src/HOL/TPTP/mash_eval.ML
author blanchet
Mon, 02 Jun 2014 11:59:50 +0200
changeset 57150 f591096a9c94
parent 57125 2f620ef839ee
child 57306 ff10067b2248
permissions -rw-r--r--
add option to keep duplicates, for more precise evaluation of relevance filters
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
48285
902ab51dd12a renaming
blanchet
parents: 48284
diff changeset
     1
(*  Title:      HOL/TPTP/mash_eval.ML
48234
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
48285
902ab51dd12a renaming
blanchet
parents: 48284
diff changeset
     5
Evaluate proof suggestions from MaSh (Machine-learning for Sledgehammer).
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     6
*)
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     7
48285
902ab51dd12a renaming
blanchet
parents: 48284
diff changeset
     8
signature MASH_EVAL =
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
     9
sig
55201
1ee776da8da7 renamed ML file
blanchet
parents: 55198
diff changeset
    10
  type params = Sledgehammer_Prover.params
48251
6cdcfbddc077 moved most of MaSh exporter code to Sledgehammer
blanchet
parents: 48250
diff changeset
    11
51015
da8aeffde7a5 compile
blanchet
parents: 51008
diff changeset
    12
  val MePoN : string
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    13
  val MaSh_IsarN : string
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    14
  val MaSh_ProverN : string
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    15
  val MeSh_IsarN : string
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    16
  val MeSh_ProverN : string
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    17
  val IsarN : string
57150
f591096a9c94 add option to keep duplicates, for more precise evaluation of relevance filters
blanchet
parents: 57125
diff changeset
    18
  val evaluate_mash_suggestions : Proof.context -> params -> int * int option -> bool ->
f591096a9c94 add option to keep duplicates, for more precise evaluation of relevance filters
blanchet
parents: 57125
diff changeset
    19
    string list -> string option -> string -> string -> string -> string -> string -> string -> unit
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    20
end;
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    21
48285
902ab51dd12a renaming
blanchet
parents: 48284
diff changeset
    22
structure MaSh_Eval : MASH_EVAL =
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
    23
struct
48235
40655464a93b MaSh evaluation driver
blanchet
parents: 48234
diff changeset
    24
50557
31313171deb5 thread no timeout properly
blanchet
parents: 50556
diff changeset
    25
open Sledgehammer_Util
48315
82d6e46c673f fixed order of accessibles + other tweaks to MaSh
blanchet
parents: 48313
diff changeset
    26
open Sledgehammer_Fact
50557
31313171deb5 thread no timeout properly
blanchet
parents: 50556
diff changeset
    27
open Sledgehammer_MePo
48381
1b7d798460bb renamed ML structures
blanchet
parents: 48379
diff changeset
    28
open Sledgehammer_MaSh
55201
1ee776da8da7 renamed ML file
blanchet
parents: 55198
diff changeset
    29
open Sledgehammer_Prover
55212
blanchet
parents: 55205
diff changeset
    30
open Sledgehammer_Prover_ATP
55198
7a538e58b64e tuned ML file name
blanchet
parents: 54816
diff changeset
    31
open Sledgehammer_Commands
48240
6a8d18798161 rationalized output
blanchet
parents: 48239
diff changeset
    32
53980
7e6a82c593f4 made SML/NJ happy (NB: toplevel ML environment is unmanaged);
wenzelm
parents: 53139
diff changeset
    33
val prefix = Library.prefix
7e6a82c593f4 made SML/NJ happy (NB: toplevel ML environment is unmanaged);
wenzelm
parents: 53139
diff changeset
    34
51008
e096c0dc538b more precise output of selected facts
blanchet
parents: 51004
diff changeset
    35
val MaSh_IsarN = MaShN ^ "-Isar"
e096c0dc538b more precise output of selected facts
blanchet
parents: 51004
diff changeset
    36
val MaSh_ProverN = MaShN ^ "-Prover"
e096c0dc538b more precise output of selected facts
blanchet
parents: 51004
diff changeset
    37
val MeSh_IsarN = MeShN ^ "-Isar"
e096c0dc538b more precise output of selected facts
blanchet
parents: 51004
diff changeset
    38
val MeSh_ProverN = MeShN ^ "-Prover"
48379
2b5ad61e2ccc renamed "iter" fact filter to "MePo" (Meng--Paulson)
blanchet
parents: 48378
diff changeset
    39
val IsarN = "Isar"
48241
688f1172d523 nicer output
blanchet
parents: 48240
diff changeset
    40
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
    41
fun in_range (from, to) j =
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
    42
  j >= from andalso (to = NONE orelse j <= the to)
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
    43
51182
962190eab40d provide two modes for MaSh driver: linearized or real visibility
blanchet
parents: 51135
diff changeset
    44
fun evaluate_mash_suggestions ctxt params range linearize methods prob_dir_name
50964
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    45
        mepo_file_name mash_isar_file_name mash_prover_file_name
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    46
        mesh_isar_file_name mesh_prover_file_name report_file_name =
48235
40655464a93b MaSh evaluation driver
blanchet
parents: 48234
diff changeset
    47
  let
55205
8450622db0c5 refactor large ML file
blanchet
parents: 55201
diff changeset
    48
    val thy = Proof_Context.theory_of ctxt
51028
7327a847f0c7 more robustness w.r.t. 0
blanchet
parents: 51015
diff changeset
    49
    val zeros = [0, 0, 0, 0, 0, 0]
50448
0a740d127187 have MaSh evaluator keep all raw problem/solution files in a directory
blanchet
parents: 50442
diff changeset
    50
    val report_path = report_file_name |> Path.explode
0a740d127187 have MaSh evaluator keep all raw problem/solution files in a directory
blanchet
parents: 50442
diff changeset
    51
    val _ = File.write report_path ""
50589
42f3630a6a0f no need for tracing
blanchet
parents: 50587
diff changeset
    52
    fun print s = File.append report_path (s ^ "\n")
55205
8450622db0c5 refactor large ML file
blanchet
parents: 55201
diff changeset
    53
    val {provers, max_facts, slice, type_enc, lam_trans, timeout, ...} = default_params thy []
48318
325c8fd0d762 more consolidation of MaSh code
blanchet
parents: 48316
diff changeset
    54
    val prover = hd provers
57108
dc0b4f50e288 more generous max number of suggestions, for more safety
blanchet
parents: 55212
diff changeset
    55
    val max_suggs = generous_max_suggestions (the max_facts)
50964
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    56
    val lines_of = Path.explode #> try File.read_lines #> these
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    57
    val file_names =
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    58
      [mepo_file_name, mash_isar_file_name, mash_prover_file_name,
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    59
       mesh_isar_file_name, mesh_prover_file_name]
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    60
    val lines as [mepo_lines, mash_isar_lines, mash_prover_lines,
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    61
                  mesh_isar_lines, mesh_prover_lines] =
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    62
      map lines_of file_names
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    63
    val num_lines = fold (Integer.max o length) lines 0
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    64
    fun pad lines = lines @ replicate (num_lines - length lines) ""
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    65
    val lines =
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    66
      pad mepo_lines ~~ pad mash_isar_lines ~~ pad mash_prover_lines ~~
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    67
      pad mesh_isar_lines ~~ pad mesh_prover_lines
50557
31313171deb5 thread no timeout properly
blanchet
parents: 50556
diff changeset
    68
    val css = clasimpset_rule_table_of ctxt
50442
4f6a4d32522c don't blacklist "case" theorems -- this causes problems in MaSh later
blanchet
parents: 50440
diff changeset
    69
    val facts = all_facts ctxt true false Symtab.empty [] [] css
50735
6b232d76cbc9 refined class handling, to prevent cycles in fact graph
blanchet
parents: 50633
diff changeset
    70
    val name_tabs = build_name_tables nickname_of_thm facts
48289
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    71
    fun with_index facts s = (find_index (curry (op =) s) facts + 1, s)
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
    72
    fun index_str (j, s) = s ^ "@" ^ string_of_int j
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    73
    val str_of_method = enclose "  " ": "
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    74
    fun str_of_result method facts ({outcome, run_time, used_facts, ...}
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
    75
                                     : prover_result) =
51004
5f2788c38127 distinguish raw and non-raw facts, using raw for 10 000s of facts and non-raw after selection of some hundreds
blanchet
parents: 50967
diff changeset
    76
      let val facts = facts |> map (fst o fst) in
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
    77
        str_of_method method ^
48289
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    78
        (if is_none outcome then
52031
9a9238342963 tuning -- renamed '_from_' to '_of_' in Sledgehammer
blanchet
parents: 51199
diff changeset
    79
           "Success (" ^ ATP_Util.string_of_time run_time ^ "): " ^
48289
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    80
           (used_facts |> map (with_index facts o fst)
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    81
                       |> sort (int_ord o pairself fst)
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
    82
                       |> map index_str
48289
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    83
                       |> space_implode " ") ^
48293
914ca0827804 renamed Sledgehammer options
blanchet
parents: 48292
diff changeset
    84
           (if length facts < the max_facts then
48289
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    85
              " (of " ^ string_of_int (length facts) ^ ")"
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    86
            else
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    87
              "")
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    88
         else
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    89
           "Failure: " ^
48293
914ca0827804 renamed Sledgehammer options
blanchet
parents: 48292
diff changeset
    90
           (facts |> take (the max_facts) |> tag_list 1
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
    91
                  |> map index_str
48289
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    92
                  |> space_implode " "))
6b65f1ad0e4b systematize lazy names in relevance filter
blanchet
parents: 48285
diff changeset
    93
      end
50964
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    94
    fun solve_goal (j, ((((mepo_line, mash_isar_line), mash_prover_line),
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
    95
                         mesh_isar_line), mesh_prover_line)) =
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
    96
      if in_range range j then
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
    97
        let
57125
2f620ef839ee added another way of invoking Python code, for experiments
blanchet
parents: 57108
diff changeset
    98
          val get_suggs = extract_suggestions ##> (take max_suggs #> map fst)
50965
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50964
diff changeset
    99
          val (name1, mepo_suggs) = get_suggs mepo_line
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50964
diff changeset
   100
          val (name2, mash_isar_suggs) = get_suggs mash_isar_line
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50964
diff changeset
   101
          val (name3, mash_prover_suggs) = get_suggs mash_prover_line
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50964
diff changeset
   102
          val (name4, mesh_isar_suggs) = get_suggs mesh_isar_line
7a7d1418301e use correct weights in MeSh driver
blanchet
parents: 50964
diff changeset
   103
          val (name5, mesh_prover_suggs) = get_suggs mesh_prover_line
50964
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
   104
          val [name] =
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
   105
            [name1, name2, name3, name4, name5]
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
   106
            |> filter (curry (op <>) "") |> distinct (op =)
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
   107
            handle General.Match => error "Input files out of sync."
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   108
          val th =
50624
4d0997abce79 improved thm order hack, in case the default names are overridden
blanchet
parents: 50620
diff changeset
   109
            case find_first (fn (_, th) => nickname_of_thm th = name) facts of
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   110
              SOME (_, th) => th
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   111
            | NONE => error ("No fact called \"" ^ name ^ "\".")
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   112
          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
   113
          val (_, hyp_ts, concl_t) = ATP_Util.strip_subgoal goal 1 ctxt
50754
74a6adcb96ac also generate queries for goals with too many Isar dependencies
blanchet
parents: 50735
diff changeset
   114
          val isar_deps = isar_dependencies_of name_tabs th
51135
e32114b25551 tuned code
blanchet
parents: 51134
diff changeset
   115
          val facts =
51199
e3447c380fe1 honor linearization option also in the evaluation driver
blanchet
parents: 51182
diff changeset
   116
            facts
e3447c380fe1 honor linearization option also in the evaluation driver
blanchet
parents: 51182
diff changeset
   117
            |> filter (fn (_, th') =>
e3447c380fe1 honor linearization option also in the evaluation driver
blanchet
parents: 51182
diff changeset
   118
                          if linearize then crude_thm_ord (th', th) = LESS
e3447c380fe1 honor linearization option also in the evaluation driver
blanchet
parents: 51182
diff changeset
   119
                          else thm_less (th', th))
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   120
          (* adapted from "mirabelle_sledgehammer.ML" *)
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   121
          fun set_file_name method (SOME dir) =
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   122
              let
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   123
                val prob_prefix =
50826
18ace05656cf start using MaSh hints
blanchet
parents: 50825
diff changeset
   124
                  "goal_" ^ string_of_int j ^ "__" ^ encode_str name ^ "__" ^
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   125
                  method
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   126
              in
55212
blanchet
parents: 55205
diff changeset
   127
                Config.put atp_dest_dir dir
blanchet
parents: 55205
diff changeset
   128
                #> Config.put atp_problem_prefix (prob_prefix ^ "__")
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   129
                #> Config.put SMT_Config.debug_files (dir ^ "/" ^ prob_prefix)
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   130
              end
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   131
            | set_file_name _ NONE = I
53139
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   132
          fun prove method suggs =
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   133
            if not (member (op =) methods method) orelse
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   134
               (null facts andalso method <> IsarN) then
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   135
              (str_of_method method ^ "Skipped", 0)
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   136
            else
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   137
              let
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   138
                fun nickify ((_, stature), th) =
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   139
                  ((K (encode_str (nickname_of_thm th)), stature), th)
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   140
                val facts =
53139
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   141
                  suggs
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   142
                  |> find_suggested_facts ctxt facts
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   143
                  |> map (fact_of_raw_fact #> nickify)
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   144
                  |> maybe_instantiate_inducts ctxt hyp_ts concl_t
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   145
                  |> take (the max_facts)
51004
5f2788c38127 distinguish raw and non-raw facts, using raw for 10 000s of facts and non-raw after selection of some hundreds
blanchet
parents: 50967
diff changeset
   146
                  |> map fact_of_raw_fact
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   147
                val ctxt = ctxt |> set_file_name method prob_dir_name
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   148
                val res as {outcome, ...} =
54141
f57f8e7a879f generate a comment storing the goal nickname in "learn_prover"
blanchet
parents: 53980
diff changeset
   149
                  run_prover_for_mash ctxt params prover name facts goal
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   150
                val ok = if is_none outcome then 1 else 0
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   151
              in (str_of_result method facts res, ok) end
50591
c5e0073558f3 avoid references altogether
blanchet
parents: 50589
diff changeset
   152
          val ress =
53139
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   153
            [fn () => prove MePoN mepo_suggs,
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   154
             fn () => prove MaSh_IsarN mash_isar_suggs,
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   155
             fn () => prove MaSh_ProverN mash_prover_suggs,
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   156
             fn () => prove MeSh_IsarN mesh_isar_suggs,
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   157
             fn () => prove MeSh_ProverN mesh_prover_suggs,
07a6e11f1631 cleanup old duplicated functionality
blanchet
parents: 52196
diff changeset
   158
             fn () => prove IsarN isar_deps]
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   159
            |> (* Par_List. *) map (fn f => f ())
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   160
        in
50591
c5e0073558f3 avoid references altogether
blanchet
parents: 50589
diff changeset
   161
          "Goal " ^ string_of_int j ^ ": " ^ name :: map fst ress
50587
bd6582be1562 synchronize access to shared reference and print proper total
blanchet
parents: 50563
diff changeset
   162
          |> cat_lines |> print;
50591
c5e0073558f3 avoid references altogether
blanchet
parents: 50589
diff changeset
   163
          map snd ress
50559
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   164
        end
89c0d2f13cca MaSh exporter can now export subsets of the facts, as consecutive ranges
blanchet
parents: 50557
diff changeset
   165
      else
51028
7327a847f0c7 more robustness w.r.t. 0
blanchet
parents: 51015
diff changeset
   166
        zeros
50953
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   167
    fun total_of method ok n =
c4c746bbf836 provide a means to skip a method
blanchet
parents: 50952
diff changeset
   168
      str_of_method method ^ string_of_int ok ^ " (" ^
50591
c5e0073558f3 avoid references altogether
blanchet
parents: 50589
diff changeset
   169
      Real.fmt (StringCvt.FIX (SOME 1))
51028
7327a847f0c7 more robustness w.r.t. 0
blanchet
parents: 51015
diff changeset
   170
               (100.0 * Real.fromInt ok / Real.fromInt (Int.max (1, n))) ^ "%)"
50557
31313171deb5 thread no timeout properly
blanchet
parents: 50556
diff changeset
   171
    val inst_inducts = Config.get ctxt instantiate_inducts
48245
854a47677335 generate ATP dependencies
blanchet
parents: 48241
diff changeset
   172
    val options =
50767
26ad3da13f47 tuned output
blanchet
parents: 50754
diff changeset
   173
      ["prover = " ^ prover,
26ad3da13f47 tuned output
blanchet
parents: 50754
diff changeset
   174
       "max_facts = " ^ string_of_int (the max_facts),
26ad3da13f47 tuned output
blanchet
parents: 50754
diff changeset
   175
       "slice" |> not slice ? prefix "dont_",
26ad3da13f47 tuned output
blanchet
parents: 50754
diff changeset
   176
       "type_enc = " ^ the_default "smart" type_enc,
26ad3da13f47 tuned output
blanchet
parents: 50754
diff changeset
   177
       "lam_trans = " ^ the_default "smart" lam_trans,
54816
10d48c2a3e32 made timeouts in Sledgehammer not be 'option's -- simplified lots of code
blanchet
parents: 54141
diff changeset
   178
       "timeout = " ^ ATP_Util.string_of_time timeout,
48241
688f1172d523 nicer output
blanchet
parents: 48240
diff changeset
   179
       "instantiate_inducts" |> not inst_inducts ? prefix "dont_"]
50587
bd6582be1562 synchronize access to shared reference and print proper total
blanchet
parents: 50563
diff changeset
   180
    val _ = print " * * *";
bd6582be1562 synchronize access to shared reference and print proper total
blanchet
parents: 50563
diff changeset
   181
    val _ = print ("Options: " ^ commas options);
50964
2a990baa09af use precomputed MaSh/MePo data whenever available
blanchet
parents: 50953
diff changeset
   182
    val oks = Par_List.map solve_goal (tag_list 1 lines)
50620
07e08250a880 fixed total
blanchet
parents: 50591
diff changeset
   183
    val n = length oks
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   184
    val [mepo_ok, mash_isar_ok, mash_prover_ok, mesh_isar_ok, mesh_prover_ok,
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   185
         isar_ok] =
51028
7327a847f0c7 more robustness w.r.t. 0
blanchet
parents: 51015
diff changeset
   186
      if n = 0 then zeros else map Integer.sum (map_transpose I oks)
48241
688f1172d523 nicer output
blanchet
parents: 48240
diff changeset
   187
  in
688f1172d523 nicer output
blanchet
parents: 48240
diff changeset
   188
    ["Successes (of " ^ string_of_int n ^ " goals)",
48379
2b5ad61e2ccc renamed "iter" fact filter to "MePo" (Meng--Paulson)
blanchet
parents: 48378
diff changeset
   189
     total_of MePoN mepo_ok n,
50952
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   190
     total_of MaSh_IsarN mash_isar_ok n,
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   191
     total_of MaSh_ProverN mash_prover_ok n,
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   192
     total_of MeSh_IsarN mesh_isar_ok n,
3fd83a0cc4bd evaluate more cases (cf. paper)
blanchet
parents: 50826
diff changeset
   193
     total_of MeSh_ProverN mesh_prover_ok n,
48379
2b5ad61e2ccc renamed "iter" fact filter to "MePo" (Meng--Paulson)
blanchet
parents: 48378
diff changeset
   194
     total_of IsarN isar_ok n]
50437
9762fe72936e store evaluation output in a file
blanchet
parents: 50436
diff changeset
   195
    |> cat_lines |> print
48241
688f1172d523 nicer output
blanchet
parents: 48240
diff changeset
   196
  end
48235
40655464a93b MaSh evaluation driver
blanchet
parents: 48234
diff changeset
   197
48234
06216c789ac9 moved MaSh into own files
blanchet
parents:
diff changeset
   198
end;