annotate src/MoreDep.v @ 85:3746a2ded8da

Tagless interpreter & cfold
author Adam Chlipala <adamc@hcoop.net>
date Mon, 06 Oct 2008 13:07:24 -0400
parents 522436ed6688
children fd505bcb5632
rev   line source
adamc@83 1 (* Copyright (c) 2008, Adam Chlipala
adamc@83 2 *
adamc@83 3 * This work is licensed under a
adamc@83 4 * Creative Commons Attribution-Noncommercial-No Derivative Works 3.0
adamc@83 5 * Unported License.
adamc@83 6 * The license text is available at:
adamc@83 7 * http://creativecommons.org/licenses/by-nc-nd/3.0/
adamc@83 8 *)
adamc@83 9
adamc@83 10 (* begin hide *)
adamc@85 11 Require Import Arith Bool List.
adamc@83 12
adamc@83 13 Require Import Tactics.
adamc@83 14
adamc@83 15 Set Implicit Arguments.
adamc@83 16 (* end hide *)
adamc@83 17
adamc@83 18
adamc@83 19 (** %\chapter{More Dependent Types}% *)
adamc@83 20
adamc@83 21 (** Subset types and their relatives help us integrate verification with programming. Though they reorganize the certified programmer's workflow, they tend not to have deep effects on proofs. We write largely the same proofs as we would for classical verification, with some of the structure moved into the programs themselves. It turns out that, when we use dependent types to their full potential, we warp the development and proving process even more than that, picking up "free theorems" to the extent that often a certified program is hardly more complex than its uncertified counterpart in Haskell or ML.
adamc@83 22
adamc@83 23 In particular, we have only scratched the tip of the iceberg that is Coq's inductive definition mechanism. The inductive types we have seen so far have their counterparts in the other proof assistants that we surveyed in Chapter 1. This chapter explores the strange new world of dependent inductive datatypes (that is, dependent inductive types outside [Prop]), a possibility which sets Coq apart from all of the competition not based on type theory. *)
adamc@83 24
adamc@84 25
adamc@84 26 (** * Length-Indexed Lists *)
adamc@84 27
adamc@84 28 (** Many introductions to dependent types start out by showing how to use them to eliminate array bounds checks. When the type of an array tells you how many elements it has, your compiler can detect out-of-bounds dereferences statically. Since we are working in a pure functional language, the next best thing is length-indexed lists, which the following code defines. *)
adamc@84 29
adamc@84 30 Section ilist.
adamc@84 31 Variable A : Set.
adamc@84 32
adamc@84 33 Inductive ilist : nat -> Set :=
adamc@84 34 | Nil : ilist O
adamc@84 35 | Cons : forall n, A -> ilist n -> ilist (S n).
adamc@84 36
adamc@84 37 (** We see that, within its section, [ilist] is given type [nat -> Set]. Previously, every inductive type we have seen has either had plain [Set] as its type or has been a predicate with some type ending in [Prop]. The full generality of inductive definitions lets us integrate the expressivity of predicates directly into our normal programming.
adamc@84 38
adamc@84 39 The [nat] argument to [ilist] tells us the length of the list. The types of [ilist]'s constructors tell us that a [Nil] list has length [O] and that a [Cons] list has length one greater than the length of its sublist. We may apply [ilist] to any natural number, even natural numbers that are only known at runtime. It is this breaking of the %\textit{%#<i>#phase distinction#</i>#%}% that characterizes [ilist] as %\textit{%#<i>#dependently typed#</i>#%}%.
adamc@84 40
adamc@84 41 In expositions of list types, we usually see the length function defined first, but here that would not be a very productive function to code. Instead, let us implement list concatenation.
adamc@84 42
adamc@84 43 [[
adamc@84 44 Fixpoint app n1 (ls1 : ilist n1) n2 (ls2 : ilist n2) {struct ls1} : ilist (n1 + n2) :=
adamc@84 45 match ls1 with
adamc@84 46 | Nil => ls2
adamc@84 47 | Cons _ x ls1' => Cons x (app ls1' ls2)
adamc@84 48 end.
adamc@84 49
adamc@84 50 Coq is not happy with this definition:
adamc@84 51
adamc@84 52 [[
adamc@84 53 The term "ls2" has type "ilist n2" while it is expected to have type
adamc@84 54 "ilist (?14 + n2)"
adamc@84 55 ]]
adamc@84 56
adamc@84 57 We see the return of a problem we have considered before. Without explicit annotations, Coq does not enrich our typing assumptions in the branches of a [match] expression. It is clear that the unification variable [?14] should be resolved to 0 in this context, so that we have [0 + n2] reducing to [n2], but Coq does not realize that. We cannot fix the problem using just the simple [return] clauses we applied in the last chapter. We need to combine a [return] clause with a new kind of annotation, an [in] clause. *)
adamc@84 58
adamc@84 59 Fixpoint app n1 (ls1 : ilist n1) n2 (ls2 : ilist n2) {struct ls1} : ilist (n1 + n2) :=
adamc@84 60 match ls1 in (ilist n1) return (ilist (n1 + n2)) with
adamc@84 61 | Nil => ls2
adamc@84 62 | Cons _ x ls1' => Cons x (app ls1' ls2)
adamc@84 63 end.
adamc@84 64
adamc@84 65 (** This version of [app] passes the type checker. Using [return] alone allowed us to express a dependency of the [match] result type on the %\textit{%#<i>#value#</i>#%}% of the discriminee. What [in] adds to our arsenal is a way of expressing a dependency on the %\textit{%#<i>#type#</i>#%}% of the discriminee. Specifically, the [n1] in the [in] clause above is a %\textit{%#<i>#binding occurrence#</i>#%}% whose scope is the [return] clause.
adamc@84 66
adamc@84 67 We may use [in] clauses only to bind names for the arguments of an inductive type family. That is, each [in] clause must be an inductive type family name applied to a sequence of underscores and variable names of the proper length. The positions for %\textit{%#<i>#parameters#</i>#%}% to the type family must all be underscores. Parameters are those arguments declared with section variables or with entries to the left of the first colon in an inductive definition. They cannot vary depending on which constructor was used to build the discriminee, so Coq prohibits pointless matches on them. It is those arguments defined in the type to the right of the colon that we may name with [in] clauses.
adamc@84 68
adamc@84 69 Our [app] function could be typed in so-called %\textit{%#<i>#stratified#</i>#%}% type systems, which avoid true dependency. We could consider the length indices to lists to live in a separate, compile-time-only universe from the lists themselves. Our next example would be harder to implement in a stratified system. We write an injection function from regular lists to length-indexed lists. A stratified implementation would need to duplicate the definition of lists across compile-time and run-time versions, and the run-time versions would need to be indexed by the compile-time versions. *)
adamc@84 70
adamc@84 71 Fixpoint inject (ls : list A) : ilist (length ls) :=
adamc@84 72 match ls return (ilist (length ls)) with
adamc@84 73 | nil => Nil
adamc@84 74 | h :: t => Cons h (inject t)
adamc@84 75 end.
adamc@84 76
adamc@84 77 (** We can define an inverse conversion and prove that it really is an inverse. *)
adamc@84 78
adamc@84 79 Fixpoint unject n (ls : ilist n) {struct ls} : list A :=
adamc@84 80 match ls with
adamc@84 81 | Nil => nil
adamc@84 82 | Cons _ h t => h :: unject t
adamc@84 83 end.
adamc@84 84
adamc@84 85 Theorem inject_inverse : forall ls, unject (inject ls) = ls.
adamc@84 86 induction ls; crush.
adamc@84 87 Qed.
adamc@84 88
adamc@84 89 (** Now let us attempt a function that is surprisingly tricky to write. In ML, the list head function raises an exception when passed an empty list. With length-indexed lists, we can rule out such invalid calls statically, and here is a first attempt at doing so.
adamc@84 90
adamc@84 91 [[
adamc@84 92 Definition hd n (ls : ilist (S n)) : A :=
adamc@84 93 match ls with
adamc@84 94 | Nil => ???
adamc@84 95 | Cons _ h _ => h
adamc@84 96 end.
adamc@84 97
adamc@84 98 It is not clear what to write for the [Nil] case, so we are stuck before we even turn our function over to the type checker. We could try omitting the [Nil] case:
adamc@84 99
adamc@84 100 [[
adamc@84 101 Definition hd n (ls : ilist (S n)) : A :=
adamc@84 102 match ls with
adamc@84 103 | Cons _ h _ => h
adamc@84 104 end.
adamc@84 105
adamc@84 106 [[
adamc@84 107 Error: Non exhaustive pattern-matching: no clause found for pattern Nil
adamc@84 108 ]]
adamc@84 109
adamc@84 110 Unlike in ML, we cannot use inexhaustive pattern matching, becuase there is no conception of a %\texttt{%#<tt>#Match#</tt>#%}% exception to be thrown. We might try using an [in] clause somehow.
adamc@84 111
adamc@84 112 [[
adamc@84 113 Definition hd n (ls : ilist (S n)) : A :=
adamc@84 114 match ls in (ilist (S n)) with
adamc@84 115 | Cons _ h _ => h
adamc@84 116 end.
adamc@84 117
adamc@84 118 [[
adamc@84 119 Error: The reference n was not found in the current environment
adamc@84 120 ]]
adamc@84 121
adamc@84 122 In this and other cases, we feel like we want [in] clauses with type family arguments that are not variables. Unfortunately, Coq only supports variables in those positions. A completely general mechanism could only be supported with a solution to the problem of higher-order unification, which is undecidable. There %\textit{%#<i>#are#</i>#%}% useful heuristics for handling non-variable indices which are gradually making their way into Coq, but we will spend some time in this and the next few chapters on effective pattern matching on dependent types using only the primitive [match] annotations.
adamc@84 123
adamc@84 124 Our final, working attempt at [hd] uses an auxiliary function and a surprising [return] annotation. *)
adamc@84 125
adamc@84 126 Definition hd' n (ls : ilist n) :=
adamc@84 127 match ls in (ilist n) return (match n with O => unit | S _ => A end) with
adamc@84 128 | Nil => tt
adamc@84 129 | Cons _ h _ => h
adamc@84 130 end.
adamc@84 131
adamc@84 132 Definition hd n (ls : ilist (S n)) : A := hd' ls.
adamc@84 133
adamc@84 134 (** We annotate our main [match] with a type that is itself a [match]. We write that the function [hd'] returns [unit] when the list is empty and returns the carried type [A] in all other cases. In the definition of [hd], we just call [hd']. Because the index of [ls] is known to be nonzero, the type checker reduces the [match] in the type of [hd'] to [A]. *)
adamc@84 135
adamc@84 136 End ilist.
adamc@85 137
adamc@85 138
adamc@85 139 (** * A Tagless Interpreter *)
adamc@85 140
adamc@85 141 (** A favorite example for motivating the power of functional programming is implementation of a simple expression language interpreter. In ML and Haskell, such interpreters are often implemented using an algebraic datatype of values, where at many points it is checked that a value was built with the right constructor of the value type. With dependent types, we can implement a %\textit{%#<i>#tagless#</i>#%}% interpreter that both removes this source of runtime ineffiency and gives us more confidence that our implementation is correct. *)
adamc@85 142
adamc@85 143 Inductive type : Set :=
adamc@85 144 | Nat : type
adamc@85 145 | Bool : type
adamc@85 146 | Prod : type -> type -> type.
adamc@85 147
adamc@85 148 Inductive exp : type -> Set :=
adamc@85 149 | NConst : nat -> exp Nat
adamc@85 150 | Plus : exp Nat -> exp Nat -> exp Nat
adamc@85 151 | Eq : exp Nat -> exp Nat -> exp Bool
adamc@85 152
adamc@85 153 | BConst : bool -> exp Bool
adamc@85 154 | And : exp Bool -> exp Bool -> exp Bool
adamc@85 155 | If : forall t, exp Bool -> exp t -> exp t -> exp t
adamc@85 156
adamc@85 157 | Pair : forall t1 t2, exp t1 -> exp t2 -> exp (Prod t1 t2)
adamc@85 158 | Fst : forall t1 t2, exp (Prod t1 t2) -> exp t1
adamc@85 159 | Snd : forall t1 t2, exp (Prod t1 t2) -> exp t2.
adamc@85 160
adamc@85 161 (** We have a standard algebraic datatype [type], defining a type language of naturals, booleans, and product (pair) types. Then we have the indexed inductive type [exp], where the argument to [exp] tells us the encoded type of an expression. In effect, we are defining the typing rules for expressions simultaneously with the syntax.
adamc@85 162
adamc@85 163 We can give types and expressions semantics in a new style, based critically on the chance for %\textit{%#<i>#type-level computation#</i>#%}%. *)
adamc@85 164
adamc@85 165 Fixpoint typeDenote (t : type) : Set :=
adamc@85 166 match t with
adamc@85 167 | Nat => nat
adamc@85 168 | Bool => bool
adamc@85 169 | Prod t1 t2 => typeDenote t1 * typeDenote t2
adamc@85 170 end%type.
adamc@85 171
adamc@85 172 (** [typeDenote] compiles types of our object language into "native" Coq types. It is deceptively easy to implement. The only new thing we see is the [%type] annotation, which tells Coq to parse the [match] expression using the notations associated with types. Without this annotation, the [*] would be interpreted as multiplication on naturals, rather than as the product type constructor. [type] is one example of an identifer bound to a %\textit{%#<i>#notation scope#</i>#%}%. We will deal more explicitly with notations and notation scopes in later chapters.
adamc@85 173
adamc@85 174 We can define a function [expDenote] that is typed in terms of [typeDenote]. *)
adamc@85 175
adamc@85 176 Fixpoint expDenote t (e : exp t) {struct e} : typeDenote t :=
adamc@85 177 match e in (exp t) return (typeDenote t) with
adamc@85 178 | NConst n => n
adamc@85 179 | Plus e1 e2 => expDenote e1 + expDenote e2
adamc@85 180 | Eq e1 e2 => if eq_nat_dec (expDenote e1) (expDenote e2) then true else false
adamc@85 181
adamc@85 182 | BConst b => b
adamc@85 183 | And e1 e2 => expDenote e1 && expDenote e2
adamc@85 184 | If _ e' e1 e2 => if expDenote e' then expDenote e1 else expDenote e2
adamc@85 185
adamc@85 186 | Pair _ _ e1 e2 => (expDenote e1, expDenote e2)
adamc@85 187 | Fst _ _ e' => fst (expDenote e')
adamc@85 188 | Snd _ _ e' => snd (expDenote e')
adamc@85 189 end.
adamc@85 190
adamc@85 191 (** Again we find that an [in] annotation is essential for type-checking a function. Besides that, the definition is routine. In fact, it is less complicated than what we would write in ML or Haskell 98, since we do not need to worry about pushing final values in and out of an algebraic datatype. The only unusual thing is the use of an expression of the form [if E then true else false] in the [Eq] case. Remember that [eq_nat_dec] has a rich dependent type, rather than a simple boolean type. Coq's native [if] is overloaded to work on a test of any two-constructor type, so we can use [if] to build a simple boolean from the [sumbool] that [eq_nat_dec] returns.
adamc@85 192
adamc@85 193 We can implement our old favorite, a constant folding function, and prove it correct. It will be useful to write a function [pairOut] that checks if an [exp] of [Prod] type is a pair, returning its two components if so. Unsurprisingly, a first attempt leads to a type error.
adamc@85 194
adamc@85 195 [[
adamc@85 196 Definition pairOut t1 t2 (e : exp (Prod t1 t2)) : option (exp t1 * exp t2) :=
adamc@85 197 match e in (exp (Prod t1 t2)) return option (exp t1 * exp t2) with
adamc@85 198 | Pair _ _ e1 e2 => Some (e1, e2)
adamc@85 199 | _ => None
adamc@85 200 end.
adamc@85 201
adamc@85 202 [[
adamc@85 203 Error: The reference t2 was not found in the current environment
adamc@85 204 ]]
adamc@85 205
adamc@85 206 We run again into the problem of not being able to specify non-variable arguments in [in] clauses. The problem would just be hopeless without a use of an [in] clause, though, since the result type of the [match] depends on an argument to [exp]. Our solution will be to use a more general type, as we did for [hd]. First, we define a type-valued function to use in assigning a type to [pairOut]. *)
adamc@85 207
adamc@85 208 Definition pairOutType (t : type) :=
adamc@85 209 match t with
adamc@85 210 | Prod t1 t2 => option (exp t1 * exp t2)
adamc@85 211 | _ => unit
adamc@85 212 end.
adamc@85 213
adamc@85 214 (** When passed a type that is a product, [pairOutType] returns our final desired type. On any other input type, [pairOutType] returns [unit], since we do not care about extracting components of non-pairs. Now we can write another helper function to provide the default behavior of [pairOut], which we will apply for inputs that are not literal pairs. *)
adamc@85 215
adamc@85 216 Definition pairOutDefault (t : type) :=
adamc@85 217 match t return (pairOutType t) with
adamc@85 218 | Prod _ _ => None
adamc@85 219 | _ => tt
adamc@85 220 end.
adamc@85 221
adamc@85 222 (** Now [pairOut] is deceptively easy to write. *)
adamc@85 223
adamc@85 224 Definition pairOut t (e : exp t) :=
adamc@85 225 match e in (exp t) return (pairOutType t) with
adamc@85 226 | Pair _ _ e1 e2 => Some (e1, e2)
adamc@85 227 | _ => pairOutDefault _
adamc@85 228 end.
adamc@85 229
adamc@85 230 (** There is one important subtlety in this definition. Coq allows us to use convenient ML-style pattern matching notation, but, internally and in proofs, we see that patterns are expanded out completely, matching one level of inductive structure at a time. Thus, the default case in the [match] above expands out to one case for each constructor of [exp] besides [Pair], and the underscore in [pairOutDefault _] is resolved differently in each case. From an ML or Haskell programmer's perspective, what we have here is type inference determining which code is run (returning either [None] or [tt]), which goes beyond what is possible with type inference guiding parametric polymorphism in Hindley-Milner languages, but is similar to what goes on with Haskell type classes.
adamc@85 231
adamc@85 232 With [pairOut] available, we can write [cfold] in a straightforward way. There are really no surprises beyond that Coq verifies that this code has such an expressive type, given the small annotation burden. *)
adamc@85 233
adamc@85 234 Fixpoint cfold t (e : exp t) {struct e} : exp t :=
adamc@85 235 match e in (exp t) return (exp t) with
adamc@85 236 | NConst n => NConst n
adamc@85 237 | Plus e1 e2 =>
adamc@85 238 let e1' := cfold e1 in
adamc@85 239 let e2' := cfold e2 in
adamc@85 240 match e1', e2' with
adamc@85 241 | NConst n1, NConst n2 => NConst (n1 + n2)
adamc@85 242 | _, _ => Plus e1' e2'
adamc@85 243 end
adamc@85 244 | Eq e1 e2 =>
adamc@85 245 let e1' := cfold e1 in
adamc@85 246 let e2' := cfold e2 in
adamc@85 247 match e1', e2' with
adamc@85 248 | NConst n1, NConst n2 => BConst (if eq_nat_dec n1 n2 then true else false)
adamc@85 249 | _, _ => Eq e1' e2'
adamc@85 250 end
adamc@85 251
adamc@85 252 | BConst b => BConst b
adamc@85 253 | And e1 e2 =>
adamc@85 254 let e1' := cfold e1 in
adamc@85 255 let e2' := cfold e2 in
adamc@85 256 match e1', e2' with
adamc@85 257 | BConst b1, BConst b2 => BConst (b1 && b2)
adamc@85 258 | _, _ => And e1' e2'
adamc@85 259 end
adamc@85 260 | If _ e e1 e2 =>
adamc@85 261 let e' := cfold e in
adamc@85 262 match e' with
adamc@85 263 | BConst true => cfold e1
adamc@85 264 | BConst false => cfold e2
adamc@85 265 | _ => If e' (cfold e1) (cfold e2)
adamc@85 266 end
adamc@85 267
adamc@85 268 | Pair _ _ e1 e2 => Pair (cfold e1) (cfold e2)
adamc@85 269 | Fst _ _ e =>
adamc@85 270 let e' := cfold e in
adamc@85 271 match pairOut e' with
adamc@85 272 | Some p => fst p
adamc@85 273 | None => Fst e'
adamc@85 274 end
adamc@85 275 | Snd _ _ e =>
adamc@85 276 let e' := cfold e in
adamc@85 277 match pairOut e' with
adamc@85 278 | Some p => snd p
adamc@85 279 | None => Snd e'
adamc@85 280 end
adamc@85 281 end.
adamc@85 282
adamc@85 283 (** The correctness theorem for [cfold] turns out to be easy to prove, once we get over one serious hurdle. *)
adamc@85 284
adamc@85 285 Theorem cfold_correct : forall t (e : exp t), expDenote e = expDenote (cfold e).
adamc@85 286 induction e; crush.
adamc@85 287
adamc@85 288 (** The first remaining subgoal is:
adamc@85 289
adamc@85 290 [[
adamc@85 291
adamc@85 292 expDenote (cfold e1) + expDenote (cfold e2) =
adamc@85 293 expDenote
adamc@85 294 match cfold e1 with
adamc@85 295 | NConst n1 =>
adamc@85 296 match cfold e2 with
adamc@85 297 | NConst n2 => NConst (n1 + n2)
adamc@85 298 | Plus _ _ => Plus (cfold e1) (cfold e2)
adamc@85 299 | Eq _ _ => Plus (cfold e1) (cfold e2)
adamc@85 300 | BConst _ => Plus (cfold e1) (cfold e2)
adamc@85 301 | And _ _ => Plus (cfold e1) (cfold e2)
adamc@85 302 | If _ _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 303 | Pair _ _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 304 | Fst _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 305 | Snd _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 306 end
adamc@85 307 | Plus _ _ => Plus (cfold e1) (cfold e2)
adamc@85 308 | Eq _ _ => Plus (cfold e1) (cfold e2)
adamc@85 309 | BConst _ => Plus (cfold e1) (cfold e2)
adamc@85 310 | And _ _ => Plus (cfold e1) (cfold e2)
adamc@85 311 | If _ _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 312 | Pair _ _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 313 | Fst _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 314 | Snd _ _ _ => Plus (cfold e1) (cfold e2)
adamc@85 315 end
adamc@85 316 ]]
adamc@85 317
adamc@85 318 We would like to do a case analysis on [cfold e1], and we attempt that in the way that has worked so far.
adamc@85 319
adamc@85 320 [[
adamc@85 321 destruct (cfold e1).
adamc@85 322
adamc@85 323 [[
adamc@85 324 User error: e1 is used in hypothesis e
adamc@85 325 ]]
adamc@85 326
adamc@85 327 Coq gives us another cryptic error message. Like so many others, this one basically means that Coq is not able to build some proof about dependent types. It is hard to generate helpful and specific error messages for problems like this, since that would require some kind of understanding of the dependency structure of a piece of code. We will encounter many examples of case-specific tricks for recovering from errors like this one.
adamc@85 328
adamc@85 329 For our current proof, we can use a tactic [dep_destruct] defined in the book [Tactics] module. General elimination/inversion of dependently-typed hypotheses is undecidable, since it must be implemented with [match] expressions that have the restriction on [in] clauses that we have already discussed. [dep_destruct] makes a best effort to handle some common cases. In a future chapter, we will learn about the explicit manipulation of equality proofs that is behind [dep_destruct]'s implementation in Ltac, but for now, we treat it as a useful black box. *)
adamc@85 330
adamc@85 331 dep_destruct (cfold e1).
adamc@85 332
adamc@85 333 (** This successfully breaks the subgoal into 5 new subgoals, one for each constructor of [exp] that could produce an [exp Nat]. Note that [dep_destruct] is successful in ruling out the other cases automatically, in effect automating some of the work that we have done manually in implementing functions like [hd] and [pairOut].
adamc@85 334
adamc@85 335 This is the only new trick we need to learn to complete the proof. We can back up and give a short, automated proof. *)
adamc@85 336
adamc@85 337 Restart.
adamc@85 338
adamc@85 339 induction e; crush;
adamc@85 340 repeat (match goal with
adamc@85 341 | [ |- context[cfold ?E] ] => dep_destruct (cfold E)
adamc@85 342 | [ |- (if ?E then _ else _) = _ ] => destruct E
adamc@85 343 end; crush).
adamc@85 344 Qed.