src/HOL/Tools/ATP/atp_problem.ML
author blanchet
Tue, 30 Aug 2011 16:07:45 +0200
changeset 45457 ccf40af26ae9
parent 45453 0a1dfc6365e9
child 45458 ae82943481e9
permissions -rw-r--r--
implement more of the polymorphic simply typed format TFF(1)
blanchet@38293
     1
(*  Title:      HOL/Tools/ATP/atp_problem.ML
blanchet@38261
     2
    Author:     Jia Meng, Cambridge University Computer Laboratory and NICTA
blanchet@37509
     3
    Author:     Jasmin Blanchette, TU Muenchen
blanchet@37509
     4
blanchet@39692
     5
Abstract representation of ATP problems and TPTP syntax.
blanchet@37509
     6
*)
blanchet@37509
     7
blanchet@38253
     8
signature ATP_PROBLEM =
blanchet@37509
     9
sig
nik@44535
    10
  datatype ('a, 'b) ho_term =
nik@44535
    11
    ATerm of 'a * ('a, 'b) ho_term list |
nik@44535
    12
    AAbs of ('a * 'b) * ('a, 'b) ho_term
blanchet@38226
    13
  datatype quantifier = AForall | AExists
blanchet@44004
    14
  datatype connective = ANot | AAnd | AOr | AImplies | AIff
blanchet@43402
    15
  datatype ('a, 'b, 'c) formula =
blanchet@43402
    16
    AQuant of quantifier * ('a * 'b option) list * ('a, 'b, 'c) formula |
blanchet@43402
    17
    AConn of connective * ('a, 'b, 'c) formula list |
blanchet@43402
    18
    AAtom of 'c
blanchet@38228
    19
blanchet@45457
    20
  datatype 'a ho_type =
blanchet@45457
    21
    AType of 'a * 'a ho_type list |
blanchet@45457
    22
    AFun of 'a ho_type * 'a ho_type
blanchet@43804
    23
blanchet@45453
    24
  datatype tff_polymorphism = TFF_Monomorphic | TFF_Polymorphic
blanchet@45453
    25
  datatype tff_explicitness = TFF_Implicit | TFF_Explicit
blanchet@45453
    26
  datatype thf_flavor = THF_Without_Choice | THF_With_Choice
blanchet@45106
    27
  datatype format =
blanchet@45106
    28
    CNF |
blanchet@45106
    29
    CNF_UEQ |
blanchet@45106
    30
    FOF |
blanchet@45453
    31
    TFF of tff_polymorphism * tff_explicitness |
blanchet@45453
    32
    THF0 of thf_flavor
blanchet@45106
    33
blanchet@43396
    34
  datatype formula_kind = Axiom | Definition | Lemma | Hypothesis | Conjecture
blanchet@43398
    35
  datatype 'a problem_line =
blanchet@43804
    36
    Decl of string * 'a * 'a ho_type |
blanchet@45261
    37
    Formula of string * formula_kind
blanchet@45261
    38
               * ('a, 'a ho_type, ('a, 'a ho_type) ho_term) formula
blanchet@45261
    39
               * (string, string ho_type) ho_term option
blanchet@45261
    40
               * (string, string ho_type) ho_term option
blanchet@38251
    41
  type 'a problem = (string * 'a problem_line list) list
blanchet@38226
    42
blanchet@43809
    43
  val tptp_cnf : string
blanchet@43809
    44
  val tptp_fof : string
blanchet@43809
    45
  val tptp_tff : string
blanchet@43809
    46
  val tptp_thf : string
blanchet@43808
    47
  val tptp_has_type : string
blanchet@43803
    48
  val tptp_type_of_types : string
blanchet@43803
    49
  val tptp_bool_type : string
blanchet@43803
    50
  val tptp_individual_type : string
blanchet@43804
    51
  val tptp_fun_type : string
blanchet@43839
    52
  val tptp_product_type : string
blanchet@43809
    53
  val tptp_forall : string
nik@44537
    54
  val tptp_ho_forall : string
blanchet@43809
    55
  val tptp_exists : string
nik@44537
    56
  val tptp_ho_exists : string
nik@45350
    57
  val tptp_choice : string
blanchet@43809
    58
  val tptp_not : string
blanchet@43809
    59
  val tptp_and : string
blanchet@43809
    60
  val tptp_or : string
blanchet@43809
    61
  val tptp_implies : string
blanchet@43809
    62
  val tptp_if : string
blanchet@43809
    63
  val tptp_iff : string
blanchet@43809
    64
  val tptp_not_iff : string
blanchet@43809
    65
  val tptp_app : string
blanchet@43809
    66
  val tptp_not_infix : string
blanchet@43809
    67
  val tptp_equal : string
blanchet@43841
    68
  val tptp_old_equal : string
blanchet@43807
    69
  val tptp_false : string
blanchet@43807
    70
  val tptp_true : string
blanchet@43841
    71
  val tptp_empty_list : string
blanchet@43841
    72
  val is_tptp_equal : string -> bool
blanchet@43839
    73
  val is_built_in_tptp_symbol : string -> bool
blanchet@43839
    74
  val is_tptp_variable : string -> bool
blanchet@43839
    75
  val is_tptp_user_symbol : string -> bool
blanchet@43783
    76
  val mk_anot : ('a, 'b, 'c) formula -> ('a, 'b, 'c) formula
blanchet@43783
    77
  val mk_aconn :
blanchet@43783
    78
    connective -> ('a, 'b, 'c) formula -> ('a, 'b, 'c) formula
blanchet@43783
    79
    -> ('a, 'b, 'c) formula
blanchet@43839
    80
  val aconn_fold :
blanchet@43839
    81
    bool option -> (bool option -> 'a -> 'b -> 'b) -> connective * 'a list
blanchet@43839
    82
    -> 'b -> 'b
blanchet@43839
    83
  val aconn_map :
blanchet@43839
    84
    bool option -> (bool option -> 'a -> ('b, 'c, 'd) formula)
blanchet@43839
    85
    -> connective * 'a list -> ('b, 'c, 'd) formula
blanchet@43839
    86
  val formula_fold :
blanchet@43839
    87
    bool option -> (bool option -> 'c -> 'd -> 'd) -> ('a, 'b, 'c) formula
blanchet@43839
    88
    -> 'd -> 'd
blanchet@43785
    89
  val formula_map : ('c -> 'd) -> ('a, 'b, 'c) formula -> ('a, 'b, 'd) formula
blanchet@45106
    90
  val is_format_thf : format -> bool
blanchet@43839
    91
  val is_format_typed : format -> bool
blanchet@44065
    92
  val tptp_lines_for_atp_problem : format -> string problem -> string list
blanchet@43933
    93
  val ensure_cnf_problem :
blanchet@43933
    94
    (string * string) problem -> (string * string) problem
blanchet@43780
    95
  val filter_cnf_ueq_problem :
blanchet@43780
    96
    (string * string) problem -> (string * string) problem
blanchet@43839
    97
  val declare_undeclared_syms_in_atp_problem :
blanchet@43839
    98
    string -> string -> (string * string) problem -> (string * string) problem
blanchet@39692
    99
  val nice_atp_problem :
blanchet@38251
   100
    bool -> ('a * (string * string) problem_line list) list
blanchet@38251
   101
    -> ('a * string problem_line list) list
blanchet@38251
   102
       * (string Symtab.table * string Symtab.table) option
blanchet@37509
   103
end;
blanchet@37509
   104
blanchet@38253
   105
structure ATP_Problem : ATP_PROBLEM =
blanchet@37509
   106
struct
blanchet@37509
   107
blanchet@43926
   108
open ATP_Util
blanchet@43926
   109
blanchet@43926
   110
blanchet@37643
   111
(** ATP problem **)
blanchet@37643
   112
nik@44535
   113
datatype ('a, 'b) ho_term =
nik@44535
   114
  ATerm of 'a * ('a, 'b) ho_term list |
nik@44535
   115
  AAbs of ('a * 'b) * ('a, 'b) ho_term
blanchet@38198
   116
datatype quantifier = AForall | AExists
blanchet@44004
   117
datatype connective = ANot | AAnd | AOr | AImplies | AIff
blanchet@43402
   118
datatype ('a, 'b, 'c) formula =
blanchet@43402
   119
  AQuant of quantifier * ('a * 'b option) list * ('a, 'b, 'c) formula |
blanchet@43402
   120
  AConn of connective * ('a, 'b, 'c) formula list |
blanchet@43402
   121
  AAtom of 'c
blanchet@38198
   122
blanchet@45457
   123
datatype 'a ho_type =
blanchet@45457
   124
  AType of 'a * 'a ho_type list |
blanchet@45457
   125
  AFun of 'a ho_type * 'a ho_type
blanchet@43804
   126
blanchet@45453
   127
datatype tff_polymorphism = TFF_Monomorphic | TFF_Polymorphic
blanchet@45453
   128
datatype tff_explicitness = TFF_Implicit | TFF_Explicit
blanchet@45453
   129
datatype thf_flavor = THF_Without_Choice | THF_With_Choice
blanchet@45354
   130
blanchet@45106
   131
datatype format =
blanchet@45106
   132
  CNF |
blanchet@45106
   133
  CNF_UEQ |
blanchet@45106
   134
  FOF |
blanchet@45453
   135
  TFF of tff_polymorphism * tff_explicitness |
blanchet@45453
   136
  THF0 of thf_flavor
blanchet@45106
   137
blanchet@43396
   138
datatype formula_kind = Axiom | Definition | Lemma | Hypothesis | Conjecture
blanchet@43398
   139
datatype 'a problem_line =
blanchet@43804
   140
  Decl of string * 'a * 'a ho_type |
nik@44535
   141
  Formula of string * formula_kind * ('a, 'a ho_type, ('a, 'a ho_type) ho_term) formula
nik@44535
   142
             * (string, string ho_type) ho_term option * (string, string ho_type) ho_term option
blanchet@37643
   143
type 'a problem = (string * 'a problem_line list) list
blanchet@37643
   144
blanchet@43587
   145
(* official TPTP syntax *)
blanchet@43809
   146
val tptp_cnf = "cnf"
blanchet@43809
   147
val tptp_fof = "fof"
blanchet@43809
   148
val tptp_tff = "tff"
blanchet@43809
   149
val tptp_thf = "thf"
blanchet@43808
   150
val tptp_has_type = ":"
blanchet@43803
   151
val tptp_type_of_types = "$tType"
blanchet@43803
   152
val tptp_bool_type = "$o"
blanchet@43803
   153
val tptp_individual_type = "$i"
blanchet@43804
   154
val tptp_fun_type = ">"
blanchet@43839
   155
val tptp_product_type = "*"
blanchet@43809
   156
val tptp_forall = "!"
nik@44537
   157
val tptp_ho_forall = "!!"
blanchet@43809
   158
val tptp_exists = "?"
nik@44537
   159
val tptp_ho_exists = "??"
nik@45350
   160
val tptp_choice = "@+"
blanchet@43809
   161
val tptp_not = "~"
blanchet@43809
   162
val tptp_and = "&"
blanchet@43809
   163
val tptp_or = "|"
blanchet@43809
   164
val tptp_implies = "=>"
blanchet@43809
   165
val tptp_if = "<="
blanchet@43809
   166
val tptp_iff = "<=>"
blanchet@43809
   167
val tptp_not_iff = "<~>"
blanchet@43809
   168
val tptp_app = "@"
blanchet@43809
   169
val tptp_not_infix = "!"
blanchet@43809
   170
val tptp_equal = "="
blanchet@43841
   171
val tptp_old_equal = "equal"
blanchet@43807
   172
val tptp_false = "$false"
blanchet@43807
   173
val tptp_true = "$true"
blanchet@43841
   174
val tptp_empty_list = "[]"
blanchet@43587
   175
blanchet@43841
   176
fun is_tptp_equal s = (s = tptp_equal orelse s = tptp_old_equal)
blanchet@43841
   177
fun is_built_in_tptp_symbol s =
blanchet@43841
   178
  s = tptp_old_equal orelse not (Char.isAlpha (String.sub (s, 0)))
blanchet@43839
   179
fun is_tptp_variable s = Char.isUpper (String.sub (s, 0))
blanchet@43839
   180
val is_tptp_user_symbol = not o (is_tptp_variable orf is_built_in_tptp_symbol)
blanchet@43780
   181
blanchet@43939
   182
fun raw_polarities_of_conn ANot = (SOME false, NONE)
blanchet@43939
   183
  | raw_polarities_of_conn AAnd = (SOME true, SOME true)
blanchet@43939
   184
  | raw_polarities_of_conn AOr = (SOME true, SOME true)
blanchet@43939
   185
  | raw_polarities_of_conn AImplies = (SOME false, SOME true)
blanchet@43939
   186
  | raw_polarities_of_conn AIff = (NONE, NONE)
blanchet@43939
   187
fun polarities_of_conn NONE = K (NONE, NONE)
blanchet@43939
   188
  | polarities_of_conn (SOME pos) =
blanchet@43939
   189
    raw_polarities_of_conn #> not pos ? pairself (Option.map not)
blanchet@43939
   190
blanchet@43783
   191
fun mk_anot (AConn (ANot, [phi])) = phi
blanchet@43783
   192
  | mk_anot phi = AConn (ANot, [phi])
blanchet@43783
   193
fun mk_aconn c phi1 phi2 = AConn (c, [phi1, phi2])
blanchet@43783
   194
blanchet@43839
   195
fun aconn_fold pos f (ANot, [phi]) = f (Option.map not pos) phi
blanchet@43839
   196
  | aconn_fold pos f (AImplies, [phi1, phi2]) =
blanchet@43839
   197
    f (Option.map not pos) phi1 #> f pos phi2
blanchet@43839
   198
  | aconn_fold pos f (AAnd, phis) = fold (f pos) phis
blanchet@43839
   199
  | aconn_fold pos f (AOr, phis) = fold (f pos) phis
blanchet@43839
   200
  | aconn_fold _ f (_, phis) = fold (f NONE) phis
blanchet@43839
   201
blanchet@43839
   202
fun aconn_map pos f (ANot, [phi]) = AConn (ANot, [f (Option.map not pos) phi])
blanchet@43839
   203
  | aconn_map pos f (AImplies, [phi1, phi2]) =
blanchet@43839
   204
    AConn (AImplies, [f (Option.map not pos) phi1, f pos phi2])
blanchet@43839
   205
  | aconn_map pos f (AAnd, phis) = AConn (AAnd, map (f pos) phis)
blanchet@43839
   206
  | aconn_map pos f (AOr, phis) = AConn (AOr, map (f pos) phis)
blanchet@43839
   207
  | aconn_map _ f (c, phis) = AConn (c, map (f NONE) phis)
blanchet@43839
   208
blanchet@43839
   209
fun formula_fold pos f =
blanchet@43839
   210
  let
blanchet@45356
   211
    fun fld pos (AQuant (_, _, phi)) = fld pos phi
blanchet@45356
   212
      | fld pos (AConn conn) = aconn_fold pos fld conn
blanchet@45356
   213
      | fld pos (AAtom tm) = f pos tm
blanchet@45356
   214
  in fld pos end
blanchet@43839
   215
blanchet@43785
   216
fun formula_map f (AQuant (q, xs, phi)) = AQuant (q, xs, formula_map f phi)
blanchet@43785
   217
  | formula_map f (AConn (c, phis)) = AConn (c, map (formula_map f) phis)
blanchet@43785
   218
  | formula_map f (AAtom tm) = AAtom (f tm)
blanchet@43785
   219
blanchet@45453
   220
fun is_format_thf (THF0 _) = true
blanchet@45106
   221
  | is_format_thf _ = false
blanchet@45354
   222
fun is_format_typed (TFF _) = true
blanchet@45453
   223
  | is_format_typed (THF0 _) = true
blanchet@45106
   224
  | is_format_typed _ = false
blanchet@43839
   225
blanchet@38854
   226
fun string_for_kind Axiom = "axiom"
blanchet@42640
   227
  | string_for_kind Definition = "definition"
blanchet@42640
   228
  | string_for_kind Lemma = "lemma"
blanchet@38854
   229
  | string_for_kind Hypothesis = "hypothesis"
blanchet@38854
   230
  | string_for_kind Conjecture = "conjecture"
blanchet@38854
   231
blanchet@45457
   232
fun strip_tff_type (AFun (AFun _, _)) =
blanchet@43804
   233
    raise Fail "unexpected higher-order type in first-order format"
blanchet@45457
   234
  | strip_tff_type (AFun (ty1, ty2)) = strip_tff_type ty2 |>> cons ty1
blanchet@45457
   235
  | strip_tff_type ty = ([], ty)
blanchet@43804
   236
blanchet@45457
   237
fun general_string_for_type ty =
blanchet@45457
   238
  let
blanchet@45457
   239
    fun str _ (AType (s, [])) = s
blanchet@45457
   240
      | str _ (AType (s, tys)) = s ^ "(" ^ commas (map (str false) tys) ^ ")"
blanchet@45457
   241
      | str rhs (AFun (ty1, ty2)) =
blanchet@45457
   242
        str false ty1 ^ " " ^ tptp_fun_type ^ " " ^ str true ty2
blanchet@45457
   243
        |> not rhs ? enclose "(" ")"
blanchet@45457
   244
  in str true ty end
blanchet@45457
   245
blanchet@45457
   246
fun string_for_type (THF0 _) ty = general_string_for_type ty
blanchet@45354
   247
  | string_for_type (TFF _) ty =
blanchet@43804
   248
    (case strip_tff_type ty of
blanchet@45457
   249
       ([], ty) => general_string_for_type ty
blanchet@45457
   250
     | ([ty1], ty2) => general_string_for_type (AFun (ty1, ty2))
blanchet@45457
   251
     | (tys, ty) =>
blanchet@45457
   252
       "(" ^ space_implode (" " ^ tptp_product_type ^ " ")
blanchet@45457
   253
       (map general_string_for_type tys) ^ ") " ^ tptp_fun_type ^ " " ^
blanchet@45457
   254
       general_string_for_type ty)
blanchet@43804
   255
  | string_for_type _ _ = raise Fail "unexpected type in untyped format"
blanchet@43804
   256
blanchet@43809
   257
fun string_for_quantifier AForall = tptp_forall
blanchet@43809
   258
  | string_for_quantifier AExists = tptp_exists
blanchet@43804
   259
blanchet@43808
   260
fun string_for_connective ANot = tptp_not
blanchet@43809
   261
  | string_for_connective AAnd = tptp_and
blanchet@43809
   262
  | string_for_connective AOr = tptp_or
blanchet@43809
   263
  | string_for_connective AImplies = tptp_implies
blanchet@43809
   264
  | string_for_connective AIff = tptp_iff
blanchet@43804
   265
blanchet@43803
   266
fun string_for_bound_var format (s, ty) =
blanchet@45457
   267
  s ^
blanchet@45457
   268
  (if is_format_typed format then
blanchet@45457
   269
     " " ^ tptp_has_type ^ " " ^
blanchet@45457
   270
     (ty |> the_default (AType (tptp_individual_type, []))
blanchet@45457
   271
         |> string_for_type format)
blanchet@45457
   272
   else
blanchet@45457
   273
     "")
blanchet@43804
   274
blanchet@44857
   275
fun string_for_term _ (ATerm (s, [])) = s
blanchet@44857
   276
  | string_for_term format (ATerm (s, ts)) =
blanchet@44857
   277
    if s = tptp_empty_list then
blanchet@44857
   278
      (* used for lists in the optional "source" field of a derivation *)
blanchet@44857
   279
      "[" ^ commas (map (string_for_term format) ts) ^ "]"
blanchet@44857
   280
    else if is_tptp_equal s then
blanchet@44857
   281
      space_implode (" " ^ tptp_equal ^ " ") (map (string_for_term format) ts)
blanchet@45106
   282
      |> is_format_thf format ? enclose "(" ")"
blanchet@44857
   283
    else
nik@45350
   284
      (case (s = tptp_ho_forall orelse s = tptp_ho_exists,
blanchet@45453
   285
             s = tptp_choice andalso format = THF0 THF_With_Choice, ts) of
nik@45350
   286
         (true, _, [AAbs ((s', ty), tm)]) =>
blanchet@44858
   287
         (* Use syntactic sugar "!" and "?" instead of "!!" and "??" whenever
blanchet@44858
   288
            possible, to work around LEO-II 1.2.8 parser limitation. *)
blanchet@44857
   289
         string_for_formula format
blanchet@44857
   290
             (AQuant (if s = tptp_ho_forall then AForall else AExists,
blanchet@44857
   291
                      [(s', SOME ty)], AAtom tm))
nik@45350
   292
       | (_, true, [AAbs ((s', ty), tm)]) =>
nik@45350
   293
         (*There is code in ATP_Translate to ensure that Eps is always applied
nik@45350
   294
           to an abstraction*)
nik@45350
   295
         tptp_choice ^ "[" ^ s' ^ " : " ^ string_for_type format ty ^ "] : " ^
nik@45350
   296
           string_for_term format tm ^ ""
nik@45350
   297
         |> enclose "(" ")"
nik@45350
   298
blanchet@44857
   299
       | _ =>
blanchet@44857
   300
         let val ss = map (string_for_term format) ts in
blanchet@45106
   301
           if is_format_thf format then
blanchet@44857
   302
             "(" ^ space_implode (" " ^ tptp_app ^ " ") (s :: ss) ^ ")"
blanchet@44857
   303
           else
blanchet@44857
   304
             s ^ "(" ^ commas ss ^ ")"
blanchet@44857
   305
         end)
blanchet@45453
   306
  | string_for_term (format as THF0 _) (AAbs ((s, ty), tm)) =
blanchet@45106
   307
    "(^[" ^ s ^ " : " ^ string_for_type format ty ^ "] : " ^
blanchet@45106
   308
    string_for_term format tm ^ ")"
blanchet@44857
   309
  | string_for_term _ _ = raise Fail "unexpected term in first-order format"
blanchet@44857
   310
and string_for_formula format (AQuant (q, xs, phi)) =
blanchet@43815
   311
    string_for_quantifier q ^
blanchet@43587
   312
    "[" ^ commas (map (string_for_bound_var format) xs) ^ "] : " ^
blanchet@43815
   313
    string_for_formula format phi
blanchet@43815
   314
    |> enclose "(" ")"
blanchet@43841
   315
  | string_for_formula format
blanchet@43841
   316
        (AConn (ANot, [AAtom (ATerm ("=" (* tptp_equal *), ts))])) =
blanchet@43809
   317
    space_implode (" " ^ tptp_not_infix ^ tptp_equal ^ " ")
blanchet@43809
   318
                  (map (string_for_term format) ts)
blanchet@45106
   319
    |> is_format_thf format ? enclose "(" ")"
blanchet@43587
   320
  | string_for_formula format (AConn (c, [phi])) =
blanchet@43815
   321
    string_for_connective c ^ " " ^
blanchet@45106
   322
    (string_for_formula format phi |> is_format_thf format ? enclose "(" ")")
blanchet@43815
   323
    |> enclose "(" ")"
blanchet@43587
   324
  | string_for_formula format (AConn (c, phis)) =
blanchet@43815
   325
    space_implode (" " ^ string_for_connective c ^ " ")
blanchet@43815
   326
                  (map (string_for_formula format) phis)
blanchet@43815
   327
    |> enclose "(" ")"
blanchet@43803
   328
  | string_for_formula format (AAtom tm) = string_for_term format tm
blanchet@38198
   329
blanchet@44558
   330
fun the_source (SOME source) = source
blanchet@44558
   331
  | the_source NONE =
blanchet@44558
   332
    ATerm ("inference",
blanchet@44558
   333
           ATerm ("isabelle", []) :: replicate 2 (ATerm ("[]", [])))
blanchet@43510
   334
blanchet@43933
   335
fun string_for_format CNF = tptp_cnf
blanchet@43933
   336
  | string_for_format CNF_UEQ = tptp_cnf
blanchet@43809
   337
  | string_for_format FOF = tptp_fof
blanchet@45354
   338
  | string_for_format (TFF _) = tptp_tff
blanchet@45453
   339
  | string_for_format (THF0 _) = tptp_thf
blanchet@43803
   340
blanchet@43804
   341
fun string_for_problem_line format (Decl (ident, sym, ty)) =
blanchet@43803
   342
    string_for_format format ^ "(" ^ ident ^ ", type,\n    " ^ sym ^ " : " ^
blanchet@43804
   343
    string_for_type format ty ^ ").\n"
blanchet@43780
   344
  | string_for_problem_line format (Formula (ident, kind, phi, source, info)) =
blanchet@43803
   345
    string_for_format format ^ "(" ^ ident ^ ", " ^ string_for_kind kind ^
blanchet@43803
   346
    ",\n    (" ^ string_for_formula format phi ^ ")" ^
blanchet@43780
   347
    (case (source, info) of
blanchet@43580
   348
       (NONE, NONE) => ""
blanchet@43803
   349
     | (SOME tm, NONE) => ", " ^ string_for_term format tm
blanchet@43580
   350
     | (_, SOME tm) =>
blanchet@44558
   351
       ", " ^ string_for_term format (the_source source) ^
blanchet@43803
   352
       ", " ^ string_for_term format tm) ^ ").\n"
blanchet@44065
   353
fun tptp_lines_for_atp_problem format problem =
blanchet@37643
   354
  "% This file was generated by Isabelle (most likely Sledgehammer)\n\
blanchet@37643
   355
  \% " ^ timestamp () ^ "\n" ::
blanchet@37643
   356
  maps (fn (_, []) => []
blanchet@37643
   357
         | (heading, lines) =>
wenzelm@41739
   358
           "\n% " ^ heading ^ " (" ^ string_of_int (length lines) ^ ")\n" ::
blanchet@43580
   359
           map (string_for_problem_line format) lines)
blanchet@38854
   360
       problem
blanchet@37643
   361
blanchet@43780
   362
blanchet@43939
   363
(** CNF (Metis) and CNF UEQ (Waldmeister) **)
blanchet@43780
   364
blanchet@43780
   365
fun is_problem_line_negated (Formula (_, _, AConn (ANot, _), _, _)) = true
blanchet@43780
   366
  | is_problem_line_negated _ = false
blanchet@43780
   367
blanchet@44034
   368
fun is_problem_line_cnf_ueq (Formula (_, _, AAtom (ATerm ((s, _), _)), _, _)) =
blanchet@44034
   369
    is_tptp_equal s
blanchet@43780
   370
  | is_problem_line_cnf_ueq _ = false
blanchet@43780
   371
blanchet@43785
   372
fun open_conjecture_term (ATerm ((s, s'), tms)) =
nik@44535
   373
    ATerm (if is_tptp_variable s then (s |> Name.desymbolize false, s')
nik@44535
   374
           else (s, s'), tms |> map open_conjecture_term)
nik@44535
   375
  | open_conjecture_term _ = raise Fail "unexpected higher-order term"
blanchet@43939
   376
fun open_formula conj =
blanchet@43939
   377
  let
blanchet@44034
   378
    (* We are conveniently assuming that all bound variable names are
blanchet@44034
   379
       distinct, which should be the case for the formulas we generate. *)
blanchet@44034
   380
    fun opn (pos as SOME true) (AQuant (AForall, _, phi)) = opn pos phi
blanchet@44034
   381
      | opn (pos as SOME false) (AQuant (AExists, _, phi)) = opn pos phi
blanchet@43939
   382
      | opn pos (AConn (ANot, [phi])) = mk_anot (opn (Option.map not pos) phi)
blanchet@43939
   383
      | opn pos (AConn (c, [phi1, phi2])) =
blanchet@43939
   384
        let val (pos1, pos2) = polarities_of_conn pos c in
blanchet@43939
   385
          AConn (c, [opn pos1 phi1, opn pos2 phi2])
blanchet@43939
   386
        end
blanchet@43939
   387
      | opn _ (AAtom t) = AAtom (t |> conj ? open_conjecture_term)
blanchet@44283
   388
      | opn _ phi = phi
blanchet@43939
   389
  in opn (SOME (not conj)) end
blanchet@43785
   390
fun open_formula_line (Formula (ident, kind, phi, source, info)) =
blanchet@43785
   391
    Formula (ident, kind, open_formula (kind = Conjecture) phi, source, info)
blanchet@43785
   392
  | open_formula_line line = line
blanchet@43780
   393
blanchet@43780
   394
fun negate_conjecture_line (Formula (ident, Conjecture, phi, source, info)) =
blanchet@43783
   395
    Formula (ident, Hypothesis, mk_anot phi, source, info)
blanchet@43780
   396
  | negate_conjecture_line line = line
blanchet@43780
   397
blanchet@43939
   398
exception CLAUSIFY of unit
blanchet@43939
   399
blanchet@43967
   400
(* This "clausification" only expands syntactic sugar, such as "phi => psi" to
blanchet@43967
   401
   "~ phi | psi" and "phi <=> psi" to "~ phi | psi" and "~ psi | phi". We don't
blanchet@43967
   402
   attempt to distribute conjunctions over disjunctions. *)
blanchet@44034
   403
fun clausify_formula pos (phi as AAtom _) = [phi |> not pos ? mk_anot]
blanchet@44034
   404
  | clausify_formula pos (AConn (ANot, [phi])) = clausify_formula (not pos) phi
blanchet@44034
   405
  | clausify_formula true (AConn (AOr, [phi1, phi2])) =
blanchet@44034
   406
    (phi1, phi2) |> pairself (clausify_formula true)
blanchet@44034
   407
                 |> uncurry (map_product (mk_aconn AOr))
blanchet@44034
   408
  | clausify_formula false (AConn (AAnd, [phi1, phi2])) =
blanchet@44034
   409
    (phi1, phi2) |> pairself (clausify_formula false)
blanchet@44034
   410
                 |> uncurry (map_product (mk_aconn AOr))
blanchet@44034
   411
  | clausify_formula true (AConn (AImplies, [phi1, phi2])) =
blanchet@44034
   412
    clausify_formula true (AConn (AOr, [mk_anot phi1, phi2]))
blanchet@44034
   413
  | clausify_formula true (AConn (AIff, phis)) =
blanchet@44034
   414
    clausify_formula true (AConn (AImplies, phis)) @
blanchet@44034
   415
    clausify_formula true (AConn (AImplies, rev phis))
blanchet@44034
   416
  | clausify_formula _ _ = raise CLAUSIFY ()
blanchet@43939
   417
blanchet@43939
   418
fun clausify_formula_line (Formula (ident, kind, phi, source, info)) =
blanchet@43967
   419
    let
blanchet@43967
   420
      val (n, phis) = phi |> try (clausify_formula true) |> these |> `length
blanchet@43967
   421
    in
blanchet@43967
   422
      map2 (fn phi => fn j =>
blanchet@44153
   423
               Formula (ident ^ replicate_string (j - 1) "x", kind, phi, source,
blanchet@44153
   424
                        info))
blanchet@43967
   425
           phis (1 upto n)
blanchet@43967
   426
    end
blanchet@43967
   427
  | clausify_formula_line _ = []
blanchet@43939
   428
blanchet@43939
   429
fun ensure_cnf_problem_line line =
blanchet@43939
   430
  line |> open_formula_line |> negate_conjecture_line |> clausify_formula_line
blanchet@43939
   431
blanchet@43939
   432
fun ensure_cnf_problem problem =
blanchet@43967
   433
  problem |> map (apsnd (maps ensure_cnf_problem_line))
blanchet@43933
   434
blanchet@43803
   435
fun filter_cnf_ueq_problem problem =
blanchet@43803
   436
  problem
blanchet@43803
   437
  |> map (apsnd (map open_formula_line
blanchet@43803
   438
                 #> filter is_problem_line_cnf_ueq
blanchet@43803
   439
                 #> map negate_conjecture_line))
blanchet@43803
   440
  |> (fn problem =>
blanchet@43780
   441
         let
blanchet@44688
   442
           val lines = problem |> maps snd
blanchet@44688
   443
           val conjs = lines |> filter is_problem_line_negated
blanchet@44688
   444
         in if length conjs = 1 andalso conjs <> lines then problem else [] end)
blanchet@38251
   445
blanchet@37643
   446
blanchet@43839
   447
(** Symbol declarations **)
blanchet@43839
   448
blanchet@43839
   449
(* TFF allows implicit declarations of types, function symbols, and predicate
blanchet@43839
   450
   symbols (with "$i" as the type of individuals), but some provers (e.g.,
blanchet@43839
   451
   SNARK) require explicit declarations. The situation is similar for THF. *)
blanchet@43839
   452
blanchet@45457
   453
val atype_of_types = AType (`I tptp_type_of_types, [])
blanchet@45457
   454
val bool_atype = AType (`I tptp_bool_type, [])
blanchet@45457
   455
val individual_atype = AType (`I tptp_individual_type, [])
blanchet@43839
   456
blanchet@43839
   457
fun default_type pred_sym =
blanchet@43839
   458
  let
blanchet@43839
   459
    fun typ 0 = if pred_sym then bool_atype else individual_atype
blanchet@43839
   460
      | typ ary = AFun (individual_atype, typ (ary - 1))
blanchet@43839
   461
  in typ end
blanchet@43839
   462
blanchet@43839
   463
fun add_declared_syms_in_problem_line (Decl (_, sym, _)) = insert (op =) sym
blanchet@43839
   464
  | add_declared_syms_in_problem_line _ = I
blanchet@43839
   465
fun declared_syms_in_problem problem =
blanchet@43839
   466
  fold (fold add_declared_syms_in_problem_line o snd) problem []
blanchet@43839
   467
blanchet@43839
   468
fun undeclared_syms_in_problem declared problem =
blanchet@43839
   469
  let
blanchet@43839
   470
    fun do_sym name ty =
blanchet@43839
   471
      if member (op =) declared name then I else AList.default (op =) (name, ty)
blanchet@45457
   472
    fun do_type (AType (name, tys)) =
blanchet@45457
   473
        do_sym name (K atype_of_types) #> fold do_type tys
blanchet@45457
   474
      | do_type (AFun (ty1, ty2)) = do_type ty1 #> do_type ty2
blanchet@43839
   475
    fun do_term pred_sym (ATerm (name as (s, _), tms)) =
nik@44535
   476
        is_tptp_user_symbol s
nik@44535
   477
        ? do_sym name (fn _ => default_type pred_sym (length tms))
nik@44535
   478
        #> fold (do_term false) tms
nik@44535
   479
      | do_term _ (AAbs ((_, ty), tm)) = do_type ty #> do_term false tm
blanchet@43839
   480
    fun do_formula (AQuant (_, xs, phi)) =
blanchet@43839
   481
        fold do_type (map_filter snd xs) #> do_formula phi
blanchet@43839
   482
      | do_formula (AConn (_, phis)) = fold do_formula phis
blanchet@43839
   483
      | do_formula (AAtom tm) = do_term true tm
blanchet@43839
   484
    fun do_problem_line (Decl (_, _, ty)) = do_type ty
blanchet@43839
   485
      | do_problem_line (Formula (_, _, phi, _, _)) = do_formula phi
blanchet@43839
   486
  in
blanchet@43839
   487
    fold (fold do_problem_line o snd) problem []
blanchet@43839
   488
    |> filter_out (is_built_in_tptp_symbol o fst o fst)
blanchet@43839
   489
  end
blanchet@43839
   490
blanchet@43839
   491
fun declare_undeclared_syms_in_atp_problem prefix heading problem =
blanchet@43839
   492
  let
blanchet@43839
   493
    fun decl_line (x as (s, _), ty) = Decl (prefix ^ s, x, ty ())
blanchet@43839
   494
    val declared = problem |> declared_syms_in_problem
blanchet@43839
   495
    val decls =
blanchet@43839
   496
      problem |> undeclared_syms_in_problem declared
blanchet@43839
   497
              |> sort_wrt (fst o fst)
blanchet@43839
   498
              |> map decl_line
blanchet@43839
   499
  in (heading, decls) :: problem end
blanchet@43839
   500
blanchet@37643
   501
(** Nice names **)
blanchet@37643
   502
blanchet@37624
   503
fun empty_name_pool readable_names =
blanchet@37643
   504
  if readable_names then SOME (Symtab.empty, Symtab.empty) else NONE
blanchet@37624
   505
blanchet@37624
   506
fun pool_fold f xs z = pair z #> fold_rev (fn x => uncurry (f x)) xs
blanchet@37624
   507
fun pool_map f xs =
blanchet@37624
   508
  pool_fold (fn x => fn ys => fn pool => f x pool |>> (fn y => y :: ys)) xs []
blanchet@37624
   509
blanchet@43088
   510
val no_qualifiers =
blanchet@43088
   511
  let
blanchet@43088
   512
    fun skip [] = []
blanchet@43088
   513
      | skip (#"." :: cs) = skip cs
blanchet@43088
   514
      | skip (c :: cs) = if Char.isAlphaNum c then skip cs else c :: keep cs
blanchet@43088
   515
    and keep [] = []
blanchet@43088
   516
      | keep (#"." :: cs) = skip cs
blanchet@43088
   517
      | keep (c :: cs) = c :: keep cs
blanchet@43088
   518
  in String.explode #> rev #> keep #> rev #> String.implode end
blanchet@43088
   519
blanchet@43626
   520
(* Long names can slow down the ATPs. *)
blanchet@43589
   521
val max_readable_name_size = 20
blanchet@43438
   522
blanchet@43841
   523
(* "equal" is reserved by some ATPs. "op" is also reserved, to avoid the
blanchet@43841
   524
   unreadable "op_1", "op_2", etc., in the problem files. "eq" is reserved to
blanchet@43841
   525
   ensure that "HOL.eq" is correctly mapped to equality (not clear whether this
blanchet@43841
   526
   is still necessary). *)
blanchet@43841
   527
val reserved_nice_names = [tptp_old_equal, "op", "eq"]
blanchet@43780
   528
blanchet@37624
   529
fun readable_name full_name s =
blanchet@37643
   530
  if s = full_name then
blanchet@37643
   531
    s
blanchet@37643
   532
  else
blanchet@43438
   533
    s |> no_qualifiers
blanchet@45266
   534
      |> perhaps (try (unprefix "'"))
blanchet@43438
   535
      |> Name.desymbolize (Char.isUpper (String.sub (full_name, 0)))
blanchet@43454
   536
      |> (fn s =>
blanchet@43530
   537
             if size s > max_readable_name_size then
blanchet@43530
   538
               String.substring (s, 0, max_readable_name_size div 2 - 4) ^
blanchet@44691
   539
               string_of_int (hash_string full_name) ^
blanchet@43530
   540
               String.extract (s, size s - max_readable_name_size div 2 + 4,
blanchet@43530
   541
                               NONE)
blanchet@43454
   542
             else
blanchet@43454
   543
               s)
blanchet@43438
   544
      |> (fn s => if member (op =) reserved_nice_names s then full_name else s)
blanchet@37624
   545
blanchet@37624
   546
fun nice_name (full_name, _) NONE = (full_name, NONE)
blanchet@37624
   547
  | nice_name (full_name, desired_name) (SOME the_pool) =
blanchet@43839
   548
    if is_built_in_tptp_symbol full_name then
blanchet@39630
   549
      (full_name, SOME the_pool)
blanchet@39630
   550
    else case Symtab.lookup (fst the_pool) full_name of
blanchet@37624
   551
      SOME nice_name => (nice_name, SOME the_pool)
blanchet@37624
   552
    | NONE =>
blanchet@37624
   553
      let
blanchet@37624
   554
        val nice_prefix = readable_name full_name desired_name
blanchet@37624
   555
        fun add j =
blanchet@37624
   556
          let
blanchet@43515
   557
            val nice_name =
blanchet@45266
   558
              nice_prefix ^ (if j = 0 then "" else string_of_int j)
blanchet@37624
   559
          in
blanchet@37624
   560
            case Symtab.lookup (snd the_pool) nice_name of
blanchet@37624
   561
              SOME full_name' =>
blanchet@37624
   562
              if full_name = full_name' then (nice_name, the_pool)
blanchet@37624
   563
              else add (j + 1)
blanchet@37624
   564
            | NONE =>
blanchet@37624
   565
              (nice_name,
blanchet@37624
   566
               (Symtab.update_new (full_name, nice_name) (fst the_pool),
blanchet@37624
   567
                Symtab.update_new (nice_name, full_name) (snd the_pool)))
blanchet@37624
   568
          end
blanchet@37624
   569
      in add 0 |> apsnd SOME end
blanchet@37624
   570
blanchet@45457
   571
fun nice_type (AType (name, tys)) =
blanchet@45457
   572
    nice_name name ##>> pool_map nice_type tys #>> AType
blanchet@43804
   573
  | nice_type (AFun (ty1, ty2)) = nice_type ty1 ##>> nice_type ty2 #>> AFun
nik@44535
   574
fun nice_term (ATerm (name, ts)) =
nik@44535
   575
    nice_name name ##>> pool_map nice_term ts #>> ATerm
nik@44535
   576
  | nice_term (AAbs ((name, ty), tm)) =
nik@44535
   577
    nice_name name ##>> nice_type ty ##>> nice_term tm #>> AAbs
blanchet@38198
   578
fun nice_formula (AQuant (q, xs, phi)) =
blanchet@43397
   579
    pool_map nice_name (map fst xs)
blanchet@43397
   580
    ##>> pool_map (fn NONE => pair NONE
blanchet@43804
   581
                    | SOME ty => nice_type ty #>> SOME) (map snd xs)
blanchet@43397
   582
    ##>> nice_formula phi
blanchet@43397
   583
    #>> (fn ((ss, ts), phi) => AQuant (q, ss ~~ ts, phi))
blanchet@38198
   584
  | nice_formula (AConn (c, phis)) =
blanchet@38198
   585
    pool_map nice_formula phis #>> curry AConn c
blanchet@38280
   586
  | nice_formula (AAtom tm) = nice_term tm #>> AAtom
blanchet@43804
   587
fun nice_problem_line (Decl (ident, sym, ty)) =
blanchet@43839
   588
    nice_name sym ##>> nice_type ty #>> (fn (sym, ty) => Decl (ident, sym, ty))
blanchet@43780
   589
  | nice_problem_line (Formula (ident, kind, phi, source, info)) =
blanchet@43780
   590
    nice_formula phi #>> (fn phi => Formula (ident, kind, phi, source, info))
blanchet@38171
   591
fun nice_problem problem =
blanchet@37643
   592
  pool_map (fn (heading, lines) =>
blanchet@38171
   593
               pool_map nice_problem_line lines #>> pair heading) problem
blanchet@39692
   594
fun nice_atp_problem readable_names problem =
blanchet@38251
   595
  nice_problem problem (empty_name_pool readable_names)
blanchet@37509
   596
blanchet@37509
   597
end;