1 (* Title: Tools/quickcheck.ML
2 Author: Stefan Berghofer, Florian Haftmann, Lukas Bulwahn, TU Muenchen
4 Generic counterexample search engine.
9 val quickcheckN: string
13 val setup: theory -> theory
15 val auto: bool Unsynchronized.ref
16 val batch_tester : string Config.T
17 val size : int Config.T
18 val iterations : int Config.T
19 val depth : int Config.T
20 val no_assms : bool Config.T
21 val report : bool Config.T
22 val timeout : real Config.T
23 val timing : bool Config.T
24 val genuine_only : bool Config.T
25 val abort_potential : bool Config.T
26 val quiet : bool Config.T
27 val verbose : bool Config.T
28 val use_subtype : bool Config.T
29 val allow_function_inversion : bool Config.T
30 val finite_types : bool Config.T
31 val finite_type_size : int Config.T
32 val tag : string Config.T
33 val locale : string Config.T
34 val set_active_testers: string list -> Context.generic -> Context.generic
35 datatype expectation = No_Expectation | No_Counterexample | Counterexample;
36 datatype test_params = Test_Params of {default_type: typ list, expect : expectation};
37 val test_params_of : Proof.context -> test_params
38 val map_test_params : (typ list * expectation -> typ list * expectation)
39 -> Context.generic -> Context.generic
40 val default_type : Proof.context -> typ list
41 datatype report = Report of
42 { iterations : int, raised_match_errors : int,
43 satisfied_assms : int list, positive_concl_tests : int }
44 (* quickcheck's result *)
47 {counterexample : (bool * (string * term) list) option,
48 evaluation_terms : (term * term) list option,
49 timings : (string * int) list,
50 reports : (int * report) list}
51 val empty_result : result
52 val found_counterexample : result -> bool
53 val add_timing : (string * int) -> result Unsynchronized.ref -> unit
54 val add_response : string list -> term list -> (bool * term list) option -> result Unsynchronized.ref -> unit
55 val add_report : int -> report option -> result Unsynchronized.ref -> unit
56 val counterexample_of : result -> (bool * (string * term) list) option
57 val timings_of : result -> (string * int) list
58 (* registering testers & generators *)
60 Proof.context -> bool -> (string * typ) list -> (term * term list) list -> result list
61 val add_tester : string * (bool Config.T * tester) -> Context.generic -> Context.generic
62 val add_batch_generator :
63 string * (Proof.context -> term list -> (int -> term list option) list)
64 -> Context.generic -> Context.generic
65 val add_batch_validator :
66 string * (Proof.context -> term list -> (int -> bool) list)
67 -> Context.generic -> Context.generic
68 (* basic operations *)
69 val message : Proof.context -> string -> unit
70 val verbose_message : Proof.context -> string -> unit
71 val limit : Time.time -> (bool * bool) -> (unit -> 'a) -> (unit -> 'a) -> unit -> 'a
72 val pretty_counterex : Proof.context -> bool ->
73 ((bool * (string * term) list) * (term * term) list) option -> Pretty.T
74 (* testing terms and proof states *)
75 val mk_batch_validator : Proof.context -> term list -> (int -> bool) list option
76 val mk_batch_tester : Proof.context -> term list -> (int -> term list option) list option
77 val active_testers : Proof.context -> tester list
79 Proof.context -> bool * bool -> (string * typ) list -> (term * term list) list -> result list option
80 val quickcheck: (string * string list) list -> int -> Proof.state -> (bool * (string * term) list) option
83 structure Quickcheck : QUICKCHECK =
86 val quickcheckN = "quickcheck"
88 val genuineN = "genuine"
90 val unknownN = "unknown"
94 val auto = Unsynchronized.ref false;
97 ProofGeneralPgip.add_preference Preferences.category_tracing
98 (Unsynchronized.setmp auto true (fn () =>
99 Preferences.bool_pref auto
101 "Run Quickcheck automatically.") ());
103 (* quickcheck report *)
105 datatype report = Report of
106 { iterations : int, raised_match_errors : int,
107 satisfied_assms : int list, positive_concl_tests : int }
109 (* Quickcheck Result *)
111 datatype result = Result of
112 { counterexample : (bool * (string * term) list) option, evaluation_terms : (term * term) list option,
113 timings : (string * int) list, reports : (int * report) list}
116 Result {counterexample = NONE, evaluation_terms = NONE, timings = [], reports = []}
118 fun counterexample_of (Result r) = #counterexample r
120 fun found_counterexample (Result r) = is_some (#counterexample r)
122 fun response_of (Result r) = case (#counterexample r, #evaluation_terms r) of
123 (SOME ts, SOME evals) => SOME (ts, evals)
124 | (NONE, NONE) => NONE
126 fun timings_of (Result r) = #timings r
128 fun set_response names eval_terms (SOME (genuine, ts)) (Result r) =
130 val (ts1, ts2) = chop (length names) ts
131 val (eval_terms', _) = chop (length ts2) eval_terms
133 Result {counterexample = SOME (genuine, (names ~~ ts1)),
134 evaluation_terms = SOME (eval_terms' ~~ ts2),
135 timings = #timings r, reports = #reports r}
137 | set_response _ _ NONE result = result
140 fun cons_timing timing (Result r) =
141 Result {counterexample = #counterexample r, evaluation_terms = #evaluation_terms r,
142 timings = cons timing (#timings r), reports = #reports r}
144 fun cons_report size (SOME report) (Result r) =
145 Result {counterexample = #counterexample r, evaluation_terms = #evaluation_terms r,
146 timings = #timings r, reports = cons (size, report) (#reports r)}
147 | cons_report _ NONE result = result
149 fun add_timing timing result_ref =
150 Unsynchronized.change result_ref (cons_timing timing)
152 fun add_report size report result_ref =
153 Unsynchronized.change result_ref (cons_report size report)
155 fun add_response names eval_terms response result_ref =
156 Unsynchronized.change result_ref (set_response names eval_terms response)
160 datatype expectation = No_Expectation | No_Counterexample | Counterexample;
162 fun merge_expectation (expect1, expect2) =
163 if expect1 = expect2 then expect1 else No_Expectation
165 (* quickcheck configuration -- default parameters, test generators *)
166 val batch_tester = Attrib.setup_config_string @{binding quickcheck_batch_tester} (K "")
167 val size = Attrib.setup_config_int @{binding quickcheck_size} (K 10)
168 val iterations = Attrib.setup_config_int @{binding quickcheck_iterations} (K 100)
169 val depth = Attrib.setup_config_int @{binding quickcheck_depth} (K 10)
171 val no_assms = Attrib.setup_config_bool @{binding quickcheck_no_assms} (K false)
172 val locale = Attrib.setup_config_string @{binding quickcheck_locale} (K "interpret expand")
173 val report = Attrib.setup_config_bool @{binding quickcheck_report} (K true)
174 val timing = Attrib.setup_config_bool @{binding quickcheck_timing} (K false)
175 val timeout = Attrib.setup_config_real @{binding quickcheck_timeout} (K 30.0)
177 val genuine_only = Attrib.setup_config_bool @{binding quickcheck_genuine_only} (K false)
178 val abort_potential = Attrib.setup_config_bool @{binding quickcheck_abort_potential} (K false)
180 val quiet = Attrib.setup_config_bool @{binding quickcheck_quiet} (K false)
181 val verbose = Attrib.setup_config_bool @{binding quickcheck_verbose} (K false)
182 val tag = Attrib.setup_config_string @{binding quickcheck_tag} (K "")
184 val use_subtype = Attrib.setup_config_bool @{binding quickcheck_use_subtype} (K false)
186 val allow_function_inversion =
187 Attrib.setup_config_bool @{binding quickcheck_allow_function_inversion} (K false)
188 val finite_types = Attrib.setup_config_bool @{binding quickcheck_finite_types} (K true)
189 val finite_type_size = Attrib.setup_config_int @{binding quickcheck_finite_type_size} (K 3)
191 datatype test_params = Test_Params of
192 {default_type: typ list, expect : expectation};
194 fun dest_test_params (Test_Params {default_type, expect}) = (default_type, expect);
196 fun make_test_params (default_type, expect) =
197 Test_Params {default_type = default_type, expect = expect};
199 fun map_test_params' f (Test_Params {default_type, expect}) =
200 make_test_params (f (default_type, expect));
202 fun merge_test_params
203 (Test_Params {default_type = default_type1, expect = expect1},
204 Test_Params {default_type = default_type2, expect = expect2}) =
206 (merge (op =) (default_type1, default_type2), merge_expectation (expect1, expect2));
209 Proof.context -> bool -> (string * typ) list -> (term * term list) list -> result list
211 structure Data = Generic_Data
214 ((string * (bool Config.T * tester)) list
215 * ((string * (Proof.context -> term list -> (int -> term list option) list)) list
216 * ((string * (Proof.context -> term list -> (int -> bool) list)) list)))
218 val empty = (([], ([], [])), Test_Params {default_type = [], expect = No_Expectation});
220 fun merge (((testers1, (batch_generators1, batch_validators1)), params1),
221 ((testers2, (batch_generators2, batch_validators2)), params2)) : T =
222 ((AList.merge (op =) (K true) (testers1, testers2),
223 (AList.merge (op =) (K true) (batch_generators1, batch_generators2),
224 AList.merge (op =) (K true) (batch_validators1, batch_validators2))),
225 merge_test_params (params1, params2));
228 val test_params_of = snd o Data.get o Context.Proof;
230 val default_type = fst o dest_test_params o test_params_of
232 val expect = snd o dest_test_params o test_params_of
234 val map_test_params = Data.map o apsnd o map_test_params'
236 val add_tester = Data.map o apfst o apfst o AList.update (op =);
238 val add_batch_generator = Data.map o apfst o apsnd o apfst o AList.update (op =);
240 val add_batch_validator = Data.map o apfst o apsnd o apsnd o AList.update (op =);
242 fun active_testers ctxt =
244 val testers = (map snd o fst o fst o Data.get o Context.Proof) ctxt
246 map snd (filter (fn (active, _) => Config.get ctxt active) testers)
249 fun set_active_testers [] gen_ctxt = gen_ctxt
250 | set_active_testers testers gen_ctxt =
252 val registered_testers = (fst o fst o Data.get) gen_ctxt
254 fold (fn (name, (config, _)) => Config.put_generic config (member (op =) testers name))
255 registered_testers gen_ctxt
258 (* generating tests *)
260 fun gen_mk_tester lookup ctxt v =
262 val name = Config.get ctxt batch_tester
263 val tester = case lookup ctxt name
264 of NONE => error ("No such quickcheck batch-tester: " ^ name)
265 | SOME tester => tester ctxt;
267 if Config.get ctxt quiet then
271 val tester = Exn.interruptible_capture tester v
272 in case Exn.get_res tester of
273 NONE => SOME (Exn.release tester)
274 | SOME tester => SOME tester
278 val mk_batch_tester =
279 gen_mk_tester (fn ctxt => AList.lookup (op =) ((fst o snd o fst o Data.get o Context.Proof) ctxt))
280 val mk_batch_validator =
281 gen_mk_tester (fn ctxt => AList.lookup (op =) ((snd o snd o fst o Data.get o Context.Proof) ctxt))
283 (* testing propositions *)
285 type compile_generator =
286 Proof.context -> (term * term list) list -> int list -> term list option * report option
288 fun limit timeout (limit_time, is_interactive) f exc () =
290 TimeLimit.timeLimit timeout f ()
291 handle TimeLimit.TimeOut =>
292 if is_interactive then exc () else raise TimeLimit.TimeOut
296 fun message ctxt s = if Config.get ctxt quiet then () else Output.urgent_message s
298 fun verbose_message ctxt s = if not (Config.get ctxt quiet) andalso Config.get ctxt verbose
299 then Output.urgent_message s else ()
301 fun test_terms ctxt (limit_time, is_interactive) insts goals =
302 case active_testers ctxt of
303 [] => error "No active testers for quickcheck"
305 limit (seconds (Config.get ctxt timeout)) (limit_time, is_interactive)
306 (fn () => Par_List.get_some (fn tester =>
307 tester ctxt (length testers > 1) insts goals |>
308 (fn result => if exists found_counterexample result then SOME result else NONE)) testers)
309 (fn () => (message ctxt "Quickcheck ran out of time"; NONE)) ();
311 fun all_axioms_of ctxt t =
313 val intros = Locale.get_intros ctxt
314 val unfolds = Locale.get_unfolds ctxt
315 fun retrieve_prems thms t =
316 case filter (fn th => Term.could_unify (Thm.concl_of th, t)) thms of
321 Pattern.match (Proof_Context.theory_of ctxt) (Thm.concl_of th, t) (Vartab.empty, Vartab.empty)
322 in SOME (map (Envir.subst_term (tyenv, tenv)) (Thm.prems_of th)) end
324 case retrieve_prems intros t of
325 NONE => retrieve_prems unfolds t
326 | SOME ts => SOME (maps (fn t => the_default [t] (all t)) ts)
331 fun locale_config_of s =
333 val cs = space_explode " " s
335 if forall (fn c => c = "expand" orelse c = "interpret") cs then cs
336 else (warning ("Invalid quickcheck_locale setting: falling back to the default setting.");
337 ["interpret", "expand"])
340 fun test_goal (time_limit, is_interactive) (insts, eval_terms) i state =
342 val lthy = Proof.context_of state;
343 val thy = Proof.theory_of state;
344 val _ = message lthy "Quickchecking..."
345 fun strip (Const ("all", _) $ Abs (_, _, t)) = strip t
347 val {goal = st, ...} = Proof.raw_goal state;
348 val (gi, frees) = Logic.goal_params (prop_of st) i;
349 val some_locale = case (Option.map #target o Named_Target.peek) lthy
352 | SOME locale => SOME locale;
353 val assms = if Config.get lthy no_assms then [] else case some_locale
354 of NONE => Assumption.all_assms_of lthy
355 | SOME locale => Assumption.local_assms_of lthy (Locale.init locale thy);
356 val proto_goal = Logic.list_implies (map Thm.term_of assms, subst_bounds (frees, strip gi));
357 fun axioms_of locale = case fst (Locale.specification_of thy locale) of
359 | SOME t => the_default [] (all_axioms_of lthy t)
360 val config = locale_config_of (Config.get lthy locale)
361 val goals = case some_locale
362 of NONE => [(proto_goal, eval_terms)]
363 | SOME locale => fold (fn c =>
364 if c = "expand" then cons (Logic.list_implies (axioms_of locale, proto_goal), eval_terms) else
365 if c = "interpret" then
366 append (map (fn (_, phi) => (Morphism.term phi proto_goal, map (Morphism.term phi) eval_terms))
367 (Locale.registrations_of (Context.Theory thy) (*FIXME*) locale)) else I) config [];
368 val _ = verbose_message lthy (Pretty.string_of
369 (Pretty.big_list ("Checking goals: ") (map (Syntax.pretty_term lthy o fst) goals)))
371 test_terms lthy (time_limit, is_interactive) insts goals
374 (* pretty printing *)
376 fun tool_name auto = (if auto then "Auto " else "") ^ "Quickcheck"
378 fun pretty_counterex ctxt auto NONE =
379 Pretty.str (tool_name auto ^ " found no counterexample." ^ Config.get ctxt tag)
380 | pretty_counterex ctxt auto (SOME ((genuine, cex), eval_terms)) =
381 Pretty.chunks ((Pretty.str (tool_name auto ^ " found a " ^
382 (if genuine then "counterexample:"
383 else "potentially spurious counterexample due to underspecified functions:")
384 ^ Config.get ctxt tag ^ "\n") ::
386 Pretty.block [Pretty.str (s ^ " ="), Pretty.brk 1, Syntax.pretty_term ctxt t]) (rev cex))
387 @ (if null eval_terms then []
388 else (Pretty.fbrk :: Pretty.str ("Evaluated terms:") ::
390 Pretty.block [Syntax.pretty_term ctxt t, Pretty.str " =", Pretty.brk 1,
391 Syntax.pretty_term ctxt u]) (rev eval_terms))));
393 fun pretty_report (Report {iterations = iterations, raised_match_errors = raised_match_errors,
394 satisfied_assms = satisfied_assms, positive_concl_tests = positive_concl_tests}) =
396 fun pretty_stat s i = Pretty.block ([Pretty.str (s ^ ": " ^ string_of_int i)])
398 ([pretty_stat "iterations" iterations,
399 pretty_stat "match exceptions" raised_match_errors]
401 (fn (i, n) => pretty_stat ("satisfied " ^ string_of_int (i + 1) ^ ". assumption") n)
403 @ [pretty_stat "positive conclusion tests" positive_concl_tests])
406 fun pretty_timings timings =
407 Pretty.chunks (Pretty.fbrk :: Pretty.str "timings:" ::
408 maps (fn (label, time) =>
409 Pretty.str (label ^ ": " ^ string_of_int time ^ " ms") :: Pretty.brk 1 :: []) (rev timings))
413 fun read_nat s = case (Library.read_int o Symbol.explode) s
414 of (k, []) => if k >= 0 then k
415 else error ("Not a natural number: " ^ s)
416 | (_, _ :: _) => error ("Not a natural number: " ^ s);
418 fun read_bool "false" = false
419 | read_bool "true" = true
420 | read_bool s = error ("Not a Boolean value: " ^ s)
423 case (Real.fromString s) of
425 | NONE => error ("Not a real number: " ^ s)
427 fun read_expectation "no_expectation" = No_Expectation
428 | read_expectation "no_counterexample" = No_Counterexample
429 | read_expectation "counterexample" = Counterexample
430 | read_expectation s = error ("Not an expectation value: " ^ s)
432 fun valid_tester_name genctxt name = AList.defined (op =) (fst (fst (Data.get genctxt))) name
434 fun parse_tester name (testers, genctxt) =
435 if valid_tester_name genctxt name then
436 (insert (op =) name testers, genctxt)
438 error ("Unknown tester: " ^ name)
440 fun parse_test_param ("tester", args) = fold parse_tester args
441 | parse_test_param ("size", [arg]) = apsnd (Config.put_generic size (read_nat arg))
442 | parse_test_param ("iterations", [arg]) = apsnd (Config.put_generic iterations (read_nat arg))
443 | parse_test_param ("depth", [arg]) = apsnd (Config.put_generic depth (read_nat arg))
444 | parse_test_param ("default_type", arg) = (fn (testers, gen_ctxt) =>
445 (testers, map_test_params
446 ((apfst o K) (map (Proof_Context.read_typ (Context.proof_of gen_ctxt)) arg)) gen_ctxt))
447 | parse_test_param ("no_assms", [arg]) = apsnd (Config.put_generic no_assms (read_bool arg))
448 | parse_test_param ("expect", [arg]) = apsnd (map_test_params ((apsnd o K) (read_expectation arg)))
449 | parse_test_param ("report", [arg]) = apsnd (Config.put_generic report (read_bool arg))
450 | parse_test_param ("genuine_only", [arg]) = apsnd (Config.put_generic genuine_only (read_bool arg))
451 | parse_test_param ("abort_potential", [arg]) = apsnd (Config.put_generic abort_potential (read_bool arg))
452 | parse_test_param ("quiet", [arg]) = apsnd (Config.put_generic quiet (read_bool arg))
453 | parse_test_param ("verbose", [arg]) = apsnd (Config.put_generic verbose (read_bool arg))
454 | parse_test_param ("tag", [arg]) = apsnd (Config.put_generic tag arg)
455 | parse_test_param ("use_subtype", [arg]) = apsnd (Config.put_generic use_subtype (read_bool arg))
456 | parse_test_param ("timeout", [arg]) = apsnd (Config.put_generic timeout (read_real arg))
457 | parse_test_param ("finite_types", [arg]) = apsnd (Config.put_generic finite_types (read_bool arg))
458 | parse_test_param ("allow_function_inversion", [arg]) =
459 apsnd (Config.put_generic allow_function_inversion (read_bool arg))
460 | parse_test_param ("finite_type_size", [arg]) =
461 apsnd (Config.put_generic finite_type_size (read_nat arg))
462 | parse_test_param (name, _) = (fn (testers, genctxt) =>
463 if valid_tester_name genctxt name then
464 (insert (op =) name testers, genctxt)
465 else error ("Unknown tester or test parameter: " ^ name));
467 fun proof_map_result f = apsnd Context.the_proof o f o Context.Proof;
469 fun parse_test_param_inst (name, arg) ((insts, eval_terms), (testers, ctxt)) =
470 case try (Proof_Context.read_typ ctxt) name
471 of SOME (TFree (v, _)) =>
472 ((AList.update (op =) (v, Proof_Context.read_typ ctxt (the_single arg)) insts, eval_terms),
474 | NONE => (case name of
475 "eval" => ((insts, eval_terms @ map (Syntax.read_term ctxt) arg), (testers, ctxt))
476 | _ => ((insts, eval_terms),
477 proof_map_result (fn gen_ctxt => parse_test_param (name, arg) (testers, gen_ctxt)) ctxt));
479 fun quickcheck_params_cmd args = Context.theory_map
480 (fn gen_ctxt => uncurry set_active_testers (fold parse_test_param args ([], gen_ctxt)));
482 fun check_expectation state results =
483 (if is_some results andalso expect (Proof.context_of state) = No_Counterexample
485 error ("quickcheck expected to find no counterexample but found one")
487 (if is_none results andalso expect (Proof.context_of state) = Counterexample
489 error ("quickcheck expected to find a counterexample but did not find one")
492 fun gen_quickcheck args i state =
494 |> Proof.map_context_result (fn ctxt =>
495 apsnd (fn (testers, ctxt) => Context.proof_map (set_active_testers testers) ctxt)
496 (fold parse_test_param_inst args (([], []), ([], ctxt))))
497 |> (fn ((insts, eval_terms), state') => test_goal (true, true) (insts, eval_terms) i state'
498 |> tap (check_expectation state') |> rpair state')
500 fun quickcheck args i state =
501 Option.map (the o get_first counterexample_of) (fst (gen_quickcheck args i state))
503 fun quickcheck_cmd args i state =
504 gen_quickcheck args i (Toplevel.proof_of state)
505 |> apfst (Option.map (the o get_first response_of))
506 |> (fn (r, state) => Output.urgent_message (Pretty.string_of
507 (pretty_counterex (Proof.context_of state) false r)));
511 (Scan.optional (@{keyword "="} |--
512 (((Parse.name || Parse.float_number) >> single) ||
513 (@{keyword "["} |-- Parse.list1 Parse.name --| @{keyword "]"}))) ["true"]);
516 @{keyword "["} |-- Parse.list1 parse_arg --| @{keyword "]"} || Scan.succeed [];
519 Outer_Syntax.command @{command_spec "quickcheck_params"} "set parameters for random testing"
520 (parse_args >> (fn args => Toplevel.theory (quickcheck_params_cmd args)));
523 Outer_Syntax.improper_command @{command_spec "quickcheck"}
524 "try to find counterexample for subgoal"
525 (parse_args -- Scan.optional Parse.nat 1
526 >> (fn (args, i) => Toplevel.no_timing o Toplevel.keep (quickcheck_cmd args i)));
528 (* automatic testing *)
530 fun try_quickcheck auto state =
532 val ctxt = Proof.context_of state;
536 |> Proof.map_context (Config.put report false #> Config.put quiet true)
537 |> try (test_goal (false, false) ([], []) i);
540 NONE => (unknownN, state)
543 val msg = pretty_counterex ctxt auto (Option.map (the o get_first response_of) results)
545 if is_some results then
549 Proof.goal_message (K (Pretty.chunks [Pretty.str "",
550 Pretty.mark Isabelle_Markup.hilite msg]))
552 tap (fn _ => Output.urgent_message (Pretty.string_of msg))))
557 |> `(fn (outcome_code, _) => outcome_code = genuineN)
559 val setup = Try.register_tool (quickcheckN, (20, auto, try_quickcheck))
563 val auto_quickcheck = Quickcheck.auto;