src/Pure/Isar/outer_syntax.ML
author wenzelm
Mon, 20 Aug 2012 14:09:09 +0200
changeset 49879 3ee314ae1e0a
parent 49786 2ea997196d04
child 49933 6e5fd4585512
permissions -rw-r--r--
added keyword kind "thy_load" (with optional list of file extensions);
wenzelm@5829
     1
(*  Title:      Pure/Isar/outer_syntax.ML
wenzelm@5829
     2
    Author:     Markus Wenzel, TU Muenchen
wenzelm@5829
     3
wenzelm@27353
     4
The global Isabelle/Isar outer syntax.
wenzelm@27353
     5
wenzelm@27353
     6
Note: the syntax for files is statically determined at the very
wenzelm@27353
     7
beginning; for interactive processing it may change dynamically.
wenzelm@5829
     8
*)
wenzelm@5829
     9
wenzelm@5829
    10
signature OUTER_SYNTAX =
wenzelm@5829
    11
sig
wenzelm@44582
    12
  type outer_syntax
wenzelm@44583
    13
  val is_markup: outer_syntax -> Thy_Output.markup -> string -> bool
wenzelm@44582
    14
  val get_syntax: unit -> (Scan.lexicon * Scan.lexicon) * outer_syntax
wenzelm@47841
    15
  val check_syntax: unit -> unit
wenzelm@49661
    16
  type command_spec = (string * Keyword.T) * Position.T
wenzelm@47836
    17
  val command: command_spec -> string ->
wenzelm@29311
    18
    (Toplevel.transition -> Toplevel.transition) parser -> unit
wenzelm@47836
    19
  val markup_command: Thy_Output.markup -> command_spec -> string ->
wenzelm@29311
    20
    (Toplevel.transition -> Toplevel.transition) parser -> unit
wenzelm@47836
    21
  val improper_command: command_spec -> string ->
wenzelm@29311
    22
    (Toplevel.transition -> Toplevel.transition) parser -> unit
wenzelm@47836
    23
  val local_theory': command_spec -> string ->
wenzelm@29380
    24
    (bool -> local_theory -> local_theory) parser -> unit
wenzelm@47836
    25
  val local_theory: command_spec -> string ->
wenzelm@29311
    26
    (local_theory -> local_theory) parser -> unit
wenzelm@47836
    27
  val local_theory_to_proof': command_spec -> string ->
wenzelm@29311
    28
    (bool -> local_theory -> Proof.state) parser -> unit
wenzelm@47836
    29
  val local_theory_to_proof: command_spec -> string ->
wenzelm@29311
    30
    (local_theory -> Proof.state) parser -> unit
wenzelm@5883
    31
  val print_outer_syntax: unit -> unit
wenzelm@36969
    32
  val scan: Position.T -> string -> Token.T list
wenzelm@25580
    33
  val parse: Position.T -> string -> Toplevel.transition list
wenzelm@26600
    34
  type isar
wenzelm@38551
    35
  val isar: TextIO.instream -> bool -> isar
wenzelm@47747
    36
  val read_span: outer_syntax -> Token.T list -> Toplevel.transition * bool
wenzelm@45399
    37
  val read_element: outer_syntax -> (unit -> theory) -> Thy_Syntax.element ->
wenzelm@44583
    38
    (Toplevel.transition * Toplevel.transition list) list
wenzelm@5829
    39
end;
wenzelm@5829
    40
wenzelm@36953
    41
structure Outer_Syntax: OUTER_SYNTAX =
wenzelm@5829
    42
struct
wenzelm@5829
    43
wenzelm@5829
    44
(** outer syntax **)
wenzelm@5829
    45
wenzelm@29311
    46
(* command parsers *)
wenzelm@5829
    47
wenzelm@29311
    48
datatype command = Command of
wenzelm@24868
    49
 {comment: string,
wenzelm@37216
    50
  markup: Thy_Output.markup option,
wenzelm@24868
    51
  int_only: bool,
wenzelm@49662
    52
  parse: (Toplevel.transition -> Toplevel.transition) parser,
wenzelm@49662
    53
  pos: Position.T,
wenzelm@49662
    54
  id: serial};
wenzelm@5829
    55
wenzelm@49662
    56
fun new_command comment markup int_only parse pos =
wenzelm@49662
    57
  Command {comment = comment, markup = markup, int_only = int_only, parse = parse,
wenzelm@49662
    58
    pos = pos, id = serial ()};
wenzelm@49662
    59
wenzelm@49662
    60
fun command_markup def (name, Command {pos, id, ...}) =
wenzelm@49662
    61
  Markup.properties (Position.entity_properties_of def id pos)
wenzelm@49662
    62
    (Isabelle_Markup.entity Isabelle_Markup.commandN name);
wenzelm@5829
    63
wenzelm@5829
    64
wenzelm@5829
    65
(* parse command *)
wenzelm@5829
    66
wenzelm@6860
    67
local
wenzelm@6199
    68
wenzelm@14925
    69
fun terminate false = Scan.succeed ()
wenzelm@45231
    70
  | terminate true =
wenzelm@45231
    71
      Parse.group (fn () => "end of input")
wenzelm@45231
    72
        (Scan.option Parse.sync -- Parse.semicolon >> K ());
wenzelm@14925
    73
wenzelm@26620
    74
fun body cmd (name, _) =
wenzelm@7026
    75
  (case cmd name of
wenzelm@29311
    76
    SOME (Command {int_only, parse, ...}) =>
wenzelm@36950
    77
      Parse.!!! (Scan.prompt (name ^ "# ") (Parse.tags |-- parse >> pair int_only))
wenzelm@49206
    78
  | NONE =>
wenzelm@49206
    79
      Scan.succeed (false, Toplevel.imperative (fn () =>
wenzelm@49206
    80
        error ("Bad parser for outer syntax command " ^ quote name))));
wenzelm@5829
    81
wenzelm@6860
    82
in
wenzelm@6860
    83
wenzelm@26620
    84
fun parse_command do_terminate cmd =
wenzelm@36950
    85
  Parse.semicolon >> K NONE ||
wenzelm@36950
    86
  Parse.sync >> K NONE ||
wenzelm@36950
    87
  (Parse.position Parse.command :-- body cmd) --| terminate do_terminate
wenzelm@6860
    88
    >> (fn ((name, pos), (int_only, f)) =>
skalberg@15531
    89
      SOME (Toplevel.empty |> Toplevel.name name |> Toplevel.position pos |>
wenzelm@6860
    90
        Toplevel.interactive int_only |> f));
wenzelm@5829
    91
wenzelm@6199
    92
end;
wenzelm@6199
    93
wenzelm@5829
    94
wenzelm@44582
    95
(* type outer_syntax *)
wenzelm@44582
    96
wenzelm@44582
    97
datatype outer_syntax = Outer_Syntax of
wenzelm@44582
    98
 {commands: command Symtab.table,
wenzelm@44582
    99
  markups: (string * Thy_Output.markup) list};
wenzelm@44582
   100
wenzelm@44582
   101
fun make_outer_syntax commands markups =
wenzelm@44582
   102
  Outer_Syntax {commands = commands, markups = markups};
wenzelm@44582
   103
wenzelm@44582
   104
val empty_outer_syntax = make_outer_syntax Symtab.empty [];
wenzelm@44582
   105
wenzelm@44582
   106
wenzelm@44582
   107
fun map_commands f (Outer_Syntax {commands, ...}) =
wenzelm@44582
   108
  let
wenzelm@44582
   109
    val commands' = f commands;
wenzelm@44582
   110
    val markups' =
wenzelm@44582
   111
      Symtab.fold (fn (name, Command {markup = SOME m, ...}) => cons (name, m) | _ => I)
wenzelm@44582
   112
        commands' [];
wenzelm@44582
   113
  in make_outer_syntax commands' markups' end;
wenzelm@44582
   114
wenzelm@44582
   115
fun dest_commands (Outer_Syntax {commands, ...}) =
wenzelm@44582
   116
  commands |> Symtab.dest |> sort_wrt #1
wenzelm@44582
   117
  |> map (fn (name, Command {comment, int_only, ...}) => (name, comment, int_only));
wenzelm@44582
   118
wenzelm@44582
   119
fun lookup_commands (Outer_Syntax {commands, ...}) = Symtab.lookup commands;
wenzelm@44582
   120
wenzelm@44582
   121
fun is_markup (Outer_Syntax {markups, ...}) kind name =
wenzelm@44582
   122
  AList.lookup (op =) markups name = SOME kind;
wenzelm@44582
   123
wenzelm@44582
   124
wenzelm@5829
   125
wenzelm@9132
   126
(** global outer syntax **)
wenzelm@5829
   127
wenzelm@49661
   128
type command_spec = (string * Keyword.T) * Position.T;
wenzelm@47836
   129
wenzelm@7026
   130
local
wenzelm@7026
   131
wenzelm@44582
   132
(*synchronized wrt. Keywords*)
wenzelm@44582
   133
val global_outer_syntax = Unsynchronized.ref empty_outer_syntax;
wenzelm@5952
   134
wenzelm@49662
   135
fun add_command (name, kind) cmd = CRITICAL (fn () =>
wenzelm@47824
   136
  let
wenzelm@47824
   137
    val thy = ML_Context.the_global_context ();
wenzelm@49662
   138
    val Command {pos, ...} = cmd;
wenzelm@47824
   139
    val _ =
wenzelm@47824
   140
      (case try (Thy_Header.the_keyword thy) name of
wenzelm@47836
   141
        SOME spec =>
wenzelm@49879
   142
          if Option.map #1 spec = SOME (Keyword.kind_files_of kind) then ()
wenzelm@49661
   143
          else error ("Inconsistent outer syntax keyword declaration " ^
wenzelm@49661
   144
            quote name ^ Position.str_of pos)
wenzelm@47824
   145
      | NONE =>
wenzelm@47836
   146
          if Context.theory_name thy = Context.PureN
wenzelm@47836
   147
          then Keyword.define (name, SOME kind)
wenzelm@49661
   148
          else error ("Undeclared outer syntax command " ^ quote name ^ Position.str_of pos));
wenzelm@49662
   149
    val _ = Position.report pos (command_markup true (name, cmd));
wenzelm@47824
   150
  in
wenzelm@47824
   151
    Unsynchronized.change global_outer_syntax (map_commands (fn commands =>
wenzelm@47824
   152
     (if not (Symtab.defined commands name) then ()
wenzelm@47824
   153
      else warning ("Redefining outer syntax command " ^ quote name);
wenzelm@47824
   154
      Symtab.update (name, cmd) commands)))
wenzelm@47824
   155
  end);
wenzelm@5829
   156
wenzelm@7026
   157
in
wenzelm@6722
   158
wenzelm@44582
   159
fun get_syntax () = CRITICAL (fn () => (Keyword.get_lexicons (), ! global_outer_syntax));
wenzelm@7026
   160
wenzelm@47841
   161
fun check_syntax () =
wenzelm@47841
   162
  let
wenzelm@47841
   163
    val ((_, major), syntax) = CRITICAL (fn () => (Keyword.dest (), ! global_outer_syntax));
wenzelm@47841
   164
  in
wenzelm@47841
   165
    (case subtract (op =) (map #1 (dest_commands syntax)) major of
wenzelm@47841
   166
      [] => ()
wenzelm@47841
   167
    | missing => error ("Missing outer syntax command(s) " ^ commas_quote missing))
wenzelm@47841
   168
  end;
wenzelm@47841
   169
wenzelm@44582
   170
fun lookup_commands_dynamic () = lookup_commands (! global_outer_syntax);
wenzelm@24868
   171
wenzelm@49662
   172
fun command (spec, pos) comment parse =
wenzelm@49662
   173
  add_command spec (new_command comment NONE false parse pos);
wenzelm@24868
   174
wenzelm@49662
   175
fun markup_command markup (spec, pos) comment parse =
wenzelm@49662
   176
  add_command spec (new_command comment (SOME markup) false parse pos);
wenzelm@24868
   177
wenzelm@49662
   178
fun improper_command (spec, pos) comment parse =
wenzelm@49662
   179
  add_command spec (new_command comment NONE true parse pos);
wenzelm@29311
   180
wenzelm@44582
   181
end;
wenzelm@44582
   182
wenzelm@5829
   183
wenzelm@26990
   184
(* local_theory commands *)
wenzelm@26990
   185
wenzelm@47836
   186
fun local_theory_command do_print trans command_spec comment parse =
wenzelm@47836
   187
  command command_spec comment (Parse.opt_target -- parse
wenzelm@26990
   188
    >> (fn (loc, f) => (if do_print then Toplevel.print else I) o trans loc f));
wenzelm@26990
   189
wenzelm@29380
   190
val local_theory' = local_theory_command false Toplevel.local_theory';
wenzelm@29311
   191
val local_theory = local_theory_command false Toplevel.local_theory;
wenzelm@26990
   192
val local_theory_to_proof' = local_theory_command true Toplevel.local_theory_to_proof';
wenzelm@29311
   193
val local_theory_to_proof = local_theory_command true Toplevel.local_theory_to_proof;
wenzelm@26990
   194
wenzelm@26990
   195
wenzelm@24872
   196
(* inspect syntax *)
wenzelm@5829
   197
wenzelm@9223
   198
fun print_outer_syntax () =
wenzelm@7026
   199
  let
wenzelm@47828
   200
    val ((keywords, _), outer_syntax) =
wenzelm@47828
   201
      CRITICAL (fn () => (Keyword.dest (), #2 (get_syntax ())));
wenzelm@27353
   202
    fun pretty_cmd (name, comment, _) =
wenzelm@7026
   203
      Pretty.block [Pretty.str (name ^ ":"), Pretty.brk 2, Pretty.str comment];
wenzelm@44582
   204
    val (int_cmds, cmds) = List.partition #3 (dest_commands outer_syntax);
wenzelm@7026
   205
  in
wenzelm@44582
   206
    [Pretty.strs ("syntax keywords:" :: map quote keywords),
wenzelm@18326
   207
      Pretty.big_list "commands:" (map pretty_cmd cmds),
wenzelm@18326
   208
      Pretty.big_list "interactive-only commands:" (map pretty_cmd int_cmds)]
wenzelm@9223
   209
    |> Pretty.chunks |> Pretty.writeln
wenzelm@7026
   210
  end;
wenzelm@5829
   211
wenzelm@5829
   212
wenzelm@5829
   213
wenzelm@9132
   214
(** toplevel parsing **)
wenzelm@5829
   215
wenzelm@9132
   216
(* basic sources *)
wenzelm@6860
   217
wenzelm@26620
   218
fun toplevel_source term do_recover cmd src =
wenzelm@9132
   219
  let
wenzelm@9132
   220
    val no_terminator =
wenzelm@36969
   221
      Scan.unless Parse.semicolon (Scan.one (Token.not_sync andf Token.not_eof));
wenzelm@23682
   222
    fun recover int =
wenzelm@23682
   223
      (int, fn _ => Scan.prompt "recover# " (Scan.repeat no_terminator) >> K [NONE]);
wenzelm@9132
   224
  in
wenzelm@9132
   225
    src
wenzelm@36969
   226
    |> Token.source_proper
wenzelm@36969
   227
    |> Source.source Token.stopper
wenzelm@36950
   228
      (Scan.bulk (Parse.$$$ "--" -- Parse.!!! Parse.doc_source >> K NONE || Parse.not_eof >> SOME))
wenzelm@23682
   229
        (Option.map recover do_recover)
wenzelm@19482
   230
    |> Source.map_filter I
wenzelm@36969
   231
    |> Source.source Token.stopper
wenzelm@36950
   232
        (Scan.bulk (fn xs => Parse.!!! (parse_command term (cmd ())) xs))
wenzelm@23682
   233
        (Option.map recover do_recover)
wenzelm@19482
   234
    |> Source.map_filter I
wenzelm@9132
   235
  end;
wenzelm@7683
   236
wenzelm@5829
   237
wenzelm@25580
   238
(* off-line scanning/parsing *)
nipkow@14091
   239
wenzelm@27839
   240
fun scan pos str =
wenzelm@16195
   241
  Source.of_string str
wenzelm@40769
   242
  |> Symbol.source
wenzelm@36969
   243
  |> Token.source {do_recover = SOME false} Keyword.get_lexicons pos
wenzelm@16195
   244
  |> Source.exhaust;
wenzelm@16195
   245
wenzelm@25580
   246
fun parse pos str =
wenzelm@25580
   247
  Source.of_string str
wenzelm@40769
   248
  |> Symbol.source
wenzelm@36969
   249
  |> Token.source {do_recover = SOME false} Keyword.get_lexicons pos
wenzelm@44582
   250
  |> toplevel_source false NONE lookup_commands_dynamic
wenzelm@25580
   251
  |> Source.exhaust;
wenzelm@25580
   252
nipkow@14091
   253
wenzelm@24868
   254
(* interactive source of toplevel transformers *)
wenzelm@24868
   255
wenzelm@26600
   256
type isar =
wenzelm@26600
   257
  (Toplevel.transition, (Toplevel.transition option,
wenzelm@36969
   258
    (Token.T, (Token.T option, (Token.T, (Token.T,
wenzelm@30576
   259
      (Symbol_Pos.T, Position.T * (Symbol.symbol, (string, unit) Source.source)
wenzelm@27770
   260
  Source.source) Source.source) Source.source) Source.source)
wenzelm@27770
   261
  Source.source) Source.source) Source.source) Source.source;
wenzelm@26600
   262
wenzelm@38551
   263
fun isar in_stream term : isar =
wenzelm@38551
   264
  Source.tty in_stream
wenzelm@40769
   265
  |> Symbol.source
wenzelm@36969
   266
  |> Token.source {do_recover = SOME true} Keyword.get_lexicons Position.none
wenzelm@44582
   267
  |> toplevel_source term (SOME true) lookup_commands_dynamic;
wenzelm@24868
   268
wenzelm@24868
   269
wenzelm@45399
   270
(* read toplevel commands -- fail-safe *)
wenzelm@27839
   271
wenzelm@45531
   272
fun read_span outer_syntax toks =
wenzelm@27839
   273
  let
wenzelm@44582
   274
    val commands = lookup_commands outer_syntax;
wenzelm@49764
   275
wenzelm@49786
   276
    val proper_range = Position.set_range (Command.proper_range toks);
wenzelm@49764
   277
    val pos =
wenzelm@49764
   278
      (case find_first Token.is_command toks of
wenzelm@49764
   279
        SOME tok => Token.position_of tok
wenzelm@49786
   280
      | NONE => proper_range);
wenzelm@49662
   281
wenzelm@49662
   282
    fun command_reports tok =
wenzelm@49764
   283
      if Token.is_command tok then
wenzelm@49662
   284
        let val name = Token.content_of tok in
wenzelm@49662
   285
          (case commands name of
wenzelm@49662
   286
            NONE => []
wenzelm@49783
   287
          | SOME cmd => [((Token.position_of tok, command_markup false (name, cmd)), "")])
wenzelm@49662
   288
        end
wenzelm@49662
   289
      else [];
wenzelm@49764
   290
wenzelm@49764
   291
    val (is_malformed, token_reports) = Thy_Syntax.reports_of_tokens toks;
wenzelm@49783
   292
    val _ = Position.reports_text (token_reports @ maps command_reports toks);
wenzelm@27839
   293
  in
wenzelm@49764
   294
    if is_malformed then (Toplevel.malformed pos "Malformed command syntax", true)
wenzelm@49764
   295
    else
wenzelm@49764
   296
      (case Source.exhaust (toplevel_source false NONE (K commands) (Source.of_list toks)) of
wenzelm@49764
   297
        [tr] =>
wenzelm@49764
   298
          if Keyword.is_control (Toplevel.name_of tr) then
wenzelm@49764
   299
            (Toplevel.malformed pos "Illegal control command", true)
wenzelm@49764
   300
          else (tr, true)
wenzelm@49786
   301
      | [] => (Toplevel.ignored (Position.set_range (Command.range toks)), false)
wenzelm@49786
   302
      | _ => (Toplevel.malformed proper_range "Exactly one command expected", true))
wenzelm@49786
   303
      handle ERROR msg => (Toplevel.malformed proper_range msg, true)
wenzelm@27839
   304
  end;
wenzelm@27839
   305
wenzelm@45399
   306
fun read_element outer_syntax init {head, proof, proper_proof} =
wenzelm@28436
   307
  let
wenzelm@45531
   308
    val read = read_span outer_syntax o Thy_Syntax.span_content;
wenzelm@45531
   309
    val (tr, proper_head) = read head |>> Toplevel.modify_init init;
wenzelm@45531
   310
    val proof_trs = map read proof |> filter #2 |> map #1;
wenzelm@28436
   311
  in
wenzelm@48288
   312
    if proper_head andalso proper_proof andalso
wenzelm@48288
   313
      not (Keyword.is_schematic_goal (Toplevel.name_of tr)) then [(tr, proof_trs)]
wenzelm@44502
   314
    else map (rpair []) (if proper_head then tr :: proof_trs else proof_trs)
wenzelm@28436
   315
  end;
wenzelm@28432
   316
wenzelm@5829
   317
end;
wenzelm@36953
   318