src/HOL/Tools/ATP/atp_systems.ML
author blanchet
Sun Aug 22 09:43:10 2010 +0200 (2010-08-22)
changeset 38631 979a0b37f981
parent 38603 a57d04dd1b25
child 38645 4d5bbec1a598
permissions -rw-r--r--
prefer TPTP "conjecture" tag to "hypothesis" on ATPs where this is possible;
the disjunctive view of "conjecture" is nonstandard but taken by E, SPASS, Vampire, etc.
blanchet@38047
     1
(*  Title:      HOL/Tools/ATP/atp_systems.ML
wenzelm@28592
     2
    Author:     Fabian Immler, TU Muenchen
blanchet@36371
     3
    Author:     Jasmin Blanchette, TU Muenchen
wenzelm@28592
     4
blanchet@36376
     5
Setup for supported ATPs.
wenzelm@28592
     6
*)
wenzelm@28592
     7
blanchet@36376
     8
signature ATP_SYSTEMS =
wenzelm@28592
     9
sig
blanchet@38023
    10
  datatype failure =
blanchet@38519
    11
    Unprovable | IncompleteUnprovable | CantConnect | TimedOut |
blanchet@38519
    12
    OutOfResources | SpassTooOld | VampireTooOld | NoPerl | NoLibwwwPerl |
blanchet@38519
    13
    MalformedInput | MalformedOutput | UnknownError
blanchet@38023
    14
blanchet@38023
    15
  type prover_config =
blanchet@38092
    16
    {exec: string * string,
blanchet@38092
    17
     required_execs: (string * string) list,
blanchet@38023
    18
     arguments: bool -> Time.time -> string,
blanchet@38023
    19
     proof_delims: (string * string) list,
blanchet@38023
    20
     known_failures: (failure * string) list,
blanchet@38589
    21
     default_max_relevant_per_iter: int,
blanchet@38589
    22
     default_theory_relevant: bool,
blanchet@38631
    23
     explicit_forall: bool,
blanchet@38631
    24
     use_conjecture_for_hypotheses: bool}
blanchet@38023
    25
blanchet@38061
    26
  val string_for_failure : failure -> string
blanchet@38061
    27
  val known_failure_in_output :
blanchet@38061
    28
    string -> (failure * string) list -> failure option
blanchet@38023
    29
  val add_prover: string * prover_config -> theory -> theory
blanchet@38023
    30
  val get_prover: theory -> string -> prover_config
blanchet@38023
    31
  val available_atps: theory -> unit
blanchet@35867
    32
  val refresh_systems_on_tptp : unit -> unit
blanchet@36371
    33
  val default_atps_param_value : unit -> string
blanchet@35867
    34
  val setup : theory -> theory
wenzelm@28592
    35
end;
wenzelm@28592
    36
blanchet@36376
    37
structure ATP_Systems : ATP_SYSTEMS =
wenzelm@28592
    38
struct
wenzelm@28596
    39
blanchet@38023
    40
(* prover configuration *)
boehmes@32864
    41
blanchet@38023
    42
datatype failure =
blanchet@38454
    43
  Unprovable | IncompleteUnprovable | CantConnect | TimedOut | OutOfResources |
blanchet@38519
    44
  SpassTooOld | VampireTooOld | NoPerl | NoLibwwwPerl | MalformedInput |
blanchet@38454
    45
  MalformedOutput | UnknownError
boehmes@32864
    46
wenzelm@32941
    47
type prover_config =
blanchet@38092
    48
  {exec: string * string,
blanchet@38092
    49
   required_execs: (string * string) list,
blanchet@37514
    50
   arguments: bool -> Time.time -> string,
blanchet@36370
    51
   proof_delims: (string * string) list,
blanchet@36370
    52
   known_failures: (failure * string) list,
blanchet@38589
    53
   default_max_relevant_per_iter: int,
blanchet@38589
    54
   default_theory_relevant: bool,
blanchet@38631
    55
   explicit_forall: bool,
blanchet@38631
    56
   use_conjecture_for_hypotheses: bool}
wenzelm@28596
    57
blanchet@38061
    58
val missing_message_tail =
blanchet@38061
    59
  " appears to be missing. You will need to install it if you want to run \
blanchet@38061
    60
  \ATPs remotely."
blanchet@38061
    61
blanchet@38061
    62
fun string_for_failure Unprovable = "The ATP problem is unprovable."
blanchet@38061
    63
  | string_for_failure IncompleteUnprovable =
blanchet@38061
    64
    "The ATP cannot prove the problem."
blanchet@38094
    65
  | string_for_failure CantConnect = "Can't connect to remote server."
blanchet@38061
    66
  | string_for_failure TimedOut = "Timed out."
blanchet@38061
    67
  | string_for_failure OutOfResources = "The ATP ran out of resources."
blanchet@38519
    68
  | string_for_failure SpassTooOld =
blanchet@38096
    69
    "Isabelle requires a more recent version of SPASS with support for the \
blanchet@38096
    70
    \TPTP syntax. To install it, download and extract the package \
blanchet@38096
    71
    \\"http://isabelle.in.tum.de/dist/contrib/spass-3.7.tar.gz\" and add the \
blanchet@38096
    72
    \\"spass-3.7\" directory's absolute path to " ^
blanchet@38061
    73
    quote (Path.implode (Path.expand (Path.appends
blanchet@38061
    74
               (Path.variable "ISABELLE_HOME_USER" ::
blanchet@38061
    75
                map Path.basic ["etc", "components"])))) ^
blanchet@38061
    76
    " on a line of its own."
blanchet@38519
    77
  | string_for_failure VampireTooOld =
blanchet@38454
    78
    "Isabelle requires a more recent version of Vampire. To install it, follow \
blanchet@38454
    79
    \the instructions from the Sledgehammer manual (\"isabelle doc\
blanchet@38454
    80
    \ sledgehammer\")."
blanchet@38061
    81
  | string_for_failure NoPerl = "Perl" ^ missing_message_tail
blanchet@38061
    82
  | string_for_failure NoLibwwwPerl =
blanchet@38061
    83
    "The Perl module \"libwww-perl\"" ^ missing_message_tail
blanchet@38061
    84
  | string_for_failure MalformedInput =
blanchet@38096
    85
    "The ATP problem is malformed. Please report this to the Isabelle \
blanchet@38096
    86
    \developers."
blanchet@38096
    87
  | string_for_failure MalformedOutput = "The ATP output is malformed."
blanchet@38096
    88
  | string_for_failure UnknownError = "An unknown ATP error occurred."
blanchet@38061
    89
blanchet@38061
    90
fun known_failure_in_output output =
blanchet@38061
    91
  find_first (fn (_, pattern) => String.isSubstring pattern output)
blanchet@38061
    92
  #> Option.map fst
blanchet@38061
    93
blanchet@38061
    94
val known_perl_failures =
blanchet@38094
    95
  [(CantConnect, "HTTP error"),
blanchet@38094
    96
   (NoPerl, "env: perl"),
blanchet@38065
    97
   (NoLibwwwPerl, "Can't locate HTTP")]
wenzelm@28596
    98
blanchet@38023
    99
(* named provers *)
blanchet@38001
   100
blanchet@38023
   101
structure Data = Theory_Data
blanchet@38023
   102
(
blanchet@38023
   103
  type T = (prover_config * stamp) Symtab.table
blanchet@38023
   104
  val empty = Symtab.empty
blanchet@38023
   105
  val extend = I
blanchet@38023
   106
  fun merge data : T = Symtab.merge (eq_snd op =) data
blanchet@38023
   107
    handle Symtab.DUP name => error ("Duplicate ATP: " ^ quote name ^ ".")
blanchet@38023
   108
)
blanchet@38017
   109
blanchet@38023
   110
fun add_prover (name, config) thy =
blanchet@38023
   111
  Data.map (Symtab.update_new (name, (config, stamp ()))) thy
blanchet@38023
   112
  handle Symtab.DUP name => error ("Duplicate ATP: " ^ quote name ^ ".")
blanchet@38017
   113
blanchet@38023
   114
fun get_prover thy name =
blanchet@38023
   115
  the (Symtab.lookup (Data.get thy) name) |> fst
blanchet@38023
   116
  handle Option.Option => error ("Unknown ATP: " ^ name ^ ".")
blanchet@37962
   117
blanchet@38023
   118
fun available_atps thy =
blanchet@38023
   119
  priority ("Available ATPs: " ^
blanchet@38023
   120
            commas (sort_strings (Symtab.keys (Data.get thy))) ^ ".")
boehmes@32864
   121
blanchet@38023
   122
fun available_atps thy =
blanchet@38023
   123
  priority ("Available ATPs: " ^
blanchet@38023
   124
            commas (sort_strings (Symtab.keys (Data.get thy))) ^ ".")
wenzelm@28596
   125
blanchet@36382
   126
fun to_generous_secs time = (Time.toMilliseconds time + 999) div 1000
blanchet@36142
   127
wenzelm@28596
   128
(* E prover *)
wenzelm@28596
   129
blanchet@36369
   130
val tstp_proof_delims =
blanchet@36369
   131
  ("# SZS output start CNFRefutation.", "# SZS output end CNFRefutation")
blanchet@36369
   132
blanchet@35969
   133
val e_config : prover_config =
blanchet@38092
   134
  {exec = ("E_HOME", "eproof"),
blanchet@38092
   135
   required_execs = [],
blanchet@37514
   136
   arguments = fn _ => fn timeout =>
blanchet@36382
   137
     "--tstp-in --tstp-out -l5 -xAutoDev -tAutoDev --silent --cpu-limit=" ^
blanchet@36382
   138
     string_of_int (to_generous_secs timeout),
blanchet@36369
   139
   proof_delims = [tstp_proof_delims],
blanchet@36265
   140
   known_failures =
blanchet@37995
   141
     [(Unprovable, "SZS status: CounterSatisfiable"),
blanchet@37995
   142
      (Unprovable, "SZS status CounterSatisfiable"),
blanchet@36370
   143
      (TimedOut, "Failure: Resource limit exceeded (time)"),
blanchet@36370
   144
      (TimedOut, "time limit exceeded"),
blanchet@36370
   145
      (OutOfResources,
blanchet@36370
   146
       "# Cannot determine problem status within resource limit"),
blanchet@36370
   147
      (OutOfResources, "SZS status: ResourceOut"),
blanchet@36370
   148
      (OutOfResources, "SZS status ResourceOut")],
blanchet@38589
   149
   default_max_relevant_per_iter = 50 (* FIXME *),
blanchet@38589
   150
   default_theory_relevant = false,
blanchet@38631
   151
   explicit_forall = false,
blanchet@38631
   152
   use_conjecture_for_hypotheses = true}
blanchet@38454
   153
blanchet@38023
   154
val e = ("e", e_config)
wenzelm@28596
   155
wenzelm@28596
   156
blanchet@36219
   157
(* The "-VarWeight=3" option helps the higher-order problems, probably by
blanchet@36219
   158
   counteracting the presence of "hAPP". *)
blanchet@37498
   159
val spass_config : prover_config =
blanchet@38092
   160
  {exec = ("ISABELLE_ATP", "scripts/spass"),
blanchet@38092
   161
   required_execs = [("SPASS_HOME", "SPASS")],
blanchet@37550
   162
   (* "div 2" accounts for the fact that SPASS is often run twice. *)
blanchet@37514
   163
   arguments = fn complete => fn timeout =>
blanchet@37962
   164
     ("-Auto -PGiven=0 -PProblem=0 -Splits=0 -FullRed=0 -DocProof \
blanchet@37550
   165
      \-VarWeight=3 -TimeLimit=" ^
blanchet@38094
   166
      string_of_int ((to_generous_secs timeout + 1) div 2))
blanchet@37514
   167
     |> not complete ? prefix "-SOS=1 ",
blanchet@36369
   168
   proof_delims = [("Here is a proof", "Formulae used in the proof")],
blanchet@36289
   169
   known_failures =
blanchet@38061
   170
     known_perl_failures @
blanchet@37413
   171
     [(IncompleteUnprovable, "SPASS beiseite: Completion found"),
blanchet@36370
   172
      (TimedOut, "SPASS beiseite: Ran out of time"),
blanchet@36965
   173
      (OutOfResources, "SPASS beiseite: Maximal number of loops exceeded"),
blanchet@37413
   174
      (MalformedInput, "Undefined symbol"),
blanchet@37414
   175
      (MalformedInput, "Free Variable"),
blanchet@38519
   176
      (SpassTooOld, "tptp2dfg")],
blanchet@38589
   177
   default_max_relevant_per_iter = 35 (* FIXME *),
blanchet@38589
   178
   default_theory_relevant = true,
blanchet@38631
   179
   explicit_forall = true,
blanchet@38631
   180
   use_conjecture_for_hypotheses = true}
blanchet@38454
   181
blanchet@38023
   182
val spass = ("spass", spass_config)
wenzelm@28596
   183
blanchet@38454
   184
blanchet@37509
   185
(* Vampire *)
blanchet@37509
   186
blanchet@37509
   187
val vampire_config : prover_config =
blanchet@38092
   188
  {exec = ("VAMPIRE_HOME", "vampire"),
blanchet@38092
   189
   required_execs = [],
blanchet@37514
   190
   arguments = fn _ => fn timeout =>
blanchet@38033
   191
     "--mode casc -t " ^ string_of_int (to_generous_secs timeout) ^
blanchet@38588
   192
     " --thanks Andrei --input_file",
blanchet@37509
   193
   proof_delims =
blanchet@37509
   194
     [("=========== Refutation ==========",
blanchet@37509
   195
       "======= End of refutation ======="),
blanchet@38033
   196
      ("% SZS output start Refutation", "% SZS output end Refutation"),
blanchet@38033
   197
      ("% SZS output start Proof", "% SZS output end Proof")],
blanchet@37509
   198
   known_failures =
blanchet@37509
   199
     [(Unprovable, "UNPROVABLE"),
blanchet@37509
   200
      (IncompleteUnprovable, "CANNOT PROVE"),
blanchet@38092
   201
      (TimedOut, "SZS status Timeout"),
blanchet@37509
   202
      (Unprovable, "Satisfiability detected"),
blanchet@38519
   203
      (VampireTooOld, "not a valid option")],
blanchet@38589
   204
   default_max_relevant_per_iter = 45 (* FIXME *),
blanchet@38589
   205
   default_theory_relevant = false,
blanchet@38631
   206
   explicit_forall = false,
blanchet@38631
   207
   use_conjecture_for_hypotheses = true}
blanchet@38454
   208
blanchet@38023
   209
val vampire = ("vampire", vampire_config)
blanchet@37509
   210
blanchet@38454
   211
blanchet@37509
   212
(* Remote prover invocation via SystemOnTPTP *)
wenzelm@28596
   213
blanchet@38061
   214
val systems = Synchronized.var "atp_systems" ([] : string list)
immler@31835
   215
immler@31835
   216
fun get_systems () =
blanchet@38061
   217
  case bash_output "\"$ISABELLE_ATP/scripts/remote_atp\" -w 2>&1" of
blanchet@36370
   218
    (answer, 0) => split_lines answer
blanchet@36370
   219
  | (answer, _) =>
blanchet@38065
   220
    error (case known_failure_in_output answer known_perl_failures of
blanchet@38065
   221
             SOME failure => string_for_failure failure
blanchet@38065
   222
           | NONE => perhaps (try (unsuffix "\n")) answer ^ ".")
immler@31835
   223
blanchet@35867
   224
fun refresh_systems_on_tptp () =
blanchet@37509
   225
  Synchronized.change systems (fn _ => get_systems ())
immler@31835
   226
blanchet@38589
   227
fun get_system prefix =
blanchet@38589
   228
  Synchronized.change_result systems
blanchet@38589
   229
      (fn systems => (if null systems then get_systems () else systems)
blanchet@38589
   230
                     |> `(find_first (String.isPrefix prefix)))
boehmes@32864
   231
wenzelm@32948
   232
fun the_system prefix =
boehmes@32864
   233
  (case get_system prefix of
blanchet@37509
   234
    NONE => error ("System " ^ quote prefix ^ " not available at SystemOnTPTP.")
wenzelm@32942
   235
  | SOME sys => sys);
immler@31835
   236
blanchet@38598
   237
fun remote_config system_prefix proof_delims known_failures
blanchet@38631
   238
                  default_max_relevant_per_iter default_theory_relevant
blanchet@38631
   239
                  use_conjecture_for_hypotheses =
blanchet@38092
   240
  {exec = ("ISABELLE_ATP", "scripts/remote_atp"),
blanchet@38092
   241
   required_execs = [],
blanchet@37514
   242
   arguments = fn _ => fn timeout =>
blanchet@38041
   243
     " -t " ^ string_of_int (to_generous_secs timeout) ^ " -s " ^
blanchet@38598
   244
     the_system system_prefix,
blanchet@36369
   245
   proof_delims = insert (op =) tstp_proof_delims proof_delims,
blanchet@38061
   246
   known_failures =
blanchet@38061
   247
     known_failures @ known_perl_failures @
blanchet@38094
   248
     [(TimedOut, "says Timeout")],
blanchet@38589
   249
   default_max_relevant_per_iter = default_max_relevant_per_iter,
blanchet@38589
   250
   default_theory_relevant = default_theory_relevant,
blanchet@38631
   251
   explicit_forall = true,
blanchet@38631
   252
   use_conjecture_for_hypotheses = use_conjecture_for_hypotheses}
wenzelm@28596
   253
blanchet@38598
   254
fun remotify_config system_prefix
blanchet@38598
   255
        ({proof_delims, known_failures, default_max_relevant_per_iter,
blanchet@38631
   256
          default_theory_relevant, use_conjecture_for_hypotheses, ...}
blanchet@38631
   257
         : prover_config) : prover_config =
blanchet@38598
   258
  remote_config system_prefix proof_delims known_failures
blanchet@38598
   259
                default_max_relevant_per_iter default_theory_relevant
blanchet@38631
   260
                use_conjecture_for_hypotheses
blanchet@38023
   261
blanchet@38598
   262
val remotify_name = prefix "remote_"
blanchet@38598
   263
fun remote_prover name system_prefix proof_delims known_failures
blanchet@38631
   264
                  default_max_relevant_per_iter default_theory_relevant
blanchet@38631
   265
                  use_conjecture_for_hypotheses =
blanchet@38598
   266
  (remotify_name name,
blanchet@38598
   267
   remote_config system_prefix proof_delims known_failures
blanchet@38631
   268
                 default_max_relevant_per_iter default_theory_relevant
blanchet@38631
   269
                 use_conjecture_for_hypotheses)
blanchet@38598
   270
fun remotify_prover (name, config) system_prefix =
blanchet@38598
   271
  (remotify_name name, remotify_config system_prefix config)
wenzelm@28592
   272
blanchet@38598
   273
val remote_e = remotify_prover e "EP---"
blanchet@38598
   274
val remote_vampire = remotify_prover vampire "Vampire---9"
blanchet@38603
   275
val remote_sine_e =
blanchet@38603
   276
  remote_prover "sine_e" "SInE---" []
blanchet@38631
   277
                [(Unprovable, "says Unknown")] 150 (* FIXME *) false true
blanchet@38598
   278
val remote_snark =
blanchet@38598
   279
  remote_prover "snark" "SNARK---" [("refutation.", "end_refutation.")] []
blanchet@38631
   280
                50 (* FIXME *) false true
blanchet@38454
   281
blanchet@38454
   282
(* Setup *)
blanchet@38454
   283
blanchet@38092
   284
fun is_installed ({exec, required_execs, ...} : prover_config) =
blanchet@38092
   285
  forall (curry (op <>) "" o getenv o fst) (exec :: required_execs)
blanchet@38041
   286
fun maybe_remote (name, config) =
blanchet@38598
   287
  name |> not (is_installed config) ? remotify_name
blanchet@36371
   288
blanchet@36371
   289
fun default_atps_param_value () =
blanchet@38041
   290
  space_implode " " ([maybe_remote e] @
blanchet@38041
   291
                     (if is_installed (snd spass) then [fst spass] else []) @
blanchet@38603
   292
                     [if forall (is_installed o snd) [e, spass] then
blanchet@38603
   293
                        remotify_name (fst vampire)
blanchet@38603
   294
                      else
blanchet@38603
   295
                        maybe_remote vampire,
blanchet@38603
   296
                      fst remote_sine_e])
blanchet@36371
   297
blanchet@38598
   298
val provers = [e, spass, vampire, remote_e, remote_vampire, remote_sine_e,
blanchet@38598
   299
               remote_snark]
blanchet@38023
   300
val setup = fold add_prover provers
blanchet@35867
   301
wenzelm@28592
   302
end;