open Identifiers open Types (* TODO: I often write "abstract" (value, borrow content, etc.) while I should * write "abstraction" (because those values are not abstract, they simply are * inside abstractions) *) module VarId = IdGen () module BorrowId = IdGen () module SymbolicValueId = IdGen () module AbstractionId = IdGen () module FunCallId = IdGen () (** A variable *) type big_int = Z.t let big_int_of_yojson (json : Yojson.Safe.t) : (big_int, string) result = match json with | `Int i -> Ok (Z.of_int i) | `Intlit is -> Ok (Z.of_string is) | _ -> Error "not an integer or an integer literal" let big_int_to_yojson (i : big_int) = `Intlit (Z.to_string i) let pp_big_int (fmt : Format.formatter) (bi : big_int) : unit = Format.pp_print_string fmt (Z.to_string bi) let show_big_int (bi : big_int) : string = Z.to_string bi type scalar_value = { value : big_int; int_ty : integer_type } [@@deriving show] (** A scalar value Note that we use unbounded integers everywhere. We then harcode the boundaries for the different types. *) (** A constant value *) type constant_value = | Scalar of scalar_value | Bool of bool | Char of char | String of string [@@deriving show] (** The kind of a symbolic value, which precises how the value was generated *) type sv_kind = | FunCallRet (** The value is the return value of a function call *) | FunCallGivenBack (** The value is a borrowed value given back by an abstraction (happens when giving a borrow to a function: when the abstraction introduced to model the function call ends we reintroduce a symbolic value in the context for the value modified by the abstraction through the borrow). *) | SynthInput (** The value is an input value of the function whose body we are currently synthesizing. *) | SynthRetGivenBack (** The value is a borrowed value that the function whose body we are synthesizing returned, and which was given back because we ended one of the lifetimes of this function (we do this to synthesize the backward functions). *) | SynthInputGivenBack (** The value was given back upon ending one of the input abstractions *) [@@deriving show] type symbolic_value = { sv_kind : sv_kind; sv_id : SymbolicValueId.id; sv_ty : rty; } [@@deriving show] (** A symbolic value *) (** Ancestor for [typed_value] iter visitor *) class ['self] iter_typed_value_base = object (_self : 'self) inherit [_] VisitorsRuntime.iter method visit_constant_value : 'env -> constant_value -> unit = fun _ _ -> () method visit_erased_region : 'env -> erased_region -> unit = fun _ _ -> () method visit_symbolic_value : 'env -> symbolic_value -> unit = fun _ _ -> () method visit_ety : 'env -> ety -> unit = fun _ _ -> () end (** Ancestor for [typed_value] map visitor for *) class ['self] map_typed_value_base = object (_self : 'self) inherit [_] VisitorsRuntime.map method visit_constant_value : 'env -> constant_value -> constant_value = fun _ cv -> cv method visit_erased_region : 'env -> erased_region -> erased_region = fun _ r -> r method visit_symbolic_value : 'env -> symbolic_value -> symbolic_value = fun _ sv -> sv method visit_ety : 'env -> ety -> ety = fun _ ty -> ty end (** An untyped value, used in the environments *) type value = | Concrete of constant_value (** Concrete (non-symbolic) value *) | Adt of adt_value (** Enumerations and structures *) | Bottom (** No value (uninitialized or moved value) *) | Borrow of borrow_content (** A borrowed value *) | Loan of loan_content (** A loaned value *) | Symbolic of symbolic_value (** Borrow projector over a symbolic value. Note that contrary to the abstraction-values case, symbolic values appearing in regular values are interpreted as *borrow* projectors, they can never be *loan* projectors. *) and adt_value = { variant_id : (VariantId.id option[@opaque]); field_values : typed_value list; } and borrow_content = | SharedBorrow of (BorrowId.id[@opaque]) (** A shared value *) | MutBorrow of (BorrowId.id[@opaque]) * typed_value (** A mutably borrowed value. *) | InactivatedMutBorrow of (BorrowId.id[@opaque]) (** An inactivated mut borrow. This is used to model [two-phase borrows](https://rustc-dev-guide.rust-lang.org/borrow_check/two_phase_borrows.html). When evaluating a two-phase mutable borrow, we first introduce an inactivated borrow which behaves like a shared borrow, until the moment we actually *use* the borrow: at this point, we end all the other shared borrows (or inactivated borrows - though there shouldn't be any other inactivated borrows if the program is well typed) of this value and replace the inactivated borrow with a mutable borrow. A simple use case of two-phase borrows: ``` let mut v = Vec::new(); v.push(v.len()); ``` This gets desugared to (something similar to) the following MIR: ``` v = Vec::new(); v1 = &mut v; v2 = &v; // We need this borrow, but v has already been mutably borrowed! l = Vec::len(move v2); Vec::push(move v1, move l); // In practice, v1 gets activated only here ``` *) and loan_content = | SharedLoan of (BorrowId.Set.t[@opaque]) * typed_value | MutLoan of (BorrowId.id[@opaque]) (** TODO: we might want to add a set of borrow ids (useful for inactivated borrows, and extremely useful when giving shared values to abstractions). *) and typed_value = { value : value; ty : ety } [@@deriving show, visitors { name = "iter_typed_value"; variety = "iter"; ancestors = [ "iter_typed_value_base" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }, visitors { name = "map_typed_value"; variety = "map"; ancestors = [ "map_typed_value_base" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }] (** "Regular" typed value (we map variables to typed values) *) type mvalue = typed_value [@@deriving show] (** "Meta"-value: information we store for the synthesis. Note that we never automatically visit the meta-values with the visitors: they really are meta information, and shouldn't be considered as part of the environment during a symbolic execution. *) type msymbolic_value = symbolic_value [@@deriving show] (** "Meta"-symbolic value. See the explanations for [mvalue] *) (** When giving shared borrows to functions (i.e., inserting shared borrows inside abstractions) we need to reborrow the shared values. When doing so, we lookup the shared values and apply some special projections to the shared value (until we can't go further, i.e., we find symbolic values which may get expanded upon reading them later), which don't generate avalues but sets of borrow ids and symbolic values. Note that as shared values can't get modified it is ok to forget the structure of the values we projected, and only keep the set of borrows (and symbolic values). TODO: we may actually need to remember the structure, in order to know which borrows are inside which other borrows... *) type abstract_shared_borrow = | AsbBorrow of (BorrowId.id[@opaque]) | AsbProjReborrows of (symbolic_value[@opaque]) * (rty[@opaque]) [@@deriving show] type abstract_shared_borrows = abstract_shared_borrow list [@@deriving show] (** A set of abstract shared borrows *) (** Ancestor for [aproj] iter visitor *) class ['self] iter_aproj_base = object (_self : 'self) inherit [_] iter_typed_value method visit_rty : 'env -> rty -> unit = fun _ _ -> () method visit_mvalue : 'env -> mvalue -> unit = fun _ _ -> () method visit_msymbolic_value : 'env -> msymbolic_value -> unit = fun _ _ -> () end (** Ancestor for [aproj] map visitor *) class ['self] map_aproj_base = object (_self : 'self) inherit [_] map_typed_value method visit_rty : 'env -> rty -> rty = fun _ ty -> ty method visit_mvalue : 'env -> mvalue -> mvalue = fun _ v -> v method visit_msymbolic_value : 'env -> msymbolic_value -> msymbolic_value = fun _ m -> m end type aproj = | AProjLoans of symbolic_value * (mvalue * aproj) list (** A projector of loans over a symbolic value. Note that the borrows of a symbolic value may be spread between different abstractions, meaning that the projector of loans might receive *several* (symbolic) given back values. This is the case in the following example: ``` fn f<'a> (...) -> (&'a mut u32, &'a mut u32); fn g<'b, 'c>(p : (&'b mut u32, &'c mut u32)); let p = f(...); g(move p); // Symbolic context after the call to g: // abs'a {'a} { [s@0 <: (&'a mut u32, &'a mut u32)] } // // abs'b {'b} { (s@0 <: (&'b mut u32, &'c mut u32)) } // abs'c {'c} { (s@0 <: (&'b mut u32, &'c mut u32)) } ``` Upon evaluating the call to `f`, we introduce a symbolic value `s@0` and a projector of loans (projector loans from the region 'c). This projector will later receive two given back values: one for 'a and one for 'b. We accumulate those values in the list of projections (note that the meta value stores the value which was given back). We can later end the projector of loans if `s@0` is not referenced anywhere in the context below a projector of borrows which intersects this projector of loans. *) | AProjBorrows of symbolic_value * rty (** Note that an AProjBorrows only operates on a value which is not below a shared loan: under a shared loan, we use [abstract_shared_borrow]. Also note that once given to a borrow projection, a symbolic value can't get updated/expanded: this means that we don't need to save any meta-value here. *) | AEndedProjLoans of msymbolic_value * (mvalue * aproj) list (** An ended projector of loans over a symbolic value. See the explanations for [AProjLoans] Note that we keep the original symbolic value as a meta-value. *) | AEndedProjBorrows of mvalue (** The only purpose of [AEndedProjBorrows] is to store, for synthesis purposes, the symbolic value which was generated and given back upon ending the borrow. *) | AIgnoredProjBorrows [@@deriving show, visitors { name = "iter_aproj"; variety = "iter"; ancestors = [ "iter_aproj_base" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }, visitors { name = "map_aproj"; variety = "map"; ancestors = [ "map_aproj_base" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }] type region = RegionVarId.id Types.region [@@deriving show] (** Ancestor for [typed_avalue] iter visitor *) class ['self] iter_typed_avalue_base = object (_self : 'self) inherit [_] iter_aproj method visit_id : 'env -> BorrowId.id -> unit = fun _ _ -> () method visit_region : 'env -> region -> unit = fun _ _ -> () method visit_abstract_shared_borrows : 'env -> abstract_shared_borrows -> unit = fun _ _ -> () end (** Ancestor for [typed_avalue] map visitor *) class ['self] map_typed_avalue_base = object (_self : 'self) inherit [_] map_aproj method visit_id : 'env -> BorrowId.id -> BorrowId.id = fun _ id -> id method visit_region : 'env -> region -> region = fun _ r -> r method visit_abstract_shared_borrows : 'env -> abstract_shared_borrows -> abstract_shared_borrows = fun _ asb -> asb end (** Abstraction values are used inside of abstractions to properly model borrowing relations introduced by function calls. When calling a function, we lose information about the borrow graph: part of it is thus "abstracted" away. *) type avalue = | AConcrete of constant_value (** TODO: remove Note that this case is not used in the projections to keep track of the borrow graph (because there are no borrows in "concrete" values!) but to correctly instantiate the backward functions (we may give back some values at different moments: we need to remember what those values were precisely). Also note that even though avalues and values are not the same, once values are projected to avalues, those avalues still have the structure of the original values (this is necessary, again, to correctly instantiate the backward functions) *) | AAdt of adt_avalue | ABottom | ALoan of aloan_content | ABorrow of aborrow_content | ASymbolic of aproj | AIgnored (** A value which doesn't contain borrows, or which borrows we don't own and thus ignore *) and adt_avalue = { variant_id : (VariantId.id option[@opaque]); field_values : typed_avalue list; } (** A loan content as stored in an abstraction. Note that the children avalues are independent of the parent avalues. For instance, the child avalue contained in an [AMutLoan] will likely contain other, independent loans. Keeping track of the hierarchy is not necessary to maintain the borrow graph (which is the primary role of the abstractions), but it is necessary to properly instantiate the backward functions when generating the pure translation. *) and aloan_content = | AMutLoan of (BorrowId.id[@opaque]) * typed_avalue (** A mutable loan owned by an abstraction. Example: ======== ``` fn f<'a>(...) -> &'a mut &'a mut u32; let px = f(...); ``` We get (after some symbolic exansion): ``` abs0 { a_mut_loan l0 (a_mut_loan l1) } px -> mut_borrow l0 (mut_borrow @s1) ``` *) | ASharedLoan of (BorrowId.Set.t[@opaque]) * typed_value * typed_avalue (** A shared loan owned by an abstraction. Example: ======== ``` fn f<'a>(...) -> &'a u32; let px = f(...); ``` We get: ``` abs0 { a_shared_loan {l0} @s0 ⊥ } px -> shared_loan l0 ``` *) | AEndedMutLoan of { child : typed_avalue; given_back : typed_avalue; given_back_meta : mvalue; } (** An ended mutable loan in an abstraction. We need it because abstractions must keep track of the values we gave back to them, so that we can correctly instantiate backward functions. Rk.: *DO NOT* use [visit_AEndedMutLoan]. If we update the order of the arguments and you forget to swap them at the level of [visit_AEndedMutLoan], you will not notice it. Example: ======== ``` abs0 { a_mut_loan l0 ⊥ } x -> mut_borrow l0 (U32 3) ``` After ending `l0`: ``` abs0 { a_ended_mut_loan { given_back = U32 3; child = ⊥; } x -> ⊥ ``` *) | AEndedSharedLoan of typed_value * typed_avalue (** Similar to [AEndedMutLoan] but in this case there are no avalues to give back. We keep the shared value because it now behaves as a "regular" value (which contains borrows we might want to end...). *) | AIgnoredMutLoan of (BorrowId.id[@opaque]) * typed_avalue (** An ignored mutable loan. We need to keep track of ignored mutable loans, because we may have to apply projections on the values given back to those loans (say you have a borrow of type `&'a mut &'b mut`, in the abstraction 'b, the outer loan is ignored, however you need to keep track of it so that when ending the borrow corresponding to 'a you can correctly project on the inner value). Example: ======== ``` fn f<'a,'b>(...) -> &'a mut &'b mut u32; let x = f(...); > abs'a { a_mut_loan l0 (a_ignored_mut_loan l1 ⊥) } > abs'b { a_ignored_mut_loan l0 (a_mut_loan l1 ⊥) } > x -> mut_borrow l0 (mut_borrow l1 @s1) ``` *) | AEndedIgnoredMutLoan of { child : typed_avalue; given_back : typed_avalue; given_back_meta : mvalue; } (** Similar to [AEndedMutLoan], for ignored loans. Rk.: *DO NOT* use [visit_AEndedIgnoredMutLoan]. See the comment for [AEndedMutLoan]. *) | AIgnoredSharedLoan of typed_avalue (** An ignored shared loan. Example: ======== ``` fn f<'a,'b>(...) -> &'a &'b u32; let x = f(...); > abs'a { a_shared_loan {l0} (shared_borrow l1) (a_ignored_shared_loan ⊥) } > abs'b { a_ignored_shared_loan (a_shared_loan {l1} @s1 ⊥) } > x -> shared_borrow l0 ``` *) (** Note that when a borrow content is ended, it is replaced by ⊥ (while we need to track ended loans more precisely, especially because of their children values). Note that contrary to [aloan_content], here the children avalues are not independent of the parent avalues. For instance, a value `AMutBorrow (_, AMutBorrow (_, ...)` (ignoring the types) really is to be seen like a `mut_borrow ... (mut_borrow ...)`. TODO: be more precise about the ignored borrows (keep track of the borrow ids)? *) and aborrow_content = | AMutBorrow of mvalue * (BorrowId.id[@opaque]) * typed_avalue (** A mutable borrow owned by an abstraction. Is used when an abstraction "consumes" borrows, when giving borrows as arguments to a function. Example: ======== ``` fn f<'a>(px : &'a mut u32); > x -> mut_loan l0 > px -> mut_borrow l0 (U32 0) f(move px); > x -> mut_loan l0 > px -> ⊥ > abs0 { a_mut_borrow l0 (U32 0) } ``` The meta-value stores the initial value on which the projector was applied, which reduced to this mut borrow. This meta-information is only used for the synthesis. TODO: do we really use it actually? *) | ASharedBorrow of (BorrowId.id[@opaque]) (** A shared borrow owned by an abstraction. Example: ======== ``` fn f<'a>(px : &'a u32); > x -> shared_loan {l0} (U32 0) > px -> shared_borrow l0 f(move px); > x -> shared_loan {l0} (U32 0) > px -> ⊥ > abs0 { a_shared_borrow l0 } ``` *) | AIgnoredMutBorrow of BorrowId.id option * typed_avalue (** An ignored mutable borrow. We need to keep track of ignored mut borrows because when ending such borrows, we need to project the loans of the given back value to insert them in the proper abstractions. Note that we need to do so only for borrows consumed by parent abstractions (hence the optional borrow id). Example: ======== ``` fn f<'a,'b>(ppx : &'a mut &'b mut u32); > x -> mut_loan l0 > px -> mut_loan l1 > ppx -> mut_borrow l1 (mut_borrow l0 (U32 0)) f(move ppx); > x -> mut_loan l0 > px -> mut_loan l1 > ppx -> ⊥ > abs'a { a_mut_borrow l1 (a_ignored_mut_borrow None (U32 0)) } // TODO: duplication > abs'b {parents={abs'a}} { a_ignored_mut_borrow (Some l1) (a_mut_borrow l0 (U32 0)) } ... // abs'a ends > x -> mut_loan l0 > px -> @s0 > ppx -> ⊥ > abs'b { > a_ended_ignored_mut_borrow (a_proj_loans (@s0 <: &'b mut u32)) // <-- loan projector > (a_mut_borrow l0 (U32 0)) > } ... // `@s0` gets expanded to `&mut l2 @s1` > x -> mut_loan l0 > px -> &mut l2 @s1 > ppx -> ⊥ > abs'b { > a_ended_ignored_mut_borrow (a_mut_loan l2) // <-- loan l2 is here > (a_mut_borrow l0 (U32 0)) > } ``` Note that we could use AIgnoredMutLoan in the case the borrow id is not None, which would allow us to simplify the rules (to not have rules to specifically handle the case of AIgnoredMutBorrow with Some borrow id) and also remove the AEndedIgnoredMutBorrow variant. For now, the rules are implemented and it allows us to make the avalues more precise and clearer, so we will keep it that way. TODO: this is annoying, we are duplicating information. Maybe we could introduce an "Ignored" value? We have to pay attention to two things: - introducing ⊥ when ignoring a value is not always possible, because we check whether the borrowed value contains ⊥ when giving back a borrowed value (if it is the case we give back ⊥, otherwise we introduce a symbolic value). This is necessary when ending nested borrows with the same lifetime: when ending the inner borrow we actually give back a value, however when ending the outer borrow we need to give back ⊥. TODO: actually we don't do that anymore, we check if the borrowed avalue contains ended regions (which is cleaner and more robust). - we may need to remember the precise values given to the abstraction so that we can properly call the backward functions when generating the pure translation. *) | AEndedMutBorrow of mvalue * typed_avalue (** The sole purpose of [AEndedMutBorrow] is to store the (symbolic) value that we gave back as a meta-value, to help with the synthesis. We also remember the child [avalue] because this structural information is useful for the synthesis (but not for the symbolic execution): in practice the child value should only contain ended borrows, ignored values, bottom values, etc. *) | AEndedIgnoredMutBorrow of { child : typed_avalue; given_back_loans_proj : typed_avalue; given_back_meta : mvalue; (** [given_back_meta] is used to store the (symbolic) value we gave back upon ending the borrow. Rk.: *DO NOT* use [visit_AEndedIgnoredMutLoan]. See the comment for [AEndedMutLoan]. *) } (** See the explanations for [AIgnoredMutBorrow] *) | AProjSharedBorrow of abstract_shared_borrows (** A projected shared borrow. When giving shared borrows as arguments to function calls, we introduce new borrows to keep track of the fact that the function might reborrow values inside. Note that as shared values are immutable, we don't really need to remember the structure of the shared values. Example: ======== Below, when calling `f`, we need to introduce one shared borrow per borrow in the argument. ``` fn f<'a,'b>(pppx : &'a &'b &'c mut u32); > x -> mut_loan l0 > px -> shared_loan {l1} (mut_borrow l0 (U32 0)) > ppx -> shared_loan {l2} (shared_borrow l1) > pppx -> shared_borrow l2 f(move pppx); > x -> mut_loan l0 > px -> shared_loan {l1, l3, l4} (mut_borrow l0 (U32 0)) > ppx -> shared_loan {l2} (shared_borrow l1) > pppx -> ⊥ > abs'a { a_proj_shared_borrow {l2} } > abs'b { a_proj_shared_borrow {l3} } // l3 reborrows l1 > abs'c { a_proj_shared_borrow {l4} } // l4 reborrows l0 ``` *) (* TODO: the type of avalues doesn't make sense for loan avalues: they currently are typed as `& (mut) T` instead of `T`... *) and typed_avalue = { value : avalue; ty : rty } [@@deriving show, visitors { name = "iter_typed_avalue"; variety = "iter"; ancestors = [ "iter_typed_avalue_base" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }, visitors { name = "map_typed_avalue"; variety = "map"; ancestors = [ "map_typed_avalue_base" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }] (** The kind of an abstraction, which keeps track of its origin *) type abs_kind = | FunCall (** The abstraction was introduced because of a function call *) | SynthInput (** The abstraction keeps track of the input values of the function we are currently synthesizing. *) | SynthRet (** The abstraction "absorbed" the value returned by the function we are currently synthesizing *) [@@deriving show] type abs = { abs_id : (AbstractionId.id[@opaque]); call_id : (FunCallId.id[@opaque]); (** The identifier of the function call which introduced this abstraction. This is not used by the symbolic execution: this is only used for pretty-printing and debugging, in the symbolic AST, generated by the symbolic execution. *) back_id : (RegionGroupId.id[@opaque]); (** The region group id to which this abstraction is linked. In most situations, it gives the id of the backward function (hence the name), but it is a bit more subtle in the case of synth input and synth ret abstractions. This is not used by the symbolic execution: it is a utility for the symbolic AST, generated by the symbolic execution. *) kind : (abs_kind[@opaque]); parents : (AbstractionId.Set.t[@opaque]); (** The parent abstractions *) original_parents : (AbstractionId.id list[@opaque]); (** The original list of parents, ordered. This is used for synthesis. *) regions : (RegionId.Set.t[@opaque]); (** Regions owned by this abstraction *) ancestors_regions : (RegionId.Set.t[@opaque]); (** Union of the regions owned by this abstraction's ancestors (not including the regions of this abstraction itself) *) avalues : typed_avalue list; (** The values in this abstraction *) } [@@deriving show, visitors { name = "iter_abs"; variety = "iter"; ancestors = [ "iter_typed_avalue" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }, visitors { name = "map_abs"; variety = "map"; ancestors = [ "map_typed_avalue" ]; nude = true (* Don't inherit [VisitorsRuntime.iter] *); concrete = true; }] (** Abstractions model the parts in the borrow graph where the borrowing relations have been abstracted because of a function call. In order to model the relations between the borrows, we use "abstraction values", which are a special kind of value. *) (** A symbolic expansion A symbolic expansion doesn't represent a value, but rather an operation that we apply to values. TODO: this should rather be name "expanded_symbolic" *) type symbolic_expansion = | SeConcrete of constant_value | SeAdt of (VariantId.id option * symbolic_value list) | SeMutRef of BorrowId.id * symbolic_value | SeSharedRef of BorrowId.Set.t * symbolic_value