Mercurial > cpdt > repo

annotate src/Coinductive.v @ 536:d65e9c1c9041

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Touch-ups in 8.4
author Adam Chlipala <adam@chlipala.net>
date Wed, 05 Aug 2015 18:07:57 -0400
parents ed829eaa91b2
children 81d63d9c1cc5
rev   line source
adam@534 1 (* Copyright (c) 2008-2012, 2015, Adam Chlipala
adamc@62 2 *
adamc@62 3 * This work is licensed under a
adamc@62 4 * Creative Commons Attribution-Noncommercial-No Derivative Works 3.0
adamc@62 5 * Unported License.
adamc@62 6 * The license text is available at:
adamc@62 7 * http://creativecommons.org/licenses/by-nc-nd/3.0/
adamc@62 8 *)
adamc@62 9
adamc@62 10 (* begin hide *)
adamc@62 11 Require Import List.
adamc@62 12
adam@534 13 Require Import Cpdt.CpdtTactics.
adamc@62 14
adam@422 15 Definition bad : unit := tt.
adam@402 16
adamc@62 17 Set Implicit Arguments.
adam@534 18 Set Asymmetric Patterns.
adamc@62 19 (* end hide *)
adamc@62 20
adamc@62 21
adamc@62 22 (** %\chapter{Infinite Data and Proofs}% *)
adamc@62 23
adam@395 24 (** In lazy functional programming languages like %\index{Haskell}%Haskell, infinite data structures are everywhere%~\cite{whyfp}%. Infinite lists and more exotic datatypes provide convenient abstractions for communication between parts of a program. Achieving similar convenience without infinite lazy structures would, in many cases, require acrobatic inversions of control flow.
adamc@62 25
adam@346 26 %\index{laziness}%Laziness is easy to implement in Haskell, where all the definitions in a program may be thought of as mutually recursive. In such an unconstrained setting, it is easy to implement an infinite loop when you really meant to build an infinite list, where any finite prefix of the list should be forceable in finite time. Haskell programmers learn how to avoid such slip-ups. In Coq, such a laissez-faire policy is not good enough.
adamc@62 27
adam@398 28 We spent some time in the last chapter discussing the %\index{Curry-Howard correspondence}%Curry-Howard isomorphism, where proofs are identified with functional programs. In such a setting, infinite loops, intended or otherwise, are disastrous. If Coq allowed the full breadth of definitions that Haskell did, we could code up an infinite loop and use it to prove any proposition vacuously. That is, the addition of general recursion would make CIC _inconsistent_. For an arbitrary proposition [P], we could write:
adamc@202 29 [[
adamc@202 30 Fixpoint bad (u : unit) : P := bad u.
adamc@205 31 ]]
adamc@205 32
adam@442 33 This would leave us with [bad tt] as a proof of [P].
adamc@62 34
adamc@62 35 There are also algorithmic considerations that make universal termination very desirable. We have seen how tactics like [reflexivity] compare terms up to equivalence under computational rules. Calls to recursive, pattern-matching functions are simplified automatically, with no need for explicit proof steps. It would be very hard to hold onto that kind of benefit if it became possible to write non-terminating programs; we would be running smack into the halting problem.
adamc@62 36
adam@422 37 One solution is to use types to contain the possibility of non-termination. For instance, we can create a "non-termination monad," inside which we must write all of our general-recursive programs; several such approaches are surveyed in Chapter 7. This is a heavyweight solution, and so we would like to avoid it whenever possible.
adamc@62 38
adam@402 39 Luckily, Coq has special support for a class of lazy data structures that happens to contain most examples found in Haskell. That mechanism,%\index{co-inductive types}% _co-inductive types_, is the subject of this chapter. *)
adamc@62 40
adamc@62 41
adamc@62 42 (** * Computing with Infinite Data *)
adamc@62 43
adam@398 44 (** Let us begin with the most basic type of infinite data, _streams_, or lazy lists.%\index{Vernacular commands!CoInductive}% *)
adamc@62 45
adamc@62 46 Section stream.
adam@351 47 Variable A : Type.
adamc@62 48
adam@351 49 CoInductive stream : Type :=
adamc@62 50 | Cons : A -> stream -> stream.
adamc@62 51 End stream.
adamc@62 52
adam@422 53 (* begin hide *)
adam@437 54 (* begin thide *)
adam@422 55 CoInductive evilStream := Nil.
adam@437 56 (* end thide *)
adam@422 57 (* end hide *)
adam@422 58
adamc@62 59 (** The definition is surprisingly simple. Starting from the definition of [list], we just need to change the keyword [Inductive] to [CoInductive]. We could have left a [Nil] constructor in our definition, but we will leave it out to force all of our streams to be infinite.
adamc@62 60
adam@471 61 How do we write down a stream constant? Obviously, simple application of constructors is not good enough, since we could only denote finite objects that way. Rather, whereas recursive definitions were necessary to _use_ values of recursive inductive types effectively, here we find that we need%\index{co-recursive definitions}% _co-recursive definitions_ to _build_ values of co-inductive types effectively.
adamc@62 62
adam@346 63 We can define a stream consisting only of zeroes.%\index{Vernacular commands!CoFixpoint}% *)
adamc@62 64
adamc@62 65 CoFixpoint zeroes : stream nat := Cons 0 zeroes.
adamc@62 66
adam@346 67 (* EX: Define a stream that alternates between [true] and [false]. *)
adam@346 68 (* begin thide *)
adam@346 69
adamc@62 70 (** We can also define a stream that alternates between [true] and [false]. *)
adamc@62 71
adam@346 72 CoFixpoint trues_falses : stream bool := Cons true falses_trues
adam@346 73 with falses_trues : stream bool := Cons false trues_falses.
adam@346 74 (* end thide *)
adamc@62 75
adamc@62 76 (** Co-inductive values are fair game as arguments to recursive functions, and we can use that fact to write a function to take a finite approximation of a stream. *)
adamc@62 77
adam@348 78 (* EX: Define a function to calculate a finite approximation of a stream, to a particular length. *)
adam@346 79 (* begin thide *)
adam@346 80
adamc@211 81 Fixpoint approx A (s : stream A) (n : nat) : list A :=
adamc@62 82 match n with
adamc@62 83 | O => nil
adamc@62 84 | S n' =>
adamc@62 85 match s with
adamc@62 86 | Cons h t => h :: approx t n'
adamc@62 87 end
adamc@62 88 end.
adamc@62 89
adamc@62 90 Eval simpl in approx zeroes 10.
adamc@211 91 (** %\vspace{-.15in}% [[
adamc@62 92 = 0 :: 0 :: 0 :: 0 :: 0 :: 0 :: 0 :: 0 :: 0 :: 0 :: nil
adamc@62 93 : list nat
adam@302 94 ]]
adam@302 95 *)
adamc@211 96
adam@346 97 Eval simpl in approx trues_falses 10.
adamc@211 98 (** %\vspace{-.15in}% [[
adamc@62 99 = true
adamc@62 100 :: false
adamc@62 101 :: true
adamc@62 102 :: false
adamc@62 103 :: true :: false :: true :: false :: true :: false :: nil
adamc@62 104 : list bool
adam@346 105 ]]
adam@346 106 *)
adamc@62 107
adam@349 108 (* end thide *)
adamc@62 109
adam@402 110 (* begin hide *)
adam@437 111 (* begin thide *)
adam@402 112 Definition looper := 0.
adam@437 113 (* end thide *)
adam@402 114 (* end hide *)
adam@402 115
adam@398 116 (** So far, it looks like co-inductive types might be a magic bullet, allowing us to import all of the Haskeller's usual tricks. However, there are important restrictions that are dual to the restrictions on the use of inductive types. Fixpoints _consume_ values of inductive types, with restrictions on which _arguments_ may be passed in recursive calls. Dually, co-fixpoints _produce_ values of co-inductive types, with restrictions on what may be done with the _results_ of co-recursive calls.
adamc@62 117
adam@402 118 The restriction for co-inductive types shows up as the%\index{guardedness condition}% _guardedness condition_. First, consider this stream definition, which would be legal in Haskell.
adamc@62 119 [[
adamc@62 120 CoFixpoint looper : stream nat := looper.
adam@346 121 ]]
adamc@205 122
adam@346 123 <<
adamc@62 124 Error:
adamc@62 125 Recursive definition of looper is ill-formed.
adamc@62 126 In environment
adamc@62 127 looper : stream nat
adamc@62 128
adamc@62 129 unguarded recursive call in "looper"
adam@346 130 >>
adamc@205 131
adam@398 132 The rule we have run afoul of here is that _every co-recursive call must be guarded by a constructor_; that is, every co-recursive call must be a direct argument to a constructor of the co-inductive type we are generating. It is a good thing that this rule is enforced. If the definition of [looper] were accepted, our [approx] function would run forever when passed [looper], and we would have fallen into inconsistency.
adamc@62 133
adam@346 134 Some familiar functions are easy to write in co-recursive fashion. *)
adamc@62 135
adamc@62 136 Section map.
adam@351 137 Variables A B : Type.
adamc@62 138 Variable f : A -> B.
adamc@62 139
adamc@62 140 CoFixpoint map (s : stream A) : stream B :=
adamc@62 141 match s with
adamc@62 142 | Cons h t => Cons (f h) (map t)
adamc@62 143 end.
adamc@62 144 End map.
adamc@62 145
adam@402 146 (* begin hide *)
adam@437 147 (* begin thide *)
adam@402 148 Definition filter := 0.
adam@437 149 (* end thide *)
adam@402 150 (* end hide *)
adam@402 151
adam@402 152 (** This code is a literal copy of that for the list [map] function, with the [nil] case removed and [Fixpoint] changed to [CoFixpoint]. Many other standard functions on lazy data structures can be implemented just as easily. Some, like [filter], cannot be implemented. Since the predicate passed to [filter] may reject every element of the stream, we cannot satisfy the guardedness condition.
adamc@62 153
adam@398 154 The implications of the condition can be subtle. To illustrate how, we start off with another co-recursive function definition that _is_ legal. The function [interleave] takes two streams and produces a new stream that alternates between their elements. *)
adamc@62 155
adamc@62 156 Section interleave.
adam@351 157 Variable A : Type.
adamc@62 158
adamc@62 159 CoFixpoint interleave (s1 s2 : stream A) : stream A :=
adamc@62 160 match s1, s2 with
adamc@62 161 | Cons h1 t1, Cons h2 t2 => Cons h1 (Cons h2 (interleave t1 t2))
adamc@62 162 end.
adamc@62 163 End interleave.
adamc@62 164
adamc@62 165 (** Now say we want to write a weird stuttering version of [map] that repeats elements in a particular way, based on interleaving. *)
adamc@62 166
adamc@62 167 Section map'.
adam@351 168 Variables A B : Type.
adamc@62 169 Variable f : A -> B.
adamc@68 170 (* begin thide *)
adam@440 171 (** %\vspace{-.15in}%[[
adamc@62 172 CoFixpoint map' (s : stream A) : stream B :=
adamc@62 173 match s with
adam@346 174 | Cons h t => interleave (Cons (f h) (map' t)) (Cons (f h) (map' t))
adamc@68 175 end.
adamc@205 176 ]]
adam@440 177 %\vspace{-.15in}%We get another error message about an unguarded recursive call. *)
adamc@62 178
adam@346 179 End map'.
adam@346 180
adam@475 181 (** What is going wrong here? Imagine that, instead of [interleave], we had called some other, less well-behaved function on streams. Here is one simpler example demonstrating the essential pitfall. We start by defining a standard function for taking the tail of a stream. Since streams are infinite, this operation is total. *)
adam@346 182
adam@346 183 Definition tl A (s : stream A) : stream A :=
adam@346 184 match s with
adam@346 185 | Cons _ s' => s'
adam@346 186 end.
adam@346 187
adam@346 188 (** Coq rejects the following definition that uses [tl].
adam@346 189 [[
adam@346 190 CoFixpoint bad : stream nat := tl (Cons 0 bad).
adam@346 191 ]]
adam@346 192
adam@422 193 Imagine that Coq had accepted our definition, and consider how we might evaluate [approx bad 1]. We would be trying to calculate the first element in the stream [bad]. However, it is not hard to see that the definition of [bad] "begs the question": unfolding the definition of [tl], we see that we essentially say "define [bad] to equal itself"! Of course such an equation admits no single well-defined solution, which does not fit well with the determinism of Gallina reduction.
adam@346 194
adam@471 195 Coq's complete rule for co-recursive definitions includes not just the basic guardedness condition, but also a requirement about where co-recursive calls may occur. In particular, a co-recursive call must be a direct argument to a constructor, _nested only inside of other constructor calls or [fun] or [match] expressions_. In the definition of [bad], we erroneously nested the co-recursive call inside a call to [tl], and we nested inside a call to [interleave] in the definition of [map'].
adam@471 196
adam@471 197 Coq helps the user out a little by performing the guardedness check after using computation to simplify terms. For instance, any co-recursive function definition can be expanded by inserting extra calls to an identity function, and this change preserves guardedness. However, in other cases computational simplification can reveal why definitions are dangerous. Consider what happens when we inline the definition of [tl] in [bad]:
adam@346 198 [[
adam@346 199 CoFixpoint bad : stream nat := bad.
adam@346 200 ]]
adam@471 201 This is the same looping definition we rejected earlier. A similar inlining process reveals an alternate view on our failed definition of [map']:
adam@346 202 [[
adam@346 203 CoFixpoint map' (s : stream A) : stream B :=
adam@346 204 match s with
adam@346 205 | Cons h t => Cons (f h) (Cons (f h) (interleave (map' t) (map' t)))
adam@346 206 end.
adam@346 207 ]]
adam@346 208 Clearly in this case the [map'] calls are not immediate arguments to constructors, so we violate the guardedness condition. *)
adamc@68 209 (* end thide *)
adamc@211 210
adam@402 211 (** A more interesting question is why that condition is the right one. We can make an intuitive argument that the original [map'] definition is perfectly reasonable and denotes a well-understood transformation on streams, such that every output would behave properly with [approx]. The guardedness condition is an example of a syntactic check for%\index{productivity}% _productivity_ of co-recursive definitions. A productive definition can be thought of as one whose outputs can be forced in finite time to any finite approximation level, as with [approx]. If we replaced the guardedness condition with more involved checks, we might be able to detect and allow a broader range of productive definitions. However, mistakes in these checks could cause inconsistency, and programmers would need to understand the new, more complex checks. Coq's design strikes a balance between consistency and simplicity with its choice of guard condition, though we can imagine other worthwhile balances being struck, too. *)
adamc@62 212
adamc@63 213
adamc@63 214 (** * Infinite Proofs *)
adamc@63 215
adamc@63 216 (** Let us say we want to give two different definitions of a stream of all ones, and then we want to prove that they are equivalent. *)
adamc@63 217
adamc@63 218 CoFixpoint ones : stream nat := Cons 1 ones.
adamc@63 219 Definition ones' := map S zeroes.
adamc@63 220
adamc@63 221 (** The obvious statement of the equality is this: *)
adamc@63 222
adamc@63 223 Theorem ones_eq : ones = ones'.
adamc@63 224
adam@422 225 (* begin hide *)
adam@437 226 (* begin thide *)
adam@422 227 Definition foo := @eq.
adam@437 228 (* end thide *)
adam@422 229 (* end hide *)
adam@422 230
adamc@63 231 (** However, faced with the initial subgoal, it is not at all clear how this theorem can be proved. In fact, it is unprovable. The [eq] predicate that we use is fundamentally limited to equalities that can be demonstrated by finite, syntactic arguments. To prove this equivalence, we will need to introduce a new relation. *)
adamc@68 232 (* begin thide *)
adamc@211 233
adamc@63 234 Abort.
adamc@63 235
adam@398 236 (** Co-inductive datatypes make sense by analogy from Haskell. What we need now is a _co-inductive proposition_. That is, we want to define a proposition whose proofs may be infinite, subject to the guardedness condition. The idea of infinite proofs does not show up in usual mathematics, but it can be very useful (unsurprisingly) for reasoning about infinite data structures. Besides examples from Haskell, infinite data and proofs will also turn out to be useful for modelling inherently infinite mathematical objects, like program executions.
adamc@63 237
adam@346 238 We are ready for our first %\index{co-inductive predicates}%co-inductive predicate. *)
adamc@63 239
adamc@63 240 Section stream_eq.
adam@351 241 Variable A : Type.
adamc@63 242
adamc@63 243 CoInductive stream_eq : stream A -> stream A -> Prop :=
adamc@63 244 | Stream_eq : forall h t1 t2,
adamc@63 245 stream_eq t1 t2
adamc@63 246 -> stream_eq (Cons h t1) (Cons h t2).
adamc@63 247 End stream_eq.
adamc@63 248
adamc@63 249 (** We say that two streams are equal if and only if they have the same heads and their tails are equal. We use the normal finite-syntactic equality for the heads, and we refer to our new equality recursively for the tails.
adamc@63 250
adamc@63 251 We can try restating the theorem with [stream_eq]. *)
adamc@63 252
adamc@63 253 Theorem ones_eq : stream_eq ones ones'.
adamc@63 254 (** Coq does not support tactical co-inductive proofs as well as it supports tactical inductive proofs. The usual starting point is the [cofix] tactic, which asks to structure this proof as a co-fixpoint. *)
adamc@211 255
adamc@63 256 cofix.
adamc@63 257 (** [[
adamc@63 258 ones_eq : stream_eq ones ones'
adamc@63 259 ============================
adamc@63 260 stream_eq ones ones'
adamc@211 261
adamc@211 262 ]]
adamc@63 263
adamc@211 264 It looks like this proof might be easier than we expected! *)
adamc@63 265
adamc@63 266 assumption.
adam@422 267 (** <<
adamc@211 268 Proof completed.
adam@422 269 >>
adamc@63 270
adamc@211 271 Unfortunately, we are due for some disappointment in our victory lap.
adamc@211 272 [[
adamc@63 273 Qed.
adam@346 274 ]]
adamc@63 275
adam@346 276 <<
adamc@63 277 Error:
adamc@63 278 Recursive definition of ones_eq is ill-formed.
adamc@63 279
adamc@63 280 In environment
adamc@63 281 ones_eq : stream_eq ones ones'
adamc@63 282
adamc@205 283 unguarded recursive call in "ones_eq"
adam@346 284 >>
adamc@205 285
adamc@211 286 Via the Curry-Howard correspondence, the same guardedness condition applies to our co-inductive proofs as to our co-inductive data structures. We should be grateful that this proof is rejected, because, if it were not, the same proof structure could be used to prove any co-inductive theorem vacuously, by direct appeal to itself!
adamc@63 287
adam@347 288 Thinking about how Coq would generate a proof term from the proof script above, we see that the problem is that we are violating the guardedness condition. During our proofs, Coq can help us check whether we have yet gone wrong in this way. We can run the command [Guarded] in any context to see if it is possible to finish the proof in a way that will yield a properly guarded proof term.%\index{Vernacular commands!Guarded}%
adamc@63 289 [[
adamc@63 290 Guarded.
adamc@205 291 ]]
adamc@205 292
adam@398 293 Running [Guarded] here gives us the same error message that we got when we tried to run [Qed]. In larger proofs, [Guarded] can be helpful in detecting problems _before_ we think we are ready to run [Qed].
adamc@63 294
adam@281 295 We need to start the co-induction by applying [stream_eq]'s constructor. To do that, we need to know that both arguments to the predicate are [Cons]es. Informally, this is trivial, but [simpl] is not able to help us. *)
adamc@63 296
adamc@63 297 Undo.
adamc@63 298 simpl.
adamc@63 299 (** [[
adamc@63 300 ones_eq : stream_eq ones ones'
adamc@63 301 ============================
adamc@63 302 stream_eq ones ones'
adamc@211 303
adamc@211 304 ]]
adamc@63 305
adamc@211 306 It turns out that we are best served by proving an auxiliary lemma. *)
adamc@211 307
adamc@63 308 Abort.
adamc@63 309
adam@450 310 (** First, we need to define a function that seems pointless at first glance. *)
adamc@63 311
adamc@63 312 Definition frob A (s : stream A) : stream A :=
adamc@63 313 match s with
adamc@63 314 | Cons h t => Cons h t
adamc@63 315 end.
adamc@63 316
adamc@63 317 (** Next, we need to prove a theorem that seems equally pointless. *)
adamc@63 318
adamc@63 319 Theorem frob_eq : forall A (s : stream A), s = frob s.
adamc@63 320 destruct s; reflexivity.
adamc@63 321 Qed.
adamc@63 322
adamc@63 323 (** But, miraculously, this theorem turns out to be just what we needed. *)
adamc@63 324
adamc@63 325 Theorem ones_eq : stream_eq ones ones'.
adamc@63 326 cofix.
adamc@63 327
adamc@63 328 (** We can use the theorem to rewrite the two streams. *)
adamc@211 329
adamc@63 330 rewrite (frob_eq ones).
adamc@63 331 rewrite (frob_eq ones').
adamc@63 332 (** [[
adamc@63 333 ones_eq : stream_eq ones ones'
adamc@63 334 ============================
adamc@63 335 stream_eq (frob ones) (frob ones')
adamc@211 336
adamc@211 337 ]]
adamc@63 338
adamc@211 339 Now [simpl] is able to reduce the streams. *)
adamc@63 340
adamc@63 341 simpl.
adamc@63 342 (** [[
adamc@63 343 ones_eq : stream_eq ones ones'
adamc@63 344 ============================
adamc@63 345 stream_eq (Cons 1 ones)
adamc@63 346 (Cons 1
adamc@63 347 ((cofix map (s : stream nat) : stream nat :=
adamc@63 348 match s with
adamc@63 349 | Cons h t => Cons (S h) (map t)
adamc@63 350 end) zeroes))
adamc@211 351
adamc@211 352 ]]
adamc@63 353
adam@422 354 Note the [cofix] notation for anonymous co-recursion, which is analogous to the [fix] notation we have already seen for recursion. Since we have exposed the [Cons] structure of each stream, we can apply the constructor of [stream_eq]. *)
adamc@63 355
adamc@63 356 constructor.
adamc@63 357 (** [[
adamc@63 358 ones_eq : stream_eq ones ones'
adamc@63 359 ============================
adamc@63 360 stream_eq ones
adamc@63 361 ((cofix map (s : stream nat) : stream nat :=
adamc@63 362 match s with
adamc@63 363 | Cons h t => Cons (S h) (map t)
adamc@63 364 end) zeroes)
adamc@211 365
adamc@211 366 ]]
adamc@63 367
adamc@211 368 Now, modulo unfolding of the definition of [map], we have matched our assumption. *)
adamc@211 369
adamc@63 370 assumption.
adamc@63 371 Qed.
adamc@63 372
adamc@63 373 (** Why did this silly-looking trick help? The answer has to do with the constraints placed on Coq's evaluation rules by the need for termination. The [cofix]-related restriction that foiled our first attempt at using [simpl] is dual to a restriction for [fix]. In particular, an application of an anonymous [fix] only reduces when the top-level structure of the recursive argument is known. Otherwise, we would be unfolding the recursive definition ad infinitum.
adamc@63 374
adam@398 375 Fixpoints only reduce when enough is known about the _definitions_ of their arguments. Dually, co-fixpoints only reduce when enough is known about _how their results will be used_. In particular, a [cofix] is only expanded when it is the discriminee of a [match]. Rewriting with our superficially silly lemma wrapped new [match]es around the two [cofix]es, triggering reduction.
adamc@63 376
adamc@63 377 If [cofix]es reduced haphazardly, it would be easy to run into infinite loops in evaluation, since we are, after all, building infinite objects.
adamc@63 378
adamc@63 379 One common source of difficulty with co-inductive proofs is bad interaction with standard Coq automation machinery. If we try to prove [ones_eq'] with automation, like we have in previous inductive proofs, we get an invalid proof. *)
adamc@63 380
adamc@63 381 Theorem ones_eq' : stream_eq ones ones'.
adamc@63 382 cofix; crush.
adam@440 383 (** %\vspace{-.25in}%[[
adamc@205 384 Guarded.
adam@302 385 ]]
adam@440 386 %\vspace{-.25in}%
adam@302 387 *)
adamc@63 388 Abort.
adam@346 389
adam@471 390 (** The standard [auto] machinery sees that our goal matches an assumption and so applies that assumption, even though this violates guardedness. A correct proof strategy for a theorem like this usually starts by [destruct]ing some parameter and running a custom tactic to figure out the first proof rule to apply for each case. Alternatively, there are tricks that can be played with "hiding" the co-inductive hypothesis.
adam@346 391
adam@346 392 %\medskip%
adam@346 393
adam@402 394 Must we always be cautious with automation in proofs by co-induction? Induction seems to have dual versions of the same pitfalls inherent in it, and yet we avoid those pitfalls by encapsulating safe Curry-Howard recursion schemes inside named induction principles. It turns out that we can usually do the same with%\index{co-induction principles}% _co-induction principles_. Let us take that tack here, so that we can arrive at an [induction x; crush]-style proof for [ones_eq'].
adam@346 395
adam@398 396 An induction principle is parameterized over a predicate characterizing what we mean to prove, _as a function of the inductive fact that we already know_. Dually, a co-induction principle ought to be parameterized over a predicate characterizing what we mean to prove, _as a function of the arguments to the co-inductive predicate that we are trying to prove_.
adam@346 397
adam@346 398 To state a useful principle for [stream_eq], it will be useful first to define the stream head function. *)
adam@346 399
adam@346 400 Definition hd A (s : stream A) : A :=
adam@346 401 match s with
adam@346 402 | Cons x _ => x
adam@346 403 end.
adam@346 404
adam@346 405 (** Now we enter a section for the co-induction principle, based on %\index{Park's principle}%Park's principle as introduced in a tutorial by Gim%\'%enez%~\cite{IT}%. *)
adam@346 406
adam@346 407 Section stream_eq_coind.
adam@351 408 Variable A : Type.
adam@346 409 Variable R : stream A -> stream A -> Prop.
adam@475 410
adam@471 411 (** This relation generalizes the theorem we want to prove, defining a set of pairs of streams that we must eventually prove contains the particular pair we care about. *)
adam@346 412
adam@346 413 Hypothesis Cons_case_hd : forall s1 s2, R s1 s2 -> hd s1 = hd s2.
adam@346 414 Hypothesis Cons_case_tl : forall s1 s2, R s1 s2 -> R (tl s1) (tl s2).
adam@475 415
adam@422 416 (** Two hypotheses characterize what makes a good choice of [R]: it enforces equality of stream heads, and it is %``%#<i>#hereditary#</i>#%''% in the sense that an [R] stream pair passes on "[R]-ness" to its tails. An established technical term for such a relation is%\index{bisimulation}% _bisimulation_. *)
adam@346 417
adam@346 418 (** Now it is straightforward to prove the principle, which says that any stream pair in [R] is equal. The reader may wish to step through the proof script to see what is going on. *)
adam@392 419
adam@346 420 Theorem stream_eq_coind : forall s1 s2, R s1 s2 -> stream_eq s1 s2.
adam@346 421 cofix; destruct s1; destruct s2; intro.
adam@346 422 generalize (Cons_case_hd H); intro Heq; simpl in Heq; rewrite Heq.
adam@346 423 constructor.
adam@346 424 apply stream_eq_coind.
adam@346 425 apply (Cons_case_tl H).
adam@346 426 Qed.
adam@346 427 End stream_eq_coind.
adam@346 428
adam@346 429 (** To see why this proof is guarded, we can print it and verify that the one co-recursive call is an immediate argument to a constructor. *)
adam@392 430
adam@346 431 Print stream_eq_coind.
adam@346 432
adam@346 433 (** We omit the output and proceed to proving [ones_eq''] again. The only bit of ingenuity is in choosing [R], and in this case the most restrictive predicate works. *)
adam@346 434
adam@346 435 Theorem ones_eq'' : stream_eq ones ones'.
adam@346 436 apply (stream_eq_coind (fun s1 s2 => s1 = ones /\ s2 = ones')); crush.
adam@346 437 Qed.
adam@346 438
adam@346 439 (** Note that this proof achieves the proper reduction behavior via [hd] and [tl], rather than [frob] as in the last proof. All three functions pattern match on their arguments, catalyzing computation steps.
adam@346 440
adam@346 441 Compared to the inductive proofs that we are used to, it still seems unsatisfactory that we had to write out a choice of [R] in the last proof. An alternate is to capture a common pattern of co-recursion in a more specialized co-induction principle. For the current example, that pattern is: prove [stream_eq s1 s2] where [s1] and [s2] are defined as their own tails. *)
adam@346 442
adam@346 443 Section stream_eq_loop.
adam@351 444 Variable A : Type.
adam@346 445 Variables s1 s2 : stream A.
adam@346 446
adam@346 447 Hypothesis Cons_case_hd : hd s1 = hd s2.
adam@346 448 Hypothesis loop1 : tl s1 = s1.
adam@346 449 Hypothesis loop2 : tl s2 = s2.
adam@346 450
adam@346 451 (** The proof of the principle includes a choice of [R], so that we no longer need to make such choices thereafter. *)
adam@346 452
adam@346 453 Theorem stream_eq_loop : stream_eq s1 s2.
adam@346 454 apply (stream_eq_coind (fun s1' s2' => s1' = s1 /\ s2' = s2)); crush.
adam@346 455 Qed.
adam@346 456 End stream_eq_loop.
adam@346 457
adam@346 458 Theorem ones_eq''' : stream_eq ones ones'.
adam@346 459 apply stream_eq_loop; crush.
adam@346 460 Qed.
adamc@68 461 (* end thide *)
adamc@63 462
adam@435 463 (** Let us put [stream_eq_coind] through its paces a bit more, considering two different ways to compute infinite streams of all factorial values. First, we import the [fact] factorial function from the standard library. *)
adam@346 464
adam@346 465 Require Import Arith.
adam@346 466 Print fact.
adam@346 467 (** %\vspace{-.15in}%[[
adam@346 468 fact =
adam@346 469 fix fact (n : nat) : nat :=
adam@346 470 match n with
adam@346 471 | 0 => 1
adam@346 472 | S n0 => S n0 * fact n0
adam@346 473 end
adam@346 474 : nat -> nat
adam@346 475 ]]
adam@346 476 *)
adam@346 477
adam@346 478 (** The simplest way to compute the factorial stream involves calling [fact] afresh at each position. *)
adam@346 479
adam@346 480 CoFixpoint fact_slow' (n : nat) := Cons (fact n) (fact_slow' (S n)).
adam@346 481 Definition fact_slow := fact_slow' 1.
adam@346 482
adam@346 483 (** A more clever, optimized method maintains an accumulator of the previous factorial, so that each new entry can be computed with a single multiplication. *)
adam@346 484
adam@346 485 CoFixpoint fact_iter' (cur acc : nat) := Cons acc (fact_iter' (S cur) (acc * cur)).
adam@346 486 Definition fact_iter := fact_iter' 2 1.
adam@346 487
adam@346 488 (** We can verify that the streams are equal up to particular finite bounds. *)
adam@346 489
adam@346 490 Eval simpl in approx fact_iter 5.
adam@346 491 (** %\vspace{-.15in}%[[
adam@346 492 = 1 :: 2 :: 6 :: 24 :: 120 :: nil
adam@346 493 : list nat
adam@346 494 ]]
adam@346 495 *)
adam@346 496 Eval simpl in approx fact_slow 5.
adam@346 497 (** %\vspace{-.15in}%[[
adam@346 498 = 1 :: 2 :: 6 :: 24 :: 120 :: nil
adam@346 499 : list nat
adam@346 500 ]]
adam@346 501
adam@471 502 Now, to prove that the two versions are equivalent, it is helpful to prove (and add as a proof hint) a quick lemma about the computational behavior of [fact]. (I intentionally skip explaining its proof at this point.) *)
adam@346 503
adam@346 504 (* begin thide *)
adam@346 505 Lemma fact_def : forall x n,
adam@346 506 fact_iter' x (fact n * S n) = fact_iter' x (fact (S n)).
adam@346 507 simpl; intros; f_equal; ring.
adam@346 508 Qed.
adam@346 509
adam@346 510 Hint Resolve fact_def.
adam@346 511
adam@346 512 (** With the hint added, it is easy to prove an auxiliary lemma relating [fact_iter'] and [fact_slow']. The key bit of ingenuity is introduction of an existential quantifier for the shared parameter [n]. *)
adam@346 513
adam@346 514 Lemma fact_eq' : forall n, stream_eq (fact_iter' (S n) (fact n)) (fact_slow' n).
adam@346 515 intro; apply (stream_eq_coind (fun s1 s2 => exists n, s1 = fact_iter' (S n) (fact n)
adam@346 516 /\ s2 = fact_slow' n)); crush; eauto.
adam@346 517 Qed.
adam@346 518
adam@346 519 (** The final theorem is a direct corollary of [fact_eq']. *)
adam@346 520
adam@346 521 Theorem fact_eq : stream_eq fact_iter fact_slow.
adam@346 522 apply fact_eq'.
adam@346 523 Qed.
adam@346 524
adam@346 525 (** As in the case of [ones_eq'], we may be unsatisfied that we needed to write down a choice of [R] that seems to duplicate information already present in a lemma statement. We can facilitate a simpler proof by defining a co-induction principle specialized to goals that begin with single universal quantifiers, and the strategy can be extended in a straightforward way to principles for other counts of quantifiers. (Our [stream_eq_loop] principle is effectively the instantiation of this technique to zero quantifiers.) *)
adam@346 526
adam@346 527 Section stream_eq_onequant.
adam@351 528 Variables A B : Type.
adam@346 529 (** We have the types [A], the domain of the one quantifier; and [B], the type of data found in the streams. *)
adam@346 530
adam@346 531 Variables f g : A -> stream B.
adam@346 532 (** The two streams we compare must be of the forms [f x] and [g x], for some shared [x]. Note that this falls out naturally when [x] is a shared universally quantified variable in a lemma statement. *)
adam@346 533
adam@346 534 Hypothesis Cons_case_hd : forall x, hd (f x) = hd (g x).
adam@346 535 Hypothesis Cons_case_tl : forall x, exists y, tl (f x) = f y /\ tl (g x) = g y.
adam@346 536 (** These conditions are inspired by the bisimulation requirements, with a more general version of the [R] choice we made for [fact_eq'] inlined into the hypotheses of [stream_eq_coind]. *)
adam@346 537
adam@346 538 Theorem stream_eq_onequant : forall x, stream_eq (f x) (g x).
adam@346 539 intro; apply (stream_eq_coind (fun s1 s2 => exists x, s1 = f x /\ s2 = g x)); crush; eauto.
adam@346 540 Qed.
adam@346 541 End stream_eq_onequant.
adam@346 542
adam@346 543 Lemma fact_eq'' : forall n, stream_eq (fact_iter' (S n) (fact n)) (fact_slow' n).
adam@346 544 apply stream_eq_onequant; crush; eauto.
adam@346 545 Qed.
adam@346 546
adam@346 547 (** We have arrived at one of our customary automated proofs, thanks to the new principle. *)
adam@346 548 (* end thide *)
adamc@64 549
adamc@64 550
adamc@64 551 (** * Simple Modeling of Non-Terminating Programs *)
adamc@64 552
adam@402 553 (** We close the chapter with a quick motivating example for more complex uses of co-inductive types. We will define a co-inductive semantics for a simple imperative programming language and use that semantics to prove the correctness of a trivial optimization that removes spurious additions by 0. We follow the technique of%\index{co-inductive big-step operational semantics}% _co-inductive big-step operational semantics_ %\cite{BigStep}%.
adamc@64 554
adam@471 555 We define a suggestive synonym for [nat], as we will consider programs over infinitely many variables, represented as [nat]s. *)
adamc@211 556
adam@347 557 Definition var := nat.
adamc@64 558
adam@436 559 (** We define a type [vars] of maps from variables to values. To define a function [set] for setting a variable's value in a map, we use the standard library function [beq_nat] for comparing natural numbers. *)
adamc@64 560
adam@347 561 Definition vars := var -> nat.
adam@347 562 Definition set (vs : vars) (v : var) (n : nat) : vars :=
adam@347 563 fun v' => if beq_nat v v' then n else vs v'.
adamc@67 564
adam@347 565 (** We define a simple arithmetic expression language with variables, and we give it a semantics via an interpreter. *)
adamc@67 566
adam@347 567 Inductive exp : Set :=
adam@347 568 | Const : nat -> exp
adam@347 569 | Var : var -> exp
adam@347 570 | Plus : exp -> exp -> exp.
adamc@64 571
adam@347 572 Fixpoint evalExp (vs : vars) (e : exp) : nat :=
adam@347 573 match e with
adam@347 574 | Const n => n
adam@347 575 | Var v => vs v
adam@347 576 | Plus e1 e2 => evalExp vs e1 + evalExp vs e2
adam@347 577 end.
adamc@64 578
adam@422 579 (** Finally, we define a language of commands. It includes variable assignment, sequencing, and a <<while>> form that repeats as long as its test expression evaluates to a nonzero value. *)
adamc@64 580
adam@347 581 Inductive cmd : Set :=
adam@347 582 | Assign : var -> exp -> cmd
adam@347 583 | Seq : cmd -> cmd -> cmd
adam@347 584 | While : exp -> cmd -> cmd.
adamc@64 585
adam@471 586 (** We could define an inductive relation to characterize the results of command evaluation. However, such a relation would not capture _nonterminating_ executions. With a co-inductive relation, we can capture both cases. The parameters of the relation are an initial state, a command, and a final state. A program that does not terminate in a particular initial state is related to _any_ final state. For more realistic languages than this one, it is often possible for programs to _crash_, in which case a semantics would generally relate their executions to no final states; so relating safely non-terminating programs to all final states provides a crucial distinction. *)
adamc@67 587
adam@347 588 CoInductive evalCmd : vars -> cmd -> vars -> Prop :=
adam@347 589 | EvalAssign : forall vs v e, evalCmd vs (Assign v e) (set vs v (evalExp vs e))
adam@347 590 | EvalSeq : forall vs1 vs2 vs3 c1 c2, evalCmd vs1 c1 vs2
adam@347 591 -> evalCmd vs2 c2 vs3
adam@347 592 -> evalCmd vs1 (Seq c1 c2) vs3
adam@347 593 | EvalWhileFalse : forall vs e c, evalExp vs e = 0
adam@347 594 -> evalCmd vs (While e c) vs
adam@347 595 | EvalWhileTrue : forall vs1 vs2 vs3 e c, evalExp vs1 e <> 0
adam@347 596 -> evalCmd vs1 c vs2
adam@347 597 -> evalCmd vs2 (While e c) vs3
adam@347 598 -> evalCmd vs1 (While e c) vs3.
adam@347 599
adam@347 600 (** Having learned our lesson in the last section, before proceeding, we build a co-induction principle for [evalCmd]. *)
adam@347 601
adam@347 602 Section evalCmd_coind.
adam@347 603 Variable R : vars -> cmd -> vars -> Prop.
adam@347 604
adam@347 605 Hypothesis AssignCase : forall vs1 vs2 v e, R vs1 (Assign v e) vs2
adam@347 606 -> vs2 = set vs1 v (evalExp vs1 e).
adam@347 607
adam@347 608 Hypothesis SeqCase : forall vs1 vs3 c1 c2, R vs1 (Seq c1 c2) vs3
adam@347 609 -> exists vs2, R vs1 c1 vs2 /\ R vs2 c2 vs3.
adam@347 610
adam@347 611 Hypothesis WhileCase : forall vs1 vs3 e c, R vs1 (While e c) vs3
adam@347 612 -> (evalExp vs1 e = 0 /\ vs3 = vs1)
adam@347 613 \/ exists vs2, evalExp vs1 e <> 0 /\ R vs1 c vs2 /\ R vs2 (While e c) vs3.
adam@347 614
adam@402 615 (** The proof is routine. We make use of a form of %\index{tactics!destruct}%[destruct] that takes an%\index{intro pattern}% _intro pattern_ in an [as] clause. These patterns control how deeply we break apart the components of an inductive value, and we refer the reader to the Coq manual for more details. *)
adam@347 616
adam@347 617 Theorem evalCmd_coind : forall vs1 c vs2, R vs1 c vs2 -> evalCmd vs1 c vs2.
adam@347 618 cofix; intros; destruct c.
adam@347 619 rewrite (AssignCase H); constructor.
adam@347 620 destruct (SeqCase H) as [? [? ?]]; econstructor; eauto.
adam@478 621 destruct (WhileCase H) as [[? ?] | [? [? [? ?]]]]; subst; econstructor; eauto.
adam@347 622 Qed.
adam@347 623 End evalCmd_coind.
adam@347 624
adam@347 625 (** Now that we have a co-induction principle, we should use it to prove something! Our example is a trivial program optimizer that finds places to replace [0 + e] with [e]. *)
adam@347 626
adam@347 627 Fixpoint optExp (e : exp) : exp :=
adam@347 628 match e with
adam@347 629 | Plus (Const 0) e => optExp e
adam@347 630 | Plus e1 e2 => Plus (optExp e1) (optExp e2)
adam@347 631 | _ => e
adam@347 632 end.
adam@347 633
adam@347 634 Fixpoint optCmd (c : cmd) : cmd :=
adam@347 635 match c with
adam@347 636 | Assign v e => Assign v (optExp e)
adam@347 637 | Seq c1 c2 => Seq (optCmd c1) (optCmd c2)
adam@347 638 | While e c => While (optExp e) (optCmd c)
adam@347 639 end.
adam@347 640
adam@347 641 (** Before proving correctness of [optCmd], we prove a lemma about [optExp]. This is where we have to do the most work, choosing pattern match opportunities automatically. *)
adam@347 642
adam@347 643 (* begin thide *)
adam@347 644 Lemma optExp_correct : forall vs e, evalExp vs (optExp e) = evalExp vs e.
adam@347 645 induction e; crush;
adam@347 646 repeat (match goal with
adam@478 647 | [ |- context[match ?E with Const _ => _ | _ => _ end] ] => destruct E
adam@347 648 | [ |- context[match ?E with O => _ | S _ => _ end] ] => destruct E
adam@347 649 end; crush).
adamc@64 650 Qed.
adamc@64 651
adam@478 652 Hint Rewrite optExp_correct.
adamc@64 653
adam@384 654 (** The final theorem is easy to establish, using our co-induction principle and a bit of Ltac smarts that we leave unexplained for now. Curious readers can consult the Coq manual, or wait for the later chapters of this book about proof automation. At a high level, we show inclusions between behaviors, going in both directions between original and optimized programs. *)
adamc@64 655
adam@384 656 Ltac finisher := match goal with
adam@384 657 | [ H : evalCmd _ _ _ |- _ ] => ((inversion H; [])
adam@384 658 || (inversion H; [|])); subst
adam@384 659 end; crush; eauto 10.
adam@384 660
adam@384 661 Lemma optCmd_correct1 : forall vs1 c vs2, evalCmd vs1 c vs2
adam@347 662 -> evalCmd vs1 (optCmd c) vs2.
adam@347 663 intros; apply (evalCmd_coind (fun vs1 c' vs2 => exists c, evalCmd vs1 c vs2
adam@347 664 /\ c' = optCmd c)); eauto; crush;
adam@347 665 match goal with
adam@347 666 | [ H : _ = optCmd ?E |- _ ] => destruct E; simpl in *; discriminate
adam@347 667 || injection H; intros; subst
adam@384 668 end; finisher.
adam@384 669 Qed.
adam@384 670
adam@384 671 Lemma optCmd_correct2 : forall vs1 c vs2, evalCmd vs1 (optCmd c) vs2
adam@384 672 -> evalCmd vs1 c vs2.
adam@384 673 intros; apply (evalCmd_coind (fun vs1 c vs2 => evalCmd vs1 (optCmd c) vs2));
adam@384 674 crush; finisher.
adam@384 675 Qed.
adam@384 676
adam@384 677 Theorem optCmd_correct : forall vs1 c vs2, evalCmd vs1 (optCmd c) vs2
adam@384 678 <-> evalCmd vs1 c vs2.
adam@384 679 intuition; apply optCmd_correct1 || apply optCmd_correct2; assumption.
adam@347 680 Qed.
adam@347 681 (* end thide *)
adamc@64 682
adam@422 683 (** In this form, the theorem tells us that the optimizer preserves observable behavior of both terminating and nonterminating programs, but we did not have to do more work than for the case of terminating programs alone. We merely took the natural inductive definition for terminating executions, made it co-inductive, and applied the appropriate co-induction principle. Curious readers might experiment with adding command constructs like <<if>>; the same proof script should continue working, after the co-induction principle is extended to the new evaluation rules. *)