wenzelm@18120: (*  Title:      Pure/old_goals.ML
wenzelm@18120:     ID:         $Id$
wenzelm@18120:     Author:     Lawrence C Paulson, Cambridge University Computer Laboratory
wenzelm@18120:     Copyright   1993  University of Cambridge
wenzelm@18120: 
wenzelm@18120: Old-style goal stack package.  The goal stack initially holds a dummy
wenzelm@18120: proof, and can never become empty.  Each goal stack consists of a list
wenzelm@18120: of levels.  The undo list is a list of goal stacks.  Finally, there
wenzelm@18120: may be a stack of pending proofs.
wenzelm@18120: *)
wenzelm@18120: 
wenzelm@18120: signature GOALS =
wenzelm@18120: sig
wenzelm@26429:   val the_context: unit -> theory
wenzelm@18120:   val premises: unit -> thm list
wenzelm@18120:   val prove_goal: theory -> string -> (thm list -> tactic list) -> thm
wenzelm@18120:   val prove_goalw: theory -> thm list -> string -> (thm list -> tactic list) -> thm
wenzelm@18120:   val topthm: unit -> thm
wenzelm@18120:   val result: unit -> thm
wenzelm@18120:   val uresult: unit -> thm
wenzelm@18120:   val getgoal: int -> term
wenzelm@18120:   val gethyps: int -> thm list
wenzelm@18120:   val prlev: int -> unit
wenzelm@18120:   val pr: unit -> unit
wenzelm@18120:   val prlim: int -> unit
wenzelm@18120:   val goal: theory -> string -> thm list
wenzelm@18120:   val goalw: theory -> thm list -> string -> thm list
wenzelm@18120:   val Goal: string -> thm list
wenzelm@18120:   val Goalw: thm list -> string -> thm list
wenzelm@18120:   val by: tactic -> unit
wenzelm@18120:   val back: unit -> unit
wenzelm@18120:   val choplev: int -> unit
wenzelm@18120:   val undo: unit -> unit
wenzelm@18120:   val qed: string -> unit
wenzelm@18120:   val qed_goal: string -> theory -> string -> (thm list -> tactic list) -> unit
wenzelm@18120:   val qed_goalw: string -> theory -> thm list -> string
wenzelm@18120:     -> (thm list -> tactic list) -> unit
wenzelm@18120:   val qed_spec_mp: string -> unit
wenzelm@18120:   val qed_goal_spec_mp: string -> theory -> string -> (thm list -> tactic list) -> unit
wenzelm@18120:   val qed_goalw_spec_mp: string -> theory -> thm list -> string
wenzelm@18120:     -> (thm list -> tactic list) -> unit
wenzelm@18120: end;
wenzelm@18120: 
wenzelm@18120: signature OLD_GOALS =
wenzelm@18120: sig
wenzelm@18120:   include GOALS
wenzelm@27256:   val simple_read_term: theory -> typ -> string -> term
wenzelm@27256:   val read_term: theory -> string -> term
wenzelm@27256:   val read_prop: theory -> string -> term
wenzelm@18120:   type proof
wenzelm@19053:   val chop: unit -> unit
wenzelm@18120:   val reset_goals: unit -> unit
wenzelm@18120:   val result_error_fn: (thm -> string -> thm) ref
wenzelm@18120:   val print_sign_exn: theory -> exn -> 'a
wenzelm@18120:   val prove_goalw_cterm: thm list->cterm->(thm list->tactic list)->thm
wenzelm@18120:   val prove_goalw_cterm_nocheck: thm list->cterm->(thm list->tactic list)->thm
wenzelm@18120:   val print_exn: exn -> 'a
wenzelm@18120:   val filter_goal: (term*term->bool) -> thm list -> int -> thm list
wenzelm@18120:   val goalw_cterm: thm list -> cterm -> thm list
wenzelm@18120:   val simple_prove_goal_cterm: cterm->(thm list->tactic list)->thm
wenzelm@18120:   val byev: tactic list -> unit
wenzelm@18120:   val save_proof: unit -> proof
wenzelm@18120:   val restore_proof: proof -> thm list
wenzelm@18120:   val push_proof: unit -> unit
wenzelm@18120:   val pop_proof: unit -> thm list
wenzelm@18120:   val rotate_proof: unit -> thm list
wenzelm@18120: end;
wenzelm@18120: 
wenzelm@18120: structure OldGoals: OLD_GOALS =
wenzelm@18120: struct
wenzelm@18120: 
wenzelm@27256: (* global context *)
wenzelm@27256: 
wenzelm@26429: val the_context = ML_Context.the_global_context;
wenzelm@26429: 
wenzelm@26429: 
wenzelm@27256: (* old ways of reading terms *)
wenzelm@27256: 
wenzelm@27256: fun simple_read_term thy T s =
wenzelm@27256:   let
wenzelm@27256:     val ctxt = ProofContext.init thy
wenzelm@27256:       |> ProofContext.allow_dummies
wenzelm@27256:       |> ProofContext.set_mode ProofContext.mode_schematic;
wenzelm@27256:     val parse = if T = propT then Syntax.parse_prop else Syntax.parse_term;
wenzelm@27256:   in parse ctxt s |> TypeInfer.constrain T |> Syntax.check_term ctxt end;
wenzelm@27256: 
wenzelm@27256: fun read_term thy = simple_read_term thy dummyT;
wenzelm@27256: fun read_prop thy = simple_read_term thy propT;
wenzelm@27256: 
wenzelm@27256: 
wenzelm@27256: 
wenzelm@18120: (*** Goal package ***)
wenzelm@18120: 
wenzelm@18120: (*Each level of goal stack includes a proof state and alternative states,
wenzelm@18120:   the output of the tactic applied to the preceeding level.  *)
wenzelm@18120: type gstack = (thm * thm Seq.seq) list;
wenzelm@18120: 
wenzelm@18120: datatype proof = Proof of gstack list * thm list * (bool*thm->thm);
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*** References ***)
wenzelm@18120: 
wenzelm@18120: (*Current assumption list -- set by "goal".*)
wenzelm@18120: val curr_prems = ref([] : thm list);
wenzelm@18120: 
wenzelm@18120: (*Return assumption list -- useful if you didn't save "goal"'s result. *)
wenzelm@18120: fun premises() = !curr_prems;
wenzelm@18120: 
wenzelm@18120: (*Current result maker -- set by "goal", used by "result".  *)
wenzelm@18120: fun init_mkresult _ = error "No goal has been supplied in subgoal module";
wenzelm@18120: val curr_mkresult = ref (init_mkresult: bool*thm->thm);
wenzelm@18120: 
wenzelm@18120: (*List of previous goal stacks, for the undo operation.  Set by setstate.
wenzelm@18120:   A list of lists!*)
wenzelm@26429: val undo_list = ref([[(asm_rl, Seq.empty)]] : gstack list);
wenzelm@18120: 
wenzelm@18120: (* Stack of proof attempts *)
wenzelm@18120: val proofstack = ref([]: proof list);
wenzelm@18120: 
wenzelm@18120: (*reset all refs*)
wenzelm@18120: fun reset_goals () =
wenzelm@18120:   (curr_prems := []; curr_mkresult := init_mkresult;
wenzelm@26429:     undo_list := [[(asm_rl, Seq.empty)]]);
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*** Setting up goal-directed proof ***)
wenzelm@18120: 
wenzelm@18120: (*Generates the list of new theories when the proof state's theory changes*)
wenzelm@18120: fun thy_error (thy,thy') =
wenzelm@29091:   let val names = Context.display_names thy' \\ Context.display_names thy
wenzelm@18120:   in  case names of
wenzelm@18120:         [name] => "\nNew theory: " ^ name
wenzelm@18120:       | _       => "\nNew theories: " ^ space_implode ", " names
wenzelm@18120:   end;
wenzelm@18120: 
wenzelm@18120: (*Default action is to print an error message; could be suppressed for
wenzelm@18120:   special applications.*)
wenzelm@18120: fun result_error_default state msg : thm =
wenzelm@18120:   Pretty.str "Bad final proof state:" :: Display.pretty_goals (!goals_limit) state @
wenzelm@18120:     [Pretty.str msg, Pretty.str "Proof failed!"] |> Pretty.chunks |> Pretty.string_of |> error;
wenzelm@18120: 
wenzelm@18120: val result_error_fn = ref result_error_default;
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*Common treatment of "goal" and "prove_goal":
wenzelm@18120:   Return assumptions, initial proof state, and function to make result.
wenzelm@18120:   "atomic" indicates if the goal should be wrapped up in the function
wenzelm@18120:   "Goal::prop=>prop" to avoid assumptions being returned separately.
wenzelm@18120: *)
wenzelm@18120: fun prepare_proof atomic rths chorn =
wenzelm@18120:   let
wenzelm@26291:       val _ = legacy_feature "Old goal command";
wenzelm@26626:       val thy = Thm.theory_of_cterm chorn;
wenzelm@26626:       val horn = Thm.term_of chorn;
wenzelm@18120:       val _ = Term.no_dummy_patterns horn handle TERM (msg, _) => error msg;
wenzelm@18120:       val (As, B) = Logic.strip_horn horn;
wenzelm@18120:       val atoms = atomic andalso
wenzelm@27332:             forall (fn t => not (can Logic.dest_implies t orelse Logic.is_all t)) As;
wenzelm@18120:       val (As,B) = if atoms then ([],horn) else (As,B);
wenzelm@18120:       val cAs = map (cterm_of thy) As;
wenzelm@26653:       val prems = map (rewrite_rule rths o Thm.forall_elim_vars 0 o Thm.assume) cAs;
wenzelm@18120:       val cB = cterm_of thy B;
wenzelm@18120:       val st0 = let val st = Goal.init cB |> fold Thm.weaken cAs
wenzelm@18120:                 in  rewrite_goals_rule rths st end
wenzelm@18120:       (*discharges assumptions from state in the order they appear in goal;
wenzelm@18120:         checks (if requested) that resulting theorem is equivalent to goal. *)
wenzelm@18120:       fun mkresult (check,state) =
wenzelm@18120:         let val state = Seq.hd (flexflex_rule state)
wenzelm@18120:                         handle THM _ => state   (*smash flexflex pairs*)
wenzelm@18120:             val ngoals = nprems_of state
wenzelm@18120:             val ath = implies_intr_list cAs state
wenzelm@18120:             val th = Goal.conclude ath
wenzelm@26626:             val thy' = Thm.theory_of_thm th
wenzelm@26626:             val {hyps, prop, ...} = Thm.rep_thm th
wenzelm@18120:             val final_th = standard th
wenzelm@18120:         in  if not check then final_th
wenzelm@26665:             else if not (Theory.eq_thy(thy,thy')) then !result_error_fn state
wenzelm@18120:                 ("Theory of proof state has changed!" ^
wenzelm@18120:                  thy_error (thy,thy'))
wenzelm@18120:             else if ngoals>0 then !result_error_fn state
wenzelm@18120:                 (string_of_int ngoals ^ " unsolved goals!")
wenzelm@18120:             else if not (null hyps) then !result_error_fn state
wenzelm@18120:                 ("Additional hypotheses:\n" ^
wenzelm@26939:                  cat_lines (map (Syntax.string_of_term_global thy) hyps))
wenzelm@18120:             else if Pattern.matches thy
wenzelm@18120:                                     (Envir.beta_norm (term_of chorn), Envir.beta_norm prop)
wenzelm@18120:                  then final_th
wenzelm@18120:             else  !result_error_fn state "proved a different theorem"
wenzelm@18120:         end
wenzelm@18120:   in
wenzelm@26665:      if Theory.eq_thy(thy, Thm.theory_of_thm st0)
wenzelm@18120:      then (prems, st0, mkresult)
wenzelm@18120:      else error ("Definitions would change the proof state's theory" ^
wenzelm@18120:                  thy_error (thy, Thm.theory_of_thm st0))
wenzelm@18120:   end
wenzelm@18120:   handle THM(s,_,_) => error("prepare_proof: exception THM was raised!\n" ^ s);
wenzelm@18120: 
wenzelm@18120: (*Prints exceptions readably to users*)
wenzelm@18120: fun print_sign_exn_unit thy e =
wenzelm@18120:   case e of
wenzelm@18120:      THM (msg,i,thms) =>
wenzelm@18120:          (writeln ("Exception THM " ^ string_of_int i ^ " raised:\n" ^ msg);
wenzelm@26928:           List.app Display.print_thm thms)
wenzelm@18120:    | THEORY (msg,thys) =>
wenzelm@18120:          (writeln ("Exception THEORY raised:\n" ^ msg);
wenzelm@18120:           List.app (writeln o Context.str_of_thy) thys)
wenzelm@18120:    | TERM (msg,ts) =>
wenzelm@18120:          (writeln ("Exception TERM raised:\n" ^ msg);
wenzelm@26939:           List.app (writeln o Syntax.string_of_term_global thy) ts)
wenzelm@18120:    | TYPE (msg,Ts,ts) =>
wenzelm@18120:          (writeln ("Exception TYPE raised:\n" ^ msg);
wenzelm@26939:           List.app (writeln o Syntax.string_of_typ_global thy) Ts;
wenzelm@26939:           List.app (writeln o Syntax.string_of_term_global thy) ts)
wenzelm@18120:    | e => raise e;
wenzelm@18120: 
wenzelm@18120: (*Prints an exception, then fails*)
wenzelm@18678: fun print_sign_exn thy e = (print_sign_exn_unit thy e; raise ERROR "");
wenzelm@18120: 
wenzelm@18120: (** the prove_goal.... commands
wenzelm@18120:     Prove theorem using the listed tactics; check it has the specified form.
wenzelm@18120:     Augment theory with all type assignments of goal.
wenzelm@18120:     Syntax is similar to "goal" command for easy keyboard use. **)
wenzelm@18120: 
wenzelm@18120: (*Version taking the goal as a cterm*)
wenzelm@18120: fun prove_goalw_cterm_general check rths chorn tacsf =
wenzelm@18120:   let val (prems, st0, mkresult) = prepare_proof false rths chorn
wenzelm@18120:       val tac = EVERY (tacsf prems)
wenzelm@18120:       fun statef() =
wenzelm@18120:           (case Seq.pull (tac st0) of
wenzelm@18120:                SOME(st,_) => st
wenzelm@18120:              | _ => error ("prove_goal: tactic failed"))
wenzelm@25685:   in  mkresult (check, cond_timeit (!Output.timing) "" statef)  end
wenzelm@26626:   handle e => (print_sign_exn_unit (Thm.theory_of_cterm chorn) e;
wenzelm@18120:                writeln ("The exception above was raised for\n" ^
wenzelm@18120:                       Display.string_of_cterm chorn); raise e);
wenzelm@18120: 
wenzelm@18120: (*Two variants: one checking the result, one not.
wenzelm@18120:   Neither prints runtime messages: they are for internal packages.*)
wenzelm@18120: fun prove_goalw_cterm rths chorn =
wenzelm@18120:         setmp Output.timing false (prove_goalw_cterm_general true rths chorn)
wenzelm@18120: and prove_goalw_cterm_nocheck rths chorn =
wenzelm@18120:         setmp Output.timing false (prove_goalw_cterm_general false rths chorn);
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*Version taking the goal as a string*)
wenzelm@18120: fun prove_goalw thy rths agoal tacsf =
wenzelm@27256:   let val chorn = cterm_of thy (read_prop thy agoal)
wenzelm@18120:   in prove_goalw_cterm_general true rths chorn tacsf end
wenzelm@27256:   handle ERROR msg => cat_error msg (*from read_prop?*)
wenzelm@18120:                 ("The error(s) above occurred for " ^ quote agoal);
wenzelm@18120: 
wenzelm@18120: (*String version with no meta-rewrite-rules*)
wenzelm@18120: fun prove_goal thy = prove_goalw thy [];
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*** Commands etc ***)
wenzelm@18120: 
wenzelm@18120: (*Return the current goal stack, if any, from undo_list*)
wenzelm@18120: fun getstate() : gstack = case !undo_list of
wenzelm@18120:       []   => error"No current state in subgoal module"
wenzelm@18120:     | x::_ => x;
wenzelm@18120: 
wenzelm@18120: (*Pops the given goal stack*)
wenzelm@18120: fun pop [] = error"Cannot go back past the beginning of the proof!"
wenzelm@18120:   | pop (pair::pairs) = (pair,pairs);
wenzelm@18120: 
wenzelm@18120: 
wenzelm@23635: (* Print a level of the goal stack *)
wenzelm@18120: 
wenzelm@18120: fun print_top ((th, _), pairs) =
wenzelm@23635:   let
wenzelm@23635:     val n = length pairs;
wenzelm@23635:     val m = (! goals_limit);
wenzelm@23635:     val ngoals = nprems_of th;
wenzelm@23635:   in
wenzelm@23635:     [Pretty.str ("Level " ^ string_of_int n ^
wenzelm@23635:       (if ngoals > 0 then " (" ^ string_of_int ngoals ^ " subgoal" ^
wenzelm@23635:         (if ngoals <> 1 then "s" else "") ^ ")"
wenzelm@23635:       else ""))] @
wenzelm@23635:     Display.pretty_goals m th
wenzelm@23635:   end |> Pretty.chunks |> Pretty.writeln;
wenzelm@18120: 
wenzelm@18120: (*Printing can raise exceptions, so the assignment occurs last.
wenzelm@18120:   Can do   setstate[(st,Seq.empty)]  to set st as the state.  *)
wenzelm@18120: fun setstate newgoals =
wenzelm@18120:   (print_top (pop newgoals);  undo_list := newgoals :: !undo_list);
wenzelm@18120: 
wenzelm@18120: (*Given a proof state transformation, return a command that updates
wenzelm@18120:     the goal stack*)
wenzelm@18120: fun make_command com = setstate (com (pop (getstate())));
wenzelm@18120: 
wenzelm@18120: (*Apply a function on proof states to the current goal stack*)
wenzelm@18120: fun apply_fun f = f (pop(getstate()));
wenzelm@18120: 
wenzelm@18120: (*Return the top theorem, representing the proof state*)
wenzelm@18120: fun topthm () = apply_fun  (fn ((th,_), _) => th);
wenzelm@18120: 
wenzelm@18120: (*Return the final result.  *)
wenzelm@18120: fun result () = !curr_mkresult (true, topthm());
wenzelm@18120: 
wenzelm@18120: (*Return the result UNCHECKED that it equals the goal -- for synthesis,
wenzelm@18120:   answer extraction, or other instantiation of Vars *)
wenzelm@18120: fun uresult () = !curr_mkresult (false, topthm());
wenzelm@18120: 
wenzelm@18120: (*Get subgoal i from goal stack*)
wenzelm@18120: fun getgoal i = Logic.get_goal (prop_of (topthm())) i;
wenzelm@18120: 
wenzelm@18120: (*Return subgoal i's hypotheses as meta-level assumptions.
wenzelm@18120:   For debugging uses of METAHYPS*)
wenzelm@18120: local exception GETHYPS of thm list
wenzelm@18120: in
wenzelm@18120: fun gethyps i =
wenzelm@18120:     (METAHYPS (fn hyps => raise (GETHYPS hyps)) i (topthm());  [])
wenzelm@18120:     handle GETHYPS hyps => hyps
wenzelm@18120: end;
wenzelm@18120: 
wenzelm@18120: (*Prints exceptions nicely at top level;
wenzelm@18120:   raises the exception in order to have a polymorphic type!*)
wenzelm@18120: fun print_exn e = (print_sign_exn_unit (Thm.theory_of_thm (topthm())) e;  raise e);
wenzelm@18120: 
wenzelm@18120: (*Which thms could apply to goal i? (debugs tactics involving filter_thms) *)
wenzelm@18120: fun filter_goal could ths i = filter_thms could (999, getgoal i, ths);
wenzelm@18120: 
wenzelm@18120: (*For inspecting earlier levels of the backward proof*)
wenzelm@18120: fun chop_level n (pair,pairs) =
wenzelm@18120:   let val level = length pairs
wenzelm@18120:   in  if n<0 andalso ~n <= level
wenzelm@18120:       then  List.drop (pair::pairs, ~n)
wenzelm@18120:       else if 0<=n andalso n<= level
wenzelm@18120:       then  List.drop (pair::pairs, level - n)
wenzelm@18120:       else  error ("Level number must lie between 0 and " ^
wenzelm@18120:                    string_of_int level)
wenzelm@18120:   end;
wenzelm@18120: 
wenzelm@18120: (*Print the given level of the proof; prlev ~1 prints previous level*)
wenzelm@23635: fun prlev n = apply_fun (print_top o pop o (chop_level n));
wenzelm@23635: fun pr () = apply_fun print_top;
wenzelm@18120: 
wenzelm@18120: (*Set goals_limit and print again*)
wenzelm@18120: fun prlim n = (goals_limit:=n; pr());
wenzelm@18120: 
wenzelm@18120: (** the goal.... commands
wenzelm@18120:     Read main goal.  Set global variables curr_prems, curr_mkresult.
wenzelm@18120:     Initial subgoal and premises are rewritten using rths. **)
wenzelm@18120: 
wenzelm@18120: (*Version taking the goal as a cterm; if you have a term t and theory thy, use
wenzelm@18120:     goalw_cterm rths (cterm_of thy t);      *)
wenzelm@18120: fun agoalw_cterm atomic rths chorn =
wenzelm@18120:   let val (prems, st0, mkresult) = prepare_proof atomic rths chorn
wenzelm@18120:   in  undo_list := [];
wenzelm@18120:       setstate [ (st0, Seq.empty) ];
wenzelm@18120:       curr_prems := prems;
wenzelm@18120:       curr_mkresult := mkresult;
wenzelm@18120:       prems
wenzelm@18120:   end;
wenzelm@18120: 
wenzelm@18120: val goalw_cterm = agoalw_cterm false;
wenzelm@18120: 
wenzelm@18120: (*Version taking the goal as a string*)
wenzelm@18120: fun agoalw atomic thy rths agoal =
wenzelm@27256:     agoalw_cterm atomic rths (cterm_of thy (read_prop thy agoal))
wenzelm@18678:     handle ERROR msg => cat_error msg (*from type_assign, etc via prepare_proof*)
wenzelm@18120:         ("The error(s) above occurred for " ^ quote agoal);
wenzelm@18120: 
wenzelm@18120: val goalw = agoalw false;
wenzelm@18120: fun goal thy = goalw thy [];
wenzelm@18120: 
wenzelm@18120: (*now the versions that wrap the goal up in `Goal' to make it atomic*)
wenzelm@26429: fun Goalw thms s = agoalw true (ML_Context.the_global_context ()) thms s;
wenzelm@18120: val Goal = Goalw [];
wenzelm@18120: 
wenzelm@18120: (*simple version with minimal amount of checking and postprocessing*)
wenzelm@18120: fun simple_prove_goal_cterm G f =
wenzelm@18120:   let
wenzelm@26291:     val _ = legacy_feature "Old goal command";
wenzelm@18120:     val As = Drule.strip_imp_prems G;
wenzelm@18120:     val B = Drule.strip_imp_concl G;
wenzelm@20229:     val asms = map Assumption.assume As;
wenzelm@18120:     fun check NONE = error "prove_goal: tactic failed"
wenzelm@18120:       | check (SOME (thm, _)) = (case nprems_of thm of
wenzelm@18120:             0 => thm
wenzelm@18120:           | i => !result_error_fn thm (string_of_int i ^ " unsolved goals!"))
wenzelm@18120:   in
wenzelm@18120:     standard (implies_intr_list As
wenzelm@18120:       (check (Seq.pull (EVERY (f asms) (trivial B)))))
wenzelm@18120:   end;
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*Proof step "by" the given tactic -- apply tactic to the proof state*)
wenzelm@18120: fun by_com tac ((th,ths), pairs) : gstack =
wenzelm@18120:   (case  Seq.pull(tac th)  of
wenzelm@18120:      NONE      => error"by: tactic failed"
wenzelm@18120:    | SOME(th2,ths2) =>
wenzelm@22360:        (if Thm.eq_thm(th,th2)
wenzelm@18120:           then warning "Warning: same as previous level"
wenzelm@22360:           else if Thm.eq_thm_thy(th,th2) then ()
wenzelm@18120:           else warning ("Warning: theory of proof state has changed" ^
wenzelm@18120:                        thy_error (Thm.theory_of_thm th, Thm.theory_of_thm th2));
wenzelm@18120:        ((th2,ths2)::(th,ths)::pairs)));
wenzelm@18120: 
wenzelm@25685: fun by tac = cond_timeit (!Output.timing) ""
wenzelm@18120:     (fn() => make_command (by_com tac));
wenzelm@18120: 
wenzelm@18120: (* byev[tac1,...,tacn] applies tac1 THEN ... THEN tacn.
wenzelm@18120:    Good for debugging proofs involving prove_goal.*)
wenzelm@18120: val byev = by o EVERY;
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*Backtracking means find an alternative result from a tactic.
wenzelm@18120:   If none at this level, try earlier levels*)
wenzelm@18120: fun backtrack [] = error"back: no alternatives"
wenzelm@18120:   | backtrack ((th,thstr) :: pairs) =
wenzelm@18120:      (case Seq.pull thstr of
wenzelm@18120:           NONE      => (writeln"Going back a level..."; backtrack pairs)
wenzelm@18120:         | SOME(th2,thstr2) =>
wenzelm@22360:            (if Thm.eq_thm(th,th2)
wenzelm@18120:               then warning "Warning: same as previous choice at this level"
wenzelm@22360:               else if Thm.eq_thm_thy(th,th2) then ()
wenzelm@18120:               else warning "Warning: theory of proof state has changed";
wenzelm@18120:             (th2,thstr2)::pairs));
wenzelm@18120: 
wenzelm@18120: fun back() = setstate (backtrack (getstate()));
wenzelm@18120: 
wenzelm@18120: (*Chop back to previous level of the proof*)
wenzelm@18120: fun choplev n = make_command (chop_level n);
wenzelm@18120: 
wenzelm@18120: (*Chopping back the goal stack*)
wenzelm@18120: fun chop () = make_command (fn (_,pairs) => pairs);
wenzelm@18120: 
wenzelm@18120: (*Restore the previous proof state;  discard current state. *)
wenzelm@18120: fun undo() = case !undo_list of
wenzelm@18120:       [] => error"No proof state"
wenzelm@18120:     | [_] => error"Already at initial state"
wenzelm@18120:     | _::newundo =>  (undo_list := newundo;  pr()) ;
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: (*** Managing the proof stack ***)
wenzelm@18120: 
wenzelm@18120: fun save_proof() = Proof(!undo_list, !curr_prems, !curr_mkresult);
wenzelm@18120: 
wenzelm@18120: fun restore_proof(Proof(ul,prems,mk)) =
wenzelm@18120:  (undo_list:= ul;  curr_prems:= prems;  curr_mkresult := mk;  prems);
wenzelm@18120: 
wenzelm@18120: 
wenzelm@18120: fun top_proof() = case !proofstack of
wenzelm@18120:         [] => error("Stack of proof attempts is empty!")
wenzelm@18120:     | p::ps => (p,ps);
wenzelm@18120: 
wenzelm@18120: (*  push a copy of the current proof state on to the stack *)
wenzelm@18120: fun push_proof() = (proofstack := (save_proof() :: !proofstack));
wenzelm@18120: 
wenzelm@18120: (* discard the top proof state of the stack *)
wenzelm@18120: fun pop_proof() =
wenzelm@18120:   let val (p,ps) = top_proof()
wenzelm@18120:       val prems = restore_proof p
wenzelm@18120:   in proofstack := ps;  pr();  prems end;
wenzelm@18120: 
wenzelm@18120: (* rotate the stack so that the top element goes to the bottom *)
wenzelm@18120: fun rotate_proof() = let val (p,ps) = top_proof()
wenzelm@18120:                     in proofstack := ps@[save_proof()];
wenzelm@18120:                        restore_proof p;
wenzelm@18120:                        pr();
wenzelm@18120:                        !curr_prems
wenzelm@18120:                     end;
wenzelm@18120: 
wenzelm@18120: 
wenzelm@26414: (** theorem bindings **)
wenzelm@18120: 
wenzelm@26414: fun qed name = ML_Context.ml_store_thm (name, result ());
wenzelm@26414: fun qed_goal name thy goal tacsf = ML_Context.ml_store_thm (name, prove_goal thy goal tacsf);
wenzelm@18120: fun qed_goalw name thy rews goal tacsf =
wenzelm@26414:   ML_Context.ml_store_thm (name, prove_goalw thy rews goal tacsf);
wenzelm@18120: fun qed_spec_mp name =
wenzelm@26414:   ML_Context.ml_store_thm (name, ObjectLogic.rulify_no_asm (result ()));
wenzelm@18120: fun qed_goal_spec_mp name thy s p =
wenzelm@18120:   bind_thm (name, ObjectLogic.rulify_no_asm (prove_goal thy s p));
wenzelm@18120: fun qed_goalw_spec_mp name thy defs s p =
wenzelm@18120:   bind_thm (name, ObjectLogic.rulify_no_asm (prove_goalw thy defs s p));
wenzelm@18120: 
wenzelm@18120: end;
wenzelm@18120: 
wenzelm@18120: structure Goals: GOALS = OldGoals;
wenzelm@18120: open Goals;