var.index ctx) ctx vars let ctx_add_local_trait_clauses (clauses : trait_clause list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (c : trait_clause) -> let basename = ctx.fmt.trait_clause_basename ctx.names_map.names_set c in ctx_add_local_trait_clause basename c.clause_id ctx) ctx clauses (** Returns the lists of names for: - the type variables - the const generic variables - the trait clauses *) let ctx_add_generic_params (generics : generic_params) (ctx : extraction_ctx) : extraction_ctx * string list * string list * string list = let { types; const_generics; trait_clauses } = generics in let ctx, tys = ctx_add_type_params types ctx in let ctx, cgs = ctx_add_const_generic_params const_generics ctx in let ctx, tcs = ctx_add_local_trait_clauses trait_clauses ctx in (ctx, tys, cgs, tcs) let ctx_add_type_decl_struct (def : type_decl) (ctx : extraction_ctx) : extraction_ctx * string = assert (match def.kind with Struct _ -> true | _ -> false); let is_opaque = false in let cons_name = ctx.fmt.struct_constructor def.name in let ctx = ctx_add is_opaque (StructId (AdtId def.def_id)) cons_name ctx in (ctx, cons_name) let ctx_add_type_decl (def : type_decl) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = def.kind = Opaque in let def_name = ctx.fmt.type_name def.name in let ctx = ctx_add is_opaque (TypeId (AdtId def.def_id)) def_name ctx in ctx let ctx_add_field (def : type_decl) (field_id : FieldId.id) (field : field) (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let name = ctx.fmt.field_name def.name field_id field.field_name in let ctx = ctx_add is_opaque (FieldId (AdtId def.def_id, field_id)) name ctx in (ctx, name) let ctx_add_fields (def : type_decl) (fields : (FieldId.id * field) list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (vid, v) -> ctx_add_field def vid v ctx) ctx fields let ctx_add_variant (def : type_decl) (variant_id : VariantId.id) (variant : variant) (ctx : extraction_ctx) : extraction_ctx * string = let is_opaque = false in let name = ctx.fmt.variant_name def.name variant.variant_name in (* Add the type name prefix for Lean *) let name = if !Config.backend = Lean then let type_name = ctx.fmt.type_name def.name in type_name ^ "." ^ name else name in let ctx = ctx_add is_opaque (VariantId (AdtId def.def_id, variant_id)) name ctx in (ctx, name) let ctx_add_variants (def : type_decl) (variants : (VariantId.id * variant) list) (ctx : extraction_ctx) : extraction_ctx * string list = List.fold_left_map (fun ctx (vid, v) -> ctx_add_variant def vid v ctx) ctx variants let ctx_add_struct (def : type_decl) (ctx : extraction_ctx) : extraction_ctx * string = assert (match def.kind with Struct _ -> true | _ -> false); let is_opaque = false in let name = ctx.fmt.struct_constructor def.name in let ctx = ctx_add is_opaque (StructId (AdtId def.def_id)) name ctx in (ctx, name) let ctx_add_decreases_proof (def : fun_decl) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.decreases_proof_name def.def_id def.basename def.num_loops def.loop_id in ctx_add is_opaque (DecreasesProofId (Regular def.def_id, def.loop_id)) name ctx let ctx_add_termination_measure (def : fun_decl) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.termination_measure_name def.def_id def.basename def.num_loops def.loop_id in ctx_add is_opaque (TerminationMeasureId (Regular def.def_id, def.loop_id)) name ctx let ctx_add_global_decl_and_body (def : A.global_decl) (ctx : extraction_ctx) : extraction_ctx = (* TODO: update once the body id can be an option *) let is_opaque = false in let name = ctx.fmt.global_name def.name in let decl = GlobalId def.def_id in let body = FunId (FromLlbc (Regular def.body_id, None, None)) in let ctx = ctx_add is_opaque decl (name ^ "_c") ctx in let ctx = ctx_add is_opaque body (name ^ "_body") ctx in ctx let ctx_add_fun_decl (trans_group : pure_fun_translation) (def : fun_decl) (ctx : extraction_ctx) : extraction_ctx = (* Sanity check: the function should not be a global body - those are handled * separately *) assert (not def.is_global_decl_body); (* Lookup the LLBC def to compute the region group information *) let def_id = def.def_id in let llbc_def = A.FunDeclId.Map.find def_id ctx.trans_ctx.fun_context.fun_decls in let sg = llbc_def.signature in let num_rgs = List.length sg.regions_hierarchy in let { keep_fwd; fwd = _; backs } = trans_group in let num_backs = List.length backs in let rg_info = match def.back_id with | None -> None | Some rg_id -> let rg = T.RegionGroupId.nth sg.regions_hierarchy rg_id in let region_names = List.map (fun rid -> (T.RegionVarId.nth sg.generics.regions rid).name) rg.regions in Some { id = rg_id; region_names } in let is_opaque = def.body = None in (* Add the function name *) let def_name = ctx.fmt.fun_name def.basename def.num_loops def.loop_id num_rgs rg_info (keep_fwd, num_backs) in let fun_id = (A.Regular def_id, def.loop_id, def.back_id) in let ctx = ctx_add is_opaque (FunId (FromLlbc fun_id)) def_name ctx in (* Add the name info *) { ctx with fun_name_info = PureUtils.RegularFunIdMap.add fun_id { keep_fwd; num_backs } ctx.fun_name_info; } let ctx_add_trait_decl (d : trait_decl) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_decl_name d in ctx_add is_opaque (TraitDeclId d.def_id) name ctx let ctx_add_trait_impl (d : trait_impl) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_impl_name d in ctx_add is_opaque (TraitImplId d.def_id) name ctx let ctx_add_trait_const (d : trait_decl) (item : string) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_const_name d item in ctx_add is_opaque (TraitItemId (d.def_id, item)) name ctx let ctx_add_trait_type (d : trait_decl) (item : string) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_type_name d item in ctx_add is_opaque (TraitItemId (d.def_id, item)) name ctx let ctx_add_trait_method (d : trait_decl) (item : string) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_method_name d item in ctx_add is_opaque (TraitItemId (d.def_id, item)) name ctx let ctx_add_trait_parent_clause (d : trait_decl) (clause : trait_clause) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_parent_clause_name d clause in ctx_add is_opaque (TraitParentClauseId (d.def_id, clause.clause_id)) name ctx let ctx_add_trait_type_clause (d : trait_decl) (item : string) (clause : trait_clause) (ctx : extraction_ctx) : extraction_ctx = let is_opaque = false in let name = ctx.fmt.trait_type_clause_name d item clause in ctx_add is_opaque (TraitItemClauseId (d.def_id, item, clause.clause_id)) name ctx type names_map_init = { keywords : string list; assumed_adts : (assumed_ty * string) list; assumed_structs : (assumed_ty * string) list; assumed_variants : (assumed_ty * VariantId.id * string) list; assumed_llbc_functions : (A.assumed_fun_id * RegionGroupId.id option * string) list; assumed_pure_functions : (pure_assumed_fun_id * string) list; } (** Initialize a names map with a proper set of keywords/names coming from the target language/prover. *) let initialize_names_map (fmt : formatter) (init : names_map_init) : names_map = let int_names = List.map fmt.int_name T.all_int_types in let keywords = List.concat [ [ fmt.bool_name; fmt.char_name; fmt.str_name ]; int_names; init.keywords; ] in let names_set = StringSet.of_list keywords in let name_to_id = StringMap.of_list (List.map (fun x -> (x, UnknownId)) keywords) in let opaque_ids = IdSet.empty in (* We fist initialize [id_to_name] as empty, because the id of a keyword is [UnknownId]. * Also note that we don't need this mapping for keywords: we insert keywords only * to check collisions. *) let id_to_name = IdMap.empty in let nm = { id_to_name; name_to_id; names_set; opaque_ids } in (* For debugging - we are creating bindings for assumed types and functions, so * it is ok if we simply use the "show" function (those aren't simply identified * by numbers) *) let id_to_string = show_id in (* Then we add: * - the assumed types * - the assumed struct constructors * - the assumed variants * - the assumed functions *) let nm = List.fold_left (fun nm (type_id, name) -> names_map_add_assumed_type id_to_string type_id name nm) nm init.assumed_adts in let nm = List.fold_left (fun nm (type_id, name) -> names_map_add_assumed_struct id_to_string type_id name nm) nm init.assumed_structs in let nm = List.fold_left (fun nm (type_id, variant_id, name) -> names_map_add_assumed_variant id_to_string type_id variant_id name nm) nm init.assumed_variants in let assumed_functions = List.map (fun (fid, rg, name) -> (FromLlbc (A.Assumed fid, None, rg), name)) init.assumed_llbc_functions @ List.map (fun (fid, name) -> (Pure fid, name)) init.assumed_pure_functions in let nm = (* In practice, the assumed function are opaque. However, assumed functions are never grouped in the opaque module, meaning we never need to prefix them: we thus consider them as non-opaque with regards to the names map. *) let is_opaque = false in List.fold_left (fun nm (fid, name) -> names_map_add_function id_to_string is_opaque fid name nm) nm assumed_functions in (* Return *) nm let compute_type_decl_name (fmt : formatter) (def : type_decl) : string = fmt.type_name def.name (** Helper function: generate a suffix for a function name, i.e., generates a suffix like "_loop", "loop1", etc. to append to a function name. *) let default_fun_loop_suffix (num_loops : int) (loop_id : LoopId.id option) : string = match loop_id with | None -> "" | Some loop_id -> (* If this is for a loop, generally speaking, we append the loop index. If this function admits only one loop, we omit it. *) if num_loops = 1 then "_loop" else "_loop" ^ LoopId.to_string loop_id (** A helper function: generates a function suffix from a region group information. TODO: move all those helpers. *) let default_fun_suffix (num_loops : int) (loop_id : LoopId.id option) (num_region_groups : int) (rg : region_group_info option) ((keep_fwd, num_backs) : bool * int) : string = let lp_suff = default_fun_loop_suffix num_loops loop_id in (* There are several cases: - [rg] is [Some]: this is a forward function: - we add "_fwd" - [rg] is [None]: this is a backward function: - this function has one extracted backward function: - if the forward function has been filtered, we add "_fwd_back": the forward function is useless, so the unique backward function takes its place, in a way - otherwise we add "_back" - this function has several backward functions: we add "_back" and an additional suffix to identify the precise backward function Note that we always add a suffix (in case there are no region groups, we could not add the "_fwd" suffix) to prevent name clashes between definitions (in particular between type and function definitions). *) let rg_suff = (* TODO: make all the backends match what is done for Lean *) match rg with | None -> ( match !Config.backend with | FStar | Coq | HOL4 -> "_fwd" | Lean -> (* In order to avoid name conflicts: * - if the forward is eliminated, we add the suffix "_fwd" (it won't be used) * - otherwise, no suffix (because the backward functions will have a suffix) *) if num_backs = 1 && not keep_fwd then "_fwd" else "") | Some rg -> assert (num_region_groups > 0 && num_backs > 0); if num_backs = 1 then (* Exactly one backward function *) match !Config.backend with | FStar | Coq | HOL4 -> if not keep_fwd then "_fwd_back" else "_back" | Lean -> if not keep_fwd then "" else "_back" else if (* Several region groups/backward functions: - if all the regions in the group have names, we use those names - otherwise we use an index *) List.for_all Option.is_some rg.region_names then (* Concatenate the region names *) "_back" ^ String.concat "" (List.map Option.get rg.region_names) else (* Use the region index *) "_back" ^ RegionGroupId.to_string rg.id in lp_suff ^ rg_suff