src/HOL/TPTP/mash_export.ML
author blanchet
Wed, 21 Aug 2013 09:25:40 +0200
changeset 53120 43d5f3d6d04e
parent 52196 2281f33e8da6
child 53121 5f727525b1ac
permissions -rw-r--r--
generate max suggestions in MaSh export driver
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     1
(*  Title:      HOL/TPTP/mash_export.ML
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     2
    Author:     Jasmin Blanchette, TU Muenchen
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     3
    Copyright   2012
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     4
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     5
Export Isabelle theory information for MaSh (Machine-learning for Sledgehammer).
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     6
*)
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     7
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     8
signature MASH_EXPORT =
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     9
sig
48251
6cdcfbddc077 moved most of MaSh exporter code to Sledgehammer
blanchet
parents: 48250
diff changeset
    10
  type params = Sledgehammer_Provers.params
48235
40655464a93b MaSh evaluation driver
blanchet
parents: 48234
diff changeset
    11
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    12
  val generate_accessibility :
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    13
    Proof.context -> theory list -> bool -> string -> unit
48318
325c8fd0d762 more consolidation of MaSh code
blanchet
parents: 48316
diff changeset
    14
  val generate_features :
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    15
    Proof.context -> string -> theory list -> string -> unit
48333
2250197977dc repair MaSh exporter
blanchet
parents: 48324
diff changeset
    16
  val generate_isar_dependencies :
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    17
    Proof.context -> theory list -> bool -> string -> unit
50484
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
    18
  val generate_prover_dependencies :
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
    19
    Proof.context -> params -> int * int option -> theory list -> bool -> string
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
    20
    -> unit
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
    21
  val generate_isar_commands :
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    22
    Proof.context -> string -> (int * int option) * int -> theory list -> bool
53120
43d5f3d6d04e generate max suggestions in MaSh export driver
blanchet
parents: 52196
diff changeset
    23
    -> int -> string -> unit
50484
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
    24
  val generate_prover_commands :
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    25
    Proof.context -> params -> (int * int option) * int -> theory list -> bool
53120
43d5f3d6d04e generate max suggestions in MaSh export driver
blanchet
parents: 52196
diff changeset
    26
    -> int -> string -> unit
48379
2b5ad61e2ccc renamed "iter" fact filter to "MePo" (Meng--Paulson)
blanchet
parents: 48378
diff changeset
    27
  val generate_mepo_suggestions :
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    28
    Proof.context -> params -> (int * int option) * int -> theory list -> bool
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    29
    -> int -> string -> unit
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
    30
  val generate_mesh_suggestions : int -> string -> string -> string -> unit
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    31
end;
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    32
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    33
structure MaSh_Export : MASH_EXPORT =
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    34
struct
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    35
50485
3c6ac2da2f45 merge aliased theorems in MaSh dependencies, modulo symmetry of equality
blanchet
parents: 50484
diff changeset
    36
open Sledgehammer_Fact
48381
1b7d798460bb renamed ML structures
blanchet
parents: 48379
diff changeset
    37
open Sledgehammer_MePo
1b7d798460bb renamed ML structures
blanchet
parents: 48379
diff changeset
    38
open Sledgehammer_MaSh
48245
854a47677335 generate ATP dependencies
blanchet
parents: 48242
diff changeset
    39
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
    40
fun in_range (from, to) j =
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
    41
  j >= from andalso (to = NONE orelse j <= the to)
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
    42
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    43
fun has_thm_thy th thy =
48316
252f45c04042 drastic overhaul of MaSh data structures + fixed a few performance issues
blanchet
parents: 48315
diff changeset
    44
  Context.theory_name thy = Context.theory_name (theory_of_thm th)
252f45c04042 drastic overhaul of MaSh data structures + fixed a few performance issues
blanchet
parents: 48315
diff changeset
    45
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    46
fun has_thys thys th = exists (has_thm_thy th) thys
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    47
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    48
fun all_facts ctxt =
48531
7da5d3b8aef4 don't export technical theorems for MaSh
blanchet
parents: 48530
diff changeset
    49
  let val css = Sledgehammer_Fact.clasimpset_rule_table_of ctxt in
50442
4f6a4d32522c don't blacklist "case" theorems -- this causes problems in MaSh later
blanchet
parents: 50434
diff changeset
    50
    Sledgehammer_Fact.all_facts ctxt true false Symtab.empty [] [] css
51135
e32114b25551 tuned code
blanchet
parents: 51034
diff changeset
    51
    |> sort (crude_thm_ord o pairself snd)
48531
7da5d3b8aef4 don't export technical theorems for MaSh
blanchet
parents: 48530
diff changeset
    52
  end
7da5d3b8aef4 don't export technical theorems for MaSh
blanchet
parents: 48530
diff changeset
    53
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    54
fun filter_accessible_from th = filter (fn (_, th') => thm_less (th', th))
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    55
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    56
fun generate_accessibility ctxt thys linearize file_name =
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    57
  let
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    58
    val path = file_name |> Path.explode
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    59
    val _ = File.write path ""
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    60
    fun do_fact (parents, fact) prevs =
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    61
      let
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    62
        val parents = if linearize then prevs else parents
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    63
        val s = encode_str fact ^ ": " ^ encode_strs parents ^ "\n"
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    64
        val _ = File.append path s
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    65
      in [fact] end
50611
99af6b652b3a linearize eval driver, to work around horrible bug in previous implementation
blanchet
parents: 50582
diff changeset
    66
    val facts =
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
    67
      all_facts ctxt
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    68
      |> filter_out (has_thys thys o snd)
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    69
      |> attach_parents_to_facts []
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    70
      |> map (apsnd (nickname_of_thm o snd))
50611
99af6b652b3a linearize eval driver, to work around horrible bug in previous implementation
blanchet
parents: 50582
diff changeset
    71
  in fold do_fact facts []; () end
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    72
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    73
fun generate_features ctxt prover thys file_name =
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    74
  let
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    75
    val path = file_name |> Path.explode
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    76
    val _ = File.write path ""
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
    77
    val facts = all_facts ctxt |> filter_out (has_thys thys o snd)
48385
2779dea0b1e0 added locality as a MaSh feature
blanchet
parents: 48381
diff changeset
    78
    fun do_fact ((_, stature), th) =
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    79
      let
50624
4d0997abce79 improved thm order hack, in case the default names are overridden
blanchet
parents: 50611
diff changeset
    80
        val name = nickname_of_thm th
48318
325c8fd0d762 more consolidation of MaSh code
blanchet
parents: 48316
diff changeset
    81
        val feats =
48385
2779dea0b1e0 added locality as a MaSh feature
blanchet
parents: 48381
diff changeset
    82
          features_of ctxt prover (theory_of_thm th) stature [prop_of th]
50582
001a0e12d7f1 tuned order to help debugging
blanchet
parents: 50561
diff changeset
    83
        val s =
50826
18ace05656cf start using MaSh hints
blanchet
parents: 50814
diff changeset
    84
          encode_str name ^ ": " ^ encode_features (sort_wrt fst feats) ^ "\n"
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    85
      in File.append path s end
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    86
  in List.app do_fact facts end
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
    87
51034
0ee6039d2c8e distinguish one more kind of proofs
blanchet
parents: 51033
diff changeset
    88
val prover_marker = "$a"
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    89
val isar_marker = "$i"
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    90
val omitted_marker = "$o"
51034
0ee6039d2c8e distinguish one more kind of proofs
blanchet
parents: 51033
diff changeset
    91
val unprovable_marker = "$u" (* axiom or definition or characteristic theorem *)
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    92
val prover_failed_marker = "$x"
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    93
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    94
fun smart_dependencies_of ctxt params_opt facts name_tabs th isar_deps_opt =
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    95
  let
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    96
    val (marker, deps) =
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    97
      case params_opt of
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    98
        SOME (params as {provers = prover :: _, ...}) =>
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
    99
        prover_dependencies_of ctxt params prover 0 facts name_tabs th
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   100
        |>> (fn true => prover_marker | false => prover_failed_marker)
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   101
      | NONE =>
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   102
        let
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   103
          val deps =
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   104
            case isar_deps_opt of
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   105
              SOME deps => deps
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   106
            | NONE => isar_dependencies_of name_tabs th
51034
0ee6039d2c8e distinguish one more kind of proofs
blanchet
parents: 51033
diff changeset
   107
        in (if null deps then unprovable_marker else isar_marker, deps) end
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   108
  in
51177
e8c9755fd14e tuned code: factored out parent computation
blanchet
parents: 51135
diff changeset
   109
    case trim_dependencies deps of
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   110
      SOME deps => (marker, deps)
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   111
    | NONE => (omitted_marker, [])
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   112
  end
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   113
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   114
fun generate_isar_or_prover_dependencies ctxt params_opt range thys linearize
50484
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
   115
                                         file_name =
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
   116
  let
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
   117
    val path = file_name |> Path.explode
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   118
    val facts = all_facts ctxt |> filter_out (has_thys thys o snd)
50735
6b232d76cbc9 refined class handling, to prevent cycles in fact graph
blanchet
parents: 50624
diff changeset
   119
    val name_tabs = build_name_tables nickname_of_thm facts
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   120
    fun do_fact (j, (_, th)) =
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   121
      if in_range range j then
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   122
        let
50624
4d0997abce79 improved thm order hack, in case the default names are overridden
blanchet
parents: 50611
diff changeset
   123
          val name = nickname_of_thm th
50561
9a733bd6c0ba added tracing to ATP exporter
blanchet
parents: 50559
diff changeset
   124
          val _ = tracing ("Fact " ^ string_of_int j ^ ": " ^ name)
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   125
          val access_facts =
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   126
            if linearize then take (j - 1) facts
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   127
            else facts |> filter_accessible_from th
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   128
          val (marker, deps) =
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   129
            smart_dependencies_of ctxt params_opt access_facts name_tabs th NONE
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   130
        in encode_str name ^ ": " ^ marker ^ " " ^ encode_strs deps ^ "\n" end
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   131
      else
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   132
        ""
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   133
    val lines = Par_List.map do_fact (tag_list 1 facts)
50519
2951841ec011 parallelized MaSh exporter
blanchet
parents: 50515
diff changeset
   134
  in File.write_list path lines end
48304
50e64af9c829 more work on MaSh
blanchet
parents: 48303
diff changeset
   135
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   136
fun generate_isar_dependencies ctxt =
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   137
  generate_isar_or_prover_dependencies ctxt NONE (1, NONE)
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   138
50484
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
   139
fun generate_prover_dependencies ctxt params =
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
   140
  generate_isar_or_prover_dependencies ctxt (SOME params)
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   141
50954
7bc58677860e added step to skip some queries
blanchet
parents: 50907
diff changeset
   142
fun is_bad_query ctxt ho_atp step j th isar_deps =
7bc58677860e added step to skip some queries
blanchet
parents: 50907
diff changeset
   143
  j mod step <> 0 orelse
50859
c0f38015a632 don't generate queries with empty dependency list
blanchet
parents: 50829
diff changeset
   144
  Thm.legacy_get_kind th = "" orelse
51020
242cd1632b0b removed spurious trimming
blanchet
parents: 50965
diff changeset
   145
  null isar_deps orelse
50523
0799339fea0f get rid of some junk facts in the MaSh evaluation driver
blanchet
parents: 50519
diff changeset
   146
  is_blacklisted_or_something ctxt ho_atp (Thm.get_name_hint th)
50515
c4a27ab89c9b shared bad MaSh query detection between MePo and MaSh, so that the generated files mirror each other
blanchet
parents: 50511
diff changeset
   147
50954
7bc58677860e added step to skip some queries
blanchet
parents: 50907
diff changeset
   148
fun generate_isar_or_prover_commands ctxt prover params_opt (range, step) thys
53120
43d5f3d6d04e generate max suggestions in MaSh export driver
blanchet
parents: 52196
diff changeset
   149
                                     linearize max_suggs file_name =
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
   150
  let
50511
8825c36cb1ce don't query blacklisted theorems in evaluation driver
blanchet
parents: 50485
diff changeset
   151
    val ho_atp = Sledgehammer_Provers.is_ho_atp ctxt prover
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
   152
    val path = file_name |> Path.explode
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
   153
    val facts = all_facts ctxt
50485
3c6ac2da2f45 merge aliased theorems in MaSh dependencies, modulo symmetry of equality
blanchet
parents: 50484
diff changeset
   154
    val (new_facts, old_facts) = facts |> List.partition (has_thys thys o snd)
50735
6b232d76cbc9 refined class handling, to prevent cycles in fact graph
blanchet
parents: 50624
diff changeset
   155
    val name_tabs = build_name_tables nickname_of_thm facts
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   156
    fun do_fact (j, ((name, (parents, ((_, stature), th))), prevs)) =
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   157
      if in_range range j then
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   158
        let
50561
9a733bd6c0ba added tracing to ATP exporter
blanchet
parents: 50559
diff changeset
   159
          val _ = tracing ("Fact " ^ string_of_int j ^ ": " ^ name)
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   160
          val feats =
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   161
            features_of ctxt prover (theory_of_thm th) stature [prop_of th]
50735
6b232d76cbc9 refined class handling, to prevent cycles in fact graph
blanchet
parents: 50624
diff changeset
   162
          val isar_deps = isar_dependencies_of name_tabs th
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   163
          val access_facts =
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   164
            (if linearize then take (j - 1) new_facts
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   165
             else new_facts |> filter_accessible_from th) @ old_facts
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   166
          val (marker, deps) =
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   167
            smart_dependencies_of ctxt params_opt access_facts name_tabs th
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   168
                                  (SOME isar_deps)
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   169
          val parents = if linearize then prevs else parents
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   170
          val core =
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   171
            encode_str name ^ ": " ^ encode_strs parents ^ "; " ^
50582
001a0e12d7f1 tuned order to help debugging
blanchet
parents: 50561
diff changeset
   172
            encode_features (sort_wrt fst feats)
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   173
          val query =
50954
7bc58677860e added step to skip some queries
blanchet
parents: 50907
diff changeset
   174
            if is_bad_query ctxt ho_atp step j th isar_deps then ""
53120
43d5f3d6d04e generate max suggestions in MaSh export driver
blanchet
parents: 52196
diff changeset
   175
            else "? " ^ string_of_int max_suggs ^ " # " ^ core ^ "\n"
50754
74a6adcb96ac also generate queries for goals with too many Isar dependencies
blanchet
parents: 50735
diff changeset
   176
          val update =
51033
177db6811f11 added markers in proofs identifying origin of proofs, in eval driver
blanchet
parents: 51020
diff changeset
   177
            "! " ^ core ^ "; " ^ marker ^ " " ^ encode_strs deps ^ "\n"
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   178
        in query ^ update end
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   179
      else
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   180
        ""
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   181
    val new_facts =
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   182
      new_facts |> attach_parents_to_facts old_facts
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   183
                |> map (`(nickname_of_thm o snd o snd))
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   184
    val hd_prevs =
50624
4d0997abce79 improved thm order hack, in case the default names are overridden
blanchet
parents: 50611
diff changeset
   185
      map (nickname_of_thm o snd) (the_list (try List.last old_facts))
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   186
    val prevss = fst (split_last (hd_prevs :: map (single o fst) new_facts))
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50523
diff changeset
   187
    val lines = Par_List.map do_fact (tag_list 1 (new_facts ~~ prevss))
50519
2951841ec011 parallelized MaSh exporter
blanchet
parents: 50515
diff changeset
   188
  in File.write_list path lines end
48239
0016290f904c generate Meng--Paulson facts for evaluation purposes
blanchet
parents: 48235
diff changeset
   189
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   190
fun generate_isar_commands ctxt prover =
50954
7bc58677860e added step to skip some queries
blanchet
parents: 50907
diff changeset
   191
  generate_isar_or_prover_commands ctxt prover NONE
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   192
50484
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
   193
fun generate_prover_commands ctxt (params as {provers = prover :: _, ...}) =
8ec31bdb9d36 adopt the neutral "prover" terminology for MaSh rather than the ambiguous/wrong ATP terminology (which sometimes excludes SMT solvers)
blanchet
parents: 50442
diff changeset
   194
  generate_isar_or_prover_commands ctxt prover (SOME params)
50411
c9023d78d1a6 export ATP and Isar commands separately
blanchet
parents: 50402
diff changeset
   195
50906
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   196
fun generate_mepo_suggestions ctxt (params as {provers = prover :: _, ...})
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   197
                              (range, step) thys linearize max_suggs file_name =
48239
0016290f904c generate Meng--Paulson facts for evaluation purposes
blanchet
parents: 48235
diff changeset
   198
  let
50515
c4a27ab89c9b shared bad MaSh query detection between MePo and MaSh, so that the generated files mirror each other
blanchet
parents: 50511
diff changeset
   199
    val ho_atp = Sledgehammer_Provers.is_ho_atp ctxt prover
48239
0016290f904c generate Meng--Paulson facts for evaluation purposes
blanchet
parents: 48235
diff changeset
   200
    val path = file_name |> Path.explode
50349
b79803ee14f3 generalized MaSh exporter to sets of theories
blanchet
parents: 48667
diff changeset
   201
    val facts = all_facts ctxt
50485
3c6ac2da2f45 merge aliased theorems in MaSh dependencies, modulo symmetry of equality
blanchet
parents: 50484
diff changeset
   202
    val (new_facts, old_facts) = facts |> List.partition (has_thys thys o snd)
50735
6b232d76cbc9 refined class handling, to prevent cycles in fact graph
blanchet
parents: 50624
diff changeset
   203
    val name_tabs = build_name_tables nickname_of_thm facts
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   204
    fun do_fact (j, ((_, th), old_facts)) =
50906
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   205
      if in_range range j then
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   206
        let
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   207
          val name = nickname_of_thm th
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   208
          val _ = tracing ("Fact " ^ string_of_int j ^ ": " ^ name)
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   209
          val goal = goal_of_thm (Proof_Context.theory_of ctxt) th
52196
2281f33e8da6 redid rac7830871177 to avoid duplicate fixed variable (e.g. lemma "P (a::nat)" proof - have "!!a::int. Q a" sledgehammer [e])
blanchet
parents: 52125
diff changeset
   210
          val (_, hyp_ts, concl_t) = ATP_Util.strip_subgoal goal 1 ctxt
50906
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   211
          val isar_deps = isar_dependencies_of name_tabs th
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   212
        in
50954
7bc58677860e added step to skip some queries
blanchet
parents: 50907
diff changeset
   213
          if is_bad_query ctxt ho_atp step j th isar_deps then
50906
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   214
            ""
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   215
          else
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   216
            let
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   217
              val suggs =
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   218
                old_facts
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51177
diff changeset
   219
                |> linearize ? filter_accessible_from th
50906
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   220
                |> Sledgehammer_MePo.mepo_suggested_facts ctxt params prover
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   221
                       max_suggs NONE hyp_ts concl_t
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   222
                |> map (nickname_of_thm o snd)
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   223
            in encode_str name ^ ": " ^ encode_strs suggs ^ "\n" end
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   224
        end
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   225
      else
67b04a8375b0 honor fact range for MePo as well
blanchet
parents: 50859
diff changeset
   226
        ""
50519
2951841ec011 parallelized MaSh exporter
blanchet
parents: 50515
diff changeset
   227
    fun accum x (yss as ys :: _) = (x :: ys) :: yss
2951841ec011 parallelized MaSh exporter
blanchet
parents: 50515
diff changeset
   228
    val old_factss = tl (fold accum new_facts [old_facts])
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   229
    val lines = Par_List.map do_fact (tag_list 1 (new_facts ~~ rev old_factss))
50519
2951841ec011 parallelized MaSh exporter
blanchet
parents: 50515
diff changeset
   230
  in File.write_list path lines end
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
   231
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   232
fun generate_mesh_suggestions max_suggs mash_file_name mepo_file_name
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   233
                              mesh_file_name =
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   234
  let
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   235
    val mesh_path = Path.explode mesh_file_name
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   236
    val _ = File.write mesh_path ""
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   237
    fun do_fact (mash_line, mepo_line) =
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   238
      let
50829
01c9a515ccdd fixed escaping for MeSh encoder
blanchet
parents: 50826
diff changeset
   239
        val (name, mash_suggs) =
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   240
          extract_suggestions mash_line
50965
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50954
diff changeset
   241
          ||> weight_mash_facts
50829
01c9a515ccdd fixed escaping for MeSh encoder
blanchet
parents: 50826
diff changeset
   242
        val (name', mepo_suggs) =
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   243
          extract_suggestions mepo_line
50965
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50954
diff changeset
   244
          ||> weight_mepo_facts
50829
01c9a515ccdd fixed escaping for MeSh encoder
blanchet
parents: 50826
diff changeset
   245
        val _ = if name = name' then () else error "Input files out of sync."
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   246
        val mess =
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   247
          [(mepo_weight, (mepo_suggs, [])),
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   248
           (mash_weight, (mash_suggs, []))]
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   249
        val mesh_suggs = mesh_facts (op =) max_suggs mess
50829
01c9a515ccdd fixed escaping for MeSh encoder
blanchet
parents: 50826
diff changeset
   250
        val mesh_line = encode_str name ^ ": " ^ encode_strs mesh_suggs ^ "\n"
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   251
      in File.append mesh_path mesh_line end
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   252
    val mash_lines = Path.explode mash_file_name |> File.read_lines
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   253
    val mepo_lines = Path.explode mepo_file_name |> File.read_lines
50907
a86708897266 graceful failure
blanchet
parents: 50906
diff changeset
   254
  in
a86708897266 graceful failure
blanchet
parents: 50906
diff changeset
   255
    if length mash_lines = length mepo_lines then
a86708897266 graceful failure
blanchet
parents: 50906
diff changeset
   256
      List.app do_fact (mash_lines ~~ mepo_lines)
a86708897266 graceful failure
blanchet
parents: 50906
diff changeset
   257
    else
a86708897266 graceful failure
blanchet
parents: 50906
diff changeset
   258
      warning "Skipped: MaSh file missing or out of sync with MePo file."
a86708897266 graceful failure
blanchet
parents: 50906
diff changeset
   259
  end
50814
4247cbd78aaf export MeSh data as well
blanchet
parents: 50754
diff changeset
   260
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
   261
end;