src/Pure/Isar/outer_lex.ML
author wenzelm
Tue Aug 16 13:42:40 2005 +0200 (2005-08-16 ago)
changeset 17069 ee08b2466a09
parent 16029 070ed43b86f8
child 17164 a786e1a1ce02
permissions -rw-r--r--
clarify is_newline vs. is_blank;
removed is_indent;
added is_comment;
wenzelm@5825
     1
(*  Title:      Pure/Isar/outer_lex.ML
wenzelm@5825
     2
    ID:         $Id$
wenzelm@5825
     3
    Author:     Markus Wenzel, TU Muenchen
wenzelm@5825
     4
wenzelm@5825
     5
Outer lexical syntax for Isabelle/Isar.
wenzelm@5825
     6
*)
wenzelm@5825
     7
wenzelm@5825
     8
signature OUTER_LEX =
wenzelm@5825
     9
sig
wenzelm@5825
    10
  datatype token_kind =
wenzelm@7477
    11
    Command | Keyword | Ident | LongIdent | SymIdent | Var | TypeIdent | TypeVar |
wenzelm@10748
    12
    Nat | String | Verbatim | Space | Comment | Sync | Malformed | EOF
aspinall@15143
    13
  eqtype token
wenzelm@5825
    14
  val str_of_kind: token_kind -> string
wenzelm@5825
    15
  val stopper: token * (token -> bool)
wenzelm@6859
    16
  val not_sync: token -> bool
wenzelm@5825
    17
  val not_eof: token -> bool
wenzelm@5825
    18
  val position_of: token -> Position.T
wenzelm@5825
    19
  val pos_of: token -> string
wenzelm@5825
    20
  val is_kind: token_kind -> token -> bool
wenzelm@7026
    21
  val keyword_with: (string -> bool) -> token -> bool
wenzelm@16029
    22
  val ident_with: (string -> bool) -> token -> bool
wenzelm@5825
    23
  val name_of: token -> string
wenzelm@5825
    24
  val is_proper: token -> bool
wenzelm@9130
    25
  val is_semicolon: token -> bool
wenzelm@17069
    26
  val is_comment: token -> bool
wenzelm@8580
    27
  val is_begin_ignore: token -> bool
wenzelm@8580
    28
  val is_end_ignore: token -> bool
wenzelm@17069
    29
  val is_blank: token -> bool
wenzelm@8651
    30
  val is_newline: token -> bool
wenzelm@14991
    31
  val unparse: token -> string
wenzelm@5825
    32
  val val_of: token -> string
wenzelm@5876
    33
  val is_sid: string -> bool
wenzelm@9130
    34
  val !!! : string -> (Position.T * 'a -> 'b) -> Position.T * 'a -> 'b
wenzelm@9130
    35
  val incr_line: ('a -> 'b * 'c) -> Position.T * 'a -> 'b * (Position.T * 'c)
wenzelm@9130
    36
  val keep_line: ('a -> 'b * 'c) -> Position.T * 'a -> 'b * (Position.T * 'c)
wenzelm@9130
    37
  val scan_blank: Position.T * Symbol.symbol list
wenzelm@9130
    38
    -> Symbol.symbol * (Position.T * Symbol.symbol list)
wenzelm@9130
    39
  val scan_string: Position.T * Symbol.symbol list -> string * (Position.T * Symbol.symbol list)
wenzelm@7026
    40
  val scan: (Scan.lexicon * Scan.lexicon) ->
wenzelm@5825
    41
    Position.T * Symbol.symbol list -> token * (Position.T * Symbol.symbol list)
wenzelm@7026
    42
  val source: bool -> (unit -> (Scan.lexicon * Scan.lexicon)) ->
wenzelm@7026
    43
    Position.T -> (Symbol.symbol, 'a) Source.source ->
wenzelm@7682
    44
    (token, Position.T * (Symbol.symbol, 'a) Source.source) Source.source
wenzelm@9130
    45
  val source_proper: (token, 'a) Source.source -> (token, (token, 'a) Source.source) Source.source
wenzelm@9130
    46
  val make_lexicon: string list -> Scan.lexicon
wenzelm@5825
    47
end;
wenzelm@5825
    48
wenzelm@5825
    49
structure OuterLex: OUTER_LEX =
wenzelm@5825
    50
struct
wenzelm@5825
    51
wenzelm@5825
    52
wenzelm@5825
    53
(** tokens **)
wenzelm@5825
    54
wenzelm@5825
    55
(* datatype token *)
wenzelm@5825
    56
wenzelm@5825
    57
datatype token_kind =
wenzelm@7477
    58
  Command | Keyword | Ident | LongIdent | SymIdent | Var | TypeIdent | TypeVar |
wenzelm@10748
    59
  Nat | String | Verbatim | Space | Comment | Sync | Malformed | EOF;
wenzelm@5825
    60
wenzelm@5825
    61
datatype token = Token of Position.T * (token_kind * string);
wenzelm@5825
    62
wenzelm@5825
    63
val str_of_kind =
wenzelm@7026
    64
 fn Command => "command"
wenzelm@7026
    65
  | Keyword => "keyword"
wenzelm@5825
    66
  | Ident => "identifier"
wenzelm@5825
    67
  | LongIdent => "long identifier"
wenzelm@5825
    68
  | SymIdent => "symbolic identifier"
wenzelm@5825
    69
  | Var => "schematic variable"
wenzelm@5825
    70
  | TypeIdent => "type variable"
wenzelm@5825
    71
  | TypeVar => "schematic type variable"
wenzelm@5825
    72
  | Nat => "number"
wenzelm@5825
    73
  | String => "string"
wenzelm@5825
    74
  | Verbatim => "verbatim text"
wenzelm@7682
    75
  | Space => "white space"
wenzelm@7682
    76
  | Comment => "comment text"
wenzelm@6859
    77
  | Sync => "sync marker"
wenzelm@10748
    78
  | Malformed => "bad input"
wenzelm@5825
    79
  | EOF => "end-of-file";
wenzelm@5825
    80
wenzelm@5825
    81
wenzelm@10748
    82
(* control tokens *)
wenzelm@6859
    83
wenzelm@6859
    84
fun not_sync (Token (_, (Sync, _))) = false
wenzelm@6859
    85
  | not_sync _ = true;
wenzelm@6859
    86
wenzelm@10748
    87
val malformed = Token (Position.none, (Malformed, ""));
aspinall@15224
    88
fun malformed_of xs = Token (Position.none, (Malformed, implode xs));
wenzelm@10748
    89
wenzelm@6859
    90
wenzelm@5825
    91
(* eof token *)
wenzelm@5825
    92
wenzelm@5825
    93
val eof = Token (Position.none, (EOF, ""));
wenzelm@5825
    94
wenzelm@5825
    95
fun is_eof (Token (_, (EOF, _))) = true
wenzelm@5825
    96
  | is_eof _ = false;
wenzelm@5825
    97
wenzelm@5825
    98
val stopper = (eof, is_eof);
wenzelm@5825
    99
val not_eof = not o is_eof;
wenzelm@5825
   100
wenzelm@5825
   101
wenzelm@5825
   102
(* get position *)
wenzelm@5825
   103
wenzelm@5825
   104
fun position_of (Token (pos, _)) = pos;
wenzelm@5825
   105
val pos_of = Position.str_of o position_of;
wenzelm@5825
   106
wenzelm@5825
   107
wenzelm@5825
   108
(* kind of token *)
wenzelm@5825
   109
wenzelm@5825
   110
fun is_kind k (Token (_, (k', _))) = k = k';
wenzelm@5825
   111
wenzelm@7026
   112
fun keyword_with pred (Token (_, (Keyword, x))) = pred x
wenzelm@7026
   113
  | keyword_with _ _ = false;
wenzelm@5825
   114
wenzelm@16029
   115
fun ident_with pred (Token (_, (Ident, x))) = pred x
wenzelm@16029
   116
  | ident_with _ _ = false;
wenzelm@16029
   117
wenzelm@7682
   118
fun is_proper (Token (_, (Space, _))) = false
wenzelm@7682
   119
  | is_proper (Token (_, (Comment, _))) = false
wenzelm@5825
   120
  | is_proper _ = true;
wenzelm@5825
   121
wenzelm@9195
   122
fun is_semicolon (Token (_, (Keyword, ";"))) = true
wenzelm@9130
   123
  | is_semicolon _ = false;
wenzelm@9130
   124
wenzelm@17069
   125
fun is_comment (Token (_, (Comment, _))) = true
wenzelm@17069
   126
  | is_comment _ = false;
wenzelm@17069
   127
wenzelm@8580
   128
fun is_begin_ignore (Token (_, (Comment, "<"))) = true
wenzelm@8580
   129
  | is_begin_ignore _ = false;
wenzelm@8580
   130
wenzelm@8580
   131
fun is_end_ignore (Token (_, (Comment, ">"))) = true
wenzelm@8580
   132
  | is_end_ignore _ = false;
wenzelm@8580
   133
wenzelm@8651
   134
wenzelm@17069
   135
(* blanks and newlines -- space tokens obey lines *)
wenzelm@8651
   136
wenzelm@17069
   137
fun is_blank (Token (_, (Space, s))) = not (String.isSuffix "\n" s)
wenzelm@17069
   138
  | is_blank _ = false;
wenzelm@17069
   139
wenzelm@17069
   140
fun is_newline (Token (_, (Space, s))) = String.isSuffix "\n" s
wenzelm@8651
   141
  | is_newline _ = false;
wenzelm@8651
   142
wenzelm@5825
   143
wenzelm@14991
   144
(* token content *)
wenzelm@9155
   145
wenzelm@9155
   146
fun name_of (tok as Token (_, (k, x))) =
wenzelm@10748
   147
  if is_semicolon tok then "terminator"
wenzelm@10748
   148
  else if x = "" then str_of_kind k
wenzelm@10748
   149
  else str_of_kind k ^ " " ^ quote x;
wenzelm@5825
   150
wenzelm@14991
   151
fun unparse (Token (_, (kind, x))) =
wenzelm@14991
   152
  (case kind of
aspinall@15209
   153
    String => x |> quote
wenzelm@14991
   154
  | Verbatim => x |> enclose "{*" "*}"
wenzelm@14991
   155
  | Comment => x |> enclose "(*" "*)"
wenzelm@14991
   156
  | _ => x);
wenzelm@14991
   157
wenzelm@5825
   158
fun val_of (Token (_, (_, x))) = x;
wenzelm@5825
   159
wenzelm@5825
   160
fun token_leq (Token (_, (_, x)), Token (_, (_, x'))) = x <= x';
wenzelm@5825
   161
wenzelm@5825
   162
wenzelm@5825
   163
wenzelm@5825
   164
(** scanners **)
wenzelm@5825
   165
wenzelm@5825
   166
fun change_prompt scan = Scan.prompt "# " scan;
wenzelm@5825
   167
wenzelm@5825
   168
wenzelm@5825
   169
(* diagnostics *)
wenzelm@5825
   170
wenzelm@5825
   171
fun lex_err msg ((pos, cs), _) = "Outer lexical error" ^ Position.str_of pos ^ ": " ^ msg cs;
wenzelm@9130
   172
fun !!! msg scan = Scan.!! (lex_err (K msg)) scan;
wenzelm@5825
   173
wenzelm@5825
   174
wenzelm@5825
   175
(* line numbering *)
wenzelm@5825
   176
wenzelm@5825
   177
fun incr_line scan = Scan.depend (fn pos => scan >> pair (Position.inc pos));
wenzelm@5825
   178
val keep_line = Scan.lift;
wenzelm@5825
   179
wenzelm@5825
   180
val scan_blank =
wenzelm@5825
   181
  incr_line ($$ "\n") ||
wenzelm@5825
   182
  keep_line (Scan.one Symbol.is_blank);
wenzelm@5825
   183
wenzelm@5825
   184
wenzelm@5825
   185
(* scan symbolic idents *)
wenzelm@5825
   186
wenzelm@5825
   187
val sym_chars = explode "!#$%&*+-/:<=>?@^_`|~";
wenzelm@5825
   188
fun is_sym_char s = s mem sym_chars;
wenzelm@5825
   189
wenzelm@8231
   190
val scan_symid =
wenzelm@8231
   191
  Scan.any1 is_sym_char >> implode ||
wenzelm@8231
   192
  Scan.one Symbol.is_symbolic;
wenzelm@5825
   193
wenzelm@8231
   194
fun is_symid str =
wenzelm@8231
   195
  (case try Symbol.explode str of
skalberg@15531
   196
    SOME [s] => Symbol.is_symbolic s orelse is_sym_char s
skalberg@15531
   197
  | SOME ss => forall is_sym_char ss
wenzelm@8231
   198
  | _ => false);
wenzelm@8231
   199
wenzelm@5876
   200
val is_sid = is_symid orf Syntax.is_identifier;
wenzelm@5825
   201
wenzelm@5825
   202
wenzelm@5825
   203
(* scan strings *)
wenzelm@5825
   204
wenzelm@5825
   205
val scan_str =
wenzelm@7682
   206
  scan_blank ||
wenzelm@9130
   207
  keep_line ($$ "\\") |-- !!! "bad escape character in string"
wenzelm@9051
   208
      (scan_blank || keep_line ($$ "\"" || $$ "\\")) ||
wenzelm@6859
   209
  keep_line (Scan.one (not_equal "\\" andf not_equal "\"" andf
wenzelm@6859
   210
    Symbol.not_sync andf Symbol.not_eof));
wenzelm@5825
   211
wenzelm@5825
   212
val scan_string =
wenzelm@5825
   213
  keep_line ($$ "\"") |--
wenzelm@9130
   214
    !!! "missing quote at end of string"
wenzelm@5825
   215
      (change_prompt ((Scan.repeat scan_str >> implode) --| keep_line ($$ "\"")));
wenzelm@5825
   216
wenzelm@5825
   217
wenzelm@5825
   218
(* scan verbatim text *)
wenzelm@5825
   219
wenzelm@5825
   220
val scan_verb =
wenzelm@5825
   221
  scan_blank ||
wenzelm@6743
   222
  keep_line ($$ "*" --| Scan.ahead (Scan.one (not_equal "}"))) ||
wenzelm@6859
   223
  keep_line (Scan.one (not_equal "*" andf Symbol.not_sync andf Symbol.not_eof));
wenzelm@5825
   224
wenzelm@5825
   225
val scan_verbatim =
wenzelm@6743
   226
  keep_line ($$ "{" -- $$ "*") |--
wenzelm@9130
   227
    !!! "missing end of verbatim text"
wenzelm@6743
   228
      (change_prompt ((Scan.repeat scan_verb >> implode) --| keep_line ($$ "*" -- $$ "}")));
wenzelm@5825
   229
wenzelm@5825
   230
wenzelm@5825
   231
(* scan space *)
wenzelm@5825
   232
wenzelm@5825
   233
val is_space = Symbol.is_blank andf not_equal "\n";
wenzelm@5825
   234
wenzelm@5825
   235
val scan_space =
wenzelm@7682
   236
  (keep_line (Scan.any1 is_space) -- Scan.optional (incr_line ($$ "\n")) "" ||
wenzelm@7682
   237
    keep_line (Scan.any is_space) -- incr_line ($$ "\n")) >> (fn (cs, c) => implode cs ^ c);
wenzelm@5825
   238
wenzelm@5825
   239
wenzelm@5825
   240
(* scan nested comments *)
wenzelm@5825
   241
wenzelm@5825
   242
val scan_cmt =
wenzelm@5825
   243
  Scan.lift scan_blank ||
wenzelm@5825
   244
  Scan.depend (fn d => keep_line ($$ "(" ^^ $$ "*") >> pair (d + 1)) ||
wenzelm@5825
   245
  Scan.depend (fn 0 => Scan.fail | d => keep_line ($$ "*" ^^ $$ ")") >> pair (d - 1)) ||
wenzelm@5825
   246
  Scan.lift (keep_line ($$ "*" --| Scan.ahead (Scan.one (not_equal ")")))) ||
wenzelm@6859
   247
  Scan.lift (keep_line (Scan.one (not_equal "*" andf Symbol.not_sync andf Symbol.not_eof)));
wenzelm@5825
   248
wenzelm@5825
   249
val scan_comment =
wenzelm@5825
   250
  keep_line ($$ "(" -- $$ "*") |--
wenzelm@9130
   251
    !!! "missing end of comment"
wenzelm@5825
   252
      (change_prompt
wenzelm@7682
   253
        (Scan.pass 0 (Scan.repeat scan_cmt >> implode) --| keep_line ($$ "*" -- $$ ")")));
wenzelm@5825
   254
wenzelm@5825
   255
wenzelm@5825
   256
(* scan token *)
wenzelm@5825
   257
wenzelm@9130
   258
fun scan (lex1, lex2) =
wenzelm@5825
   259
  let
wenzelm@9130
   260
    val scanner = Scan.state :-- (fn pos =>
wenzelm@9130
   261
      let
wenzelm@9130
   262
        fun token k x = Token (pos, (k, x));
wenzelm@9130
   263
        fun sync _ = token Sync Symbol.sync;
wenzelm@9130
   264
      in
wenzelm@9130
   265
        scan_string >> token String ||
wenzelm@9130
   266
        scan_verbatim >> token Verbatim ||
wenzelm@9130
   267
        scan_space >> token Space ||
wenzelm@9130
   268
        scan_comment >> token Comment ||
wenzelm@9130
   269
        keep_line (Scan.one Symbol.is_sync >> sync) ||
wenzelm@9130
   270
        keep_line (Scan.max token_leq
wenzelm@9130
   271
          (Scan.max token_leq
wenzelm@9130
   272
            (Scan.literal lex1 >> (token Keyword o implode))
wenzelm@9130
   273
            (Scan.literal lex2 >> (token Command o implode)))
wenzelm@9130
   274
          (Syntax.scan_longid >> token LongIdent ||
wenzelm@9130
   275
            Syntax.scan_id >> token Ident ||
wenzelm@9130
   276
            Syntax.scan_var >> token Var ||
wenzelm@9130
   277
            Syntax.scan_tid >> token TypeIdent ||
wenzelm@9130
   278
            Syntax.scan_tvar >> token TypeVar ||
wenzelm@9130
   279
            Syntax.scan_nat >> token Nat ||
wenzelm@9130
   280
            scan_symid >> token SymIdent))
wenzelm@9130
   281
      end) >> #2;
wenzelm@14729
   282
  in !! (lex_err (fn cs => "bad input " ^ quote (Symbol.beginning 10 cs))) scanner end;
wenzelm@5825
   283
wenzelm@5825
   284
wenzelm@9130
   285
(* token sources *)
wenzelm@5825
   286
wenzelm@6859
   287
val is_junk = (not o Symbol.is_blank) andf Symbol.not_sync andf Symbol.not_eof;
aspinall@15224
   288
fun recover xs = (keep_line (Scan.any is_junk) >> (fn ts => [malformed_of ts])) xs;
wenzelm@5825
   289
wenzelm@5825
   290
fun source do_recover get_lex pos src =
wenzelm@5825
   291
  Source.source' pos Symbol.stopper (Scan.bulk (fn xs => scan (get_lex ()) xs))
skalberg@15531
   292
    (if do_recover then SOME recover else NONE) src;
wenzelm@5825
   293
wenzelm@9130
   294
fun source_proper src = src |> Source.filter is_proper;
wenzelm@9130
   295
wenzelm@9130
   296
wenzelm@9130
   297
(* lexicons *)
wenzelm@9130
   298
wenzelm@9130
   299
val make_lexicon = Scan.make_lexicon o map Symbol.explode;
wenzelm@5825
   300
wenzelm@5825
   301
end;