(* ========================================================================= *) (* Load in Petros Papapanagiotou's Boyer-Moore code and try examples. *) (* ========================================================================= *) loads "Boyer_Moore/boyer-moore.ml";; (* ------------------------------------------------------------------------- *) (* Slight variant of Petros's eval.ml file. *) (* ------------------------------------------------------------------------- *) (* ========================================================================= *) (* ------------------------------------------------------------------------- *) (* Shortcuts for the various evaluation versions: *) (* ------------------------------------------------------------------------- *) let BM = BOYER_MOORE;; (* Pure re-implementation of R.Boulton's work. *) let BME = BOYER_MOORE_EXT;; (* Extended with early termination heuristics and HOL Light features. *) let BMR = BOYER_MOORE_RE [];; let BMG = BOYER_MOORE_GEN [];; (* Further extended with M.Aderhold's generalization techniques. *) let BMF = BOYER_MOORE_FINAL [];; let RBM = new_rewrite_rule o BOYER_MOORE;; let RBME = new_rewrite_rule o BOYER_MOORE_EXT;; let RBMG = new_rewrite_rule o BOYER_MOORE_GEN [];; (* ------------------------------------------------------------------------- *) (* Add a theorem as a new function definition and rewrite rule. *) (* Adding it as a rewrite rule should no longer be necessary after the *) (* latest (July 2009) bugfixes but it doesn't do any harm either. *) (* ------------------------------------------------------------------------- *) let new_stuff x = (new_def x ; new_rewrite_rule x);; (* ------------------------------------------------------------------------- *) (* Test sets extracted from the proven theorems in HOL Light's arith.ml and *) (* list.ml. *) (* ------------------------------------------------------------------------- *) loads "Boyer_Moore/testset/arith.ml";; (* Arithmetic test set *) loads "Boyer_Moore/testset/list.ml";; (* List test set *) (* ------------------------------------------------------------------------- *) (* Reloads all the necessary definitions and rules for the evaluation of the *) (* test sets above. *) (* ------------------------------------------------------------------------- *) let bm_reset () = system_defs := []; system_rewrites := []; new_stuff ADD; new_stuff MULT; new_stuff SUB; new_stuff LE; new_stuff LT; new_stuff GE; new_stuff GT; new_rewrite_rule (ARITH_RULE `1=SUC(0)`); new_stuff EXP; new_stuff FACT; new_stuff ODD; new_stuff EVEN; new_rewrite_rule NOT_SUC; new_rewrite_rule SUC_INJ; new_rewrite_rule PRE; new_rewrite_rule (prove (`!n. ~(SUC n = n)`, INDUCT_TAC THEN ASM_REWRITE_TAC[SUC_INJ;NOT_SUC])); new_rewrite_rule (prove (`!a b. a + SUC b = SUC (a + b)`,REPEAT GEN_TAC THEN BMF_TAC[])); new_stuff HD; new_stuff TL; new_stuff APPEND; new_stuff REVERSE; new_stuff LENGTH; new_stuff MAP; new_stuff LAST; new_stuff REPLICATE; new_stuff NULL; new_stuff ALL; new_stuff EX; new_stuff ITLIST; new_stuff MEM; new_stuff ALL2_DEF; new_rewrite_rule ALL2; new_stuff MAP2_DEF; new_rewrite_rule MAP2; new_stuff EL; new_stuff FILTER; new_stuff ASSOC; new_stuff ITLIST2_DEF; new_rewrite_rule ITLIST2; new_stuff ZIP_DEF; new_rewrite_rule ZIP; new_rewrite_rule NOT_CONS_NIL; new_rewrite_rule CONS_11 ;; bm_reset();; (* ------------------------------------------------------------------------- *) (* Test functions. They use the Unix library to measure time. *) (* Unfortunately this means that they do not load properly in Cygwin. *) (* ------------------------------------------------------------------------- *) #load "unix.cma";; open Unix;; open Printf;; (* ------------------------------------------------------------------------- *) (* Reference of the remaining theory to be proven. Load a list of theorems *) (* that you want the evaluation to run through. *) (* eg. remaining_theory := !mytheory;; *) (* Then use nexttm (see below) to evaluate one of the BOYER_MOORE_* *) (* procedures over the list. *) (* ------------------------------------------------------------------------- *) let remaining_theory = ref ([]:term list);; let currenttm = ref `p`;; (* ------------------------------------------------------------------------- *) (* Tries a theorem-proving procedure f on arg. *) (* Returns a truth value of whether the procedure succeeded in finding a *) (* proof and a pair of timestamps taken from the start and the end of the *) (* procedure. *) (* ------------------------------------------------------------------------- *) let bm_time f arg = let t1=Unix.times () in let resu = try (if (can dest_thm (f arg)) then true else false) with Failure _ -> false in let t2=Unix.times () in (resu,(t1,t2));; (* printf "User time: %f - system time: %f\n%!" (t2.tms_utime -. t1.tms_utime) (t2.tms_stime -. t1.tms_stime);; *) (* ------------------------------------------------------------------------- *) (* Uses bm_time to try a Boyer-Moore theorem-proving procedure f on tm. *) (* Prints out all the evaluation information that is collected and returns *) (* the list of generalizations made during the proof. *) (* ------------------------------------------------------------------------- *) let bm_test f tm = let pfpt = (print_term tm ; print_newline() ; proof_printer false) in let (resu,(t1,t2)) = bm_time f tm in let pfpt = proof_printer pfpt in printf "Proven: %b - Time: %f - Steps: %d - Inductions: %d - Gen terms: %d - Over gens: %d \\\\\n" resu (t2.tms_utime -. t1.tms_utime) (fst !bm_steps) (snd !bm_steps) (length !my_gen_terms) (!counterexamples) ; !my_gen_terms;; (* ------------------------------------------------------------------------- *) (* Another version of bm_test but with a more compact printout. *) (* Returns unit (). *) (* ------------------------------------------------------------------------- *) let bm_test2 f tm = let pfpt = (print_term tm ; print_newline() ; proof_printer false) in let (resu,(t1,t2)) = bm_time f tm in let pfpt = proof_printer pfpt in printf "& %b & %f & %d & %d & %d & %d \\\\\n" resu (t2.tms_utime -. t1.tms_utime) (fst !bm_steps) (snd !bm_steps) (length !my_gen_terms) (!counterexamples) ; ();; (* ------------------------------------------------------------------------- *) (* Convenient function for evaluation. *) (* Uses f to try and prove the next term in !remaining_theory by bm_test2 *) (* ------------------------------------------------------------------------- *) let nexttm f = if (!remaining_theory = []) then failwith "No more" else currenttm := hd !remaining_theory ; remaining_theory := tl !remaining_theory ; bm_test2 f !currenttm;; (* ------------------------------------------------------------------------- *) (* Reruns evaluation on the same term that was last loaded with nexttm. *) (* ------------------------------------------------------------------------- *) let sametm f = bm_test2 f !currenttm;; (* ========================================================================= *) (* ------------------------------------------------------------------------- *) (* Just one example. *) (* ------------------------------------------------------------------------- *) bm_test BME `m + n:num = n + m`;; (* ------------------------------------------------------------------------- *) (* Note that these don't all terminate, so need more delicacy really. *) (* Should carefully reconstruct the cases in Petros's thesis, also maybe *) (* using a timeout. *) (* ------------------------------------------------------------------------- *) (**** do_list (bm_test BME) (!mytheory);; do_list (bm_test BME) (!mytheory2);; ****)