yacc-parser.mly 10.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13
/******************************************************************************/
/*                                                                            */
/*                                   Menhir                                   */
/*                                                                            */
/*                       François Pottier, Inria Paris                        */
/*              Yann Régis-Gianas, PPS, Université Paris Diderot              */
/*                                                                            */
/*  Copyright Inria. All rights reserved. This file is distributed under the  */
/*  terms of the GNU General Public License version 2, as described in the    */
/*  file LICENSE.                                                             */
/*                                                                            */
/******************************************************************************/

14 15
/* This is the crude version of the parser. It is meant to be processed
   by ocamlyacc. Its existence is necessary for bootstrapping. It is kept
16
   in sync with [fancy-parser], with a few differences:
POTTIER Francois's avatar
POTTIER Francois committed
17
   0.  [yacc-parser] produces dummy position information;
18 19 20
   1. [fancy-parser] exploits many features of Menhir;
   2. [fancy-parser] performs slightly more refined error handling;
   3. [fancy-parser] supports anonymous rules. */
21 22 23 24 25 26 27 28

%{

open Syntax
open Positions

%}

POTTIER Francois's avatar
POTTIER Francois committed
29
%token TOKEN TYPE LEFT RIGHT NONASSOC START PREC PUBLIC COLON BAR EOF EQUAL
30
%token INLINE LPAREN RPAREN COMMA QUESTION STAR PLUS PARAMETER ON_ERROR_REDUCE
POTTIER Francois's avatar
POTTIER Francois committed
31
%token <string Positions.located> LID UID
32 33
%token <Stretch.t> HEADER
%token <Stretch.ocamltype> OCAMLTYPE
34
%token <Stretch.t Lazy.t> PERCENTPERCENT
35
%token <Syntax.identifier option array -> Syntax.action> ACTION
36 37
%token <Syntax.attribute> ATTRIBUTE GRAMMARATTRIBUTE
%token PERCENTATTRIBUTE
38
%start grammar
39 40
%type <ParserAux.early_producer> producer
%type <ParserAux.early_production> production
41
%type <Syntax.partial_grammar> grammar
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56

/* These declarations solve a shift-reduce conflict in favor of
   shifting: when the declaration of a non-terminal symbol begins with
   a leading bar, it is understood as an (insignificant) leading
   optional bar, *not* as an empty right-hand side followed by a bar.
   This ambiguity arises due to the existence of a new notation for
   letting several productions share a single semantic action. */

%nonassoc no_optional_bar
%nonassoc BAR

%%

/* ------------------------------------------------------------------------- */
/* A grammar consists of declarations and rules, followed by an optional
57
   postlude, which we do not parse. */
58 59

grammar:
60
  declarations PERCENTPERCENT rules postlude
POTTIER Francois's avatar
POTTIER Francois committed
61 62
    {
      {
63 64 65
        pg_filename          = ""; (* filled in by the caller *)
        pg_declarations      = List.rev $1;
        pg_rules             = $3;
66
        pg_postlude          = $4
67 68 69
      }
    }

70
postlude:
71 72
  EOF
    { None }
73
| PERCENTPERCENT /* followed by actual postlude */
74 75 76
    { Some (Lazy.force $1) }

/* ------------------------------------------------------------------------- */
77
/* A declaration is an %{ OCaml header %}, or a %token, %start,
78 79 80 81 82 83 84 85 86 87 88 89 90
   %type, %left, %right, or %nonassoc declaration. */

declarations:
  /* epsilon */
    { [] }
| declarations declaration
    { $2 @ $1 }

declaration:
| HEADER /* lexically delimited by %{ ... %} */
    { [ unknown_pos (DCode $1) ] }

| TOKEN optional_ocamltype terminals
91
    { List.map (Positions.map (fun (terminal, attrs) -> DToken ($2, terminal, attrs))) $3 }
92 93 94 95

| START nonterminals
    { List.map (Positions.map (fun nonterminal -> DStart nonterminal)) $2 }

96
| TYPE OCAMLTYPE actuals
97 98 99 100 101
    { List.map (Positions.map (fun nt -> DType ($2, nt)))
        (List.map Parameters.with_pos $3) }

| START OCAMLTYPE nonterminals
    /* %start <ocamltype> foo is syntactic sugar for %start foo %type <ocamltype> foo */
POTTIER Francois's avatar
POTTIER Francois committed
102
    { Misc.mapd (fun ntloc ->
103 104 105
        Positions.mapd (fun nt -> DStart nt, DType ($2, ParameterVar ntloc)) ntloc) $3 }

| priority_keyword symbols
106
    { let prec = ParserAux.new_precedence_level (rhs_start_pos 1, rhs_end_pos 1) in
107 108 109 110 111
      List.map (Positions.map (fun symbol -> DTokenProperties (symbol, $1, prec))) $2 }

| PARAMETER OCAMLTYPE
    { [ unknown_pos (DParameter $2) ] }

112 113 114 115 116 117
| GRAMMARATTRIBUTE
    { [ unknown_pos (DGrammarAttribute $1) ] }

| PERCENTATTRIBUTE actuals attributes
    { [ unknown_pos (DSymbolAttributes ($2, $3)) ] }

118
| ON_ERROR_REDUCE actuals
119 120
    { let prec = ParserAux.new_on_error_reduce_level() in
      List.map (Positions.map (fun nt -> DOnErrorReduce (nt, prec)))
121 122
        (List.map Parameters.with_pos $2) }

123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162
optional_ocamltype:
  /* epsilon */
    { None }
| OCAMLTYPE /* lexically delimited by angle brackets */
    { Some $1 }

priority_keyword:
  LEFT
    { LeftAssoc }
| RIGHT
    { RightAssoc }
| NONASSOC
    { NonAssoc }

/* ------------------------------------------------------------------------- */
/* A symbol is a terminal or nonterminal symbol. One would like to
   require nonterminal symbols to begin with a lowercase letter, so as
   to lexically distinguish them from terminal symbols, which must
   begin with an uppercase letter. However, for compatibility with
   ocamlyacc, this is impossible. It can be required only for
   nonterminal symbols that are also start symbols. */

symbols:
  /* epsilon */
    { [] }
| symbols optional_comma symbol
    { $3 :: $1 }

symbol:
  LID
    { $1 }
| UID
    { $1 }

optional_comma:
  /* epsilon */
    { () }
| COMMA
    { () }

163 164 165 166 167
attributes:
  /* epsilon */
    { [] }
| ATTRIBUTE attributes { $1 :: $2 }

168 169 170 171 172 173 174
/* ------------------------------------------------------------------------- */
/* Terminals must begin with an uppercase letter. Nonterminals that are
   declared to be start symbols must begin with a lowercase letter. */

terminals:
  /* epsilon */
    { [] }
175 176
| terminals optional_comma UID attributes
    { (Positions.map (fun uid -> (uid, $4)) $3) :: $1 }
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197

nonterminals:
  /* epsilon */
    { [] }
| nonterminals LID
    { $2 :: $1 }

/* ------------------------------------------------------------------------- */
/* A rule defines a symbol. It is optionally declared %public, and optionally
   carries a number of formal parameters. The right-hand side of the definition
   consists of a list of production groups. */

rules:
  /* epsilon */
    { [] }
| rules rule
    { $2 :: $1 }

rule:
  flags
  symbol
198
  attributes
199
  optional_formal_parameters
200
  COLON
201 202
  optional_bar
  production_group production_groups
203
    {
204
      let public, inline = $1 in
POTTIER Francois's avatar
POTTIER Francois committed
205
      { pr_public_flag = public;
206 207 208
        pr_inline_flag = inline;
        pr_nt          = Positions.value $2;
        pr_positions   = [ Positions.position $2 ];
209 210 211
        pr_attributes  = $3;
        pr_parameters  = $4;
        pr_branches    = List.flatten ($7 :: List.rev $8)
212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
      }
    }

flags:
  /* epsilon */
    { false, false }
| PUBLIC
    { true, false }
| INLINE
    { false, true }
| PUBLIC INLINE
    { true, true }
| INLINE PUBLIC
    { true, true }

/* ------------------------------------------------------------------------- */
/* Parameters are surroundered with parentheses and delimited by commas.
   The syntax of actual parameters allows applications, whereas the syntax
   of formal parameters does not. It also allows use of the "?", "+", and
   "*" shortcuts. */

optional_formal_parameters:
  /* epsilon */
    { [] }
| LPAREN formal_parameters RPAREN
    { $2 }

formal_parameters:
  symbol
    { [ Positions.value $1 ] }
| symbol COMMA formal_parameters
    { Positions.value $1 :: $3 }

245
optional_actuals:
246 247
  /* epsilon */
    { [] }
248
| LPAREN actuals_comma RPAREN
249 250
    { $2 }

251
actuals_comma:
POTTIER Francois's avatar
POTTIER Francois committed
252
  actual
253
    { [ $1 ] }
254
| actual COMMA actuals_comma
255 256
    { $1 :: $3 }

257 258
actual:
  symbol optional_actuals
259
    { Parameters.app $1 $2 }
260
| actual modifier
261
    { ParameterApp ($2, [ $1 ]) }
262

263
actuals:
264 265
  /* epsilon */
    { [] }
266
| actuals optional_comma actual
267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298
    { $3::$1 }

optional_bar:
  /* epsilon */ %prec no_optional_bar
    { () }
| BAR
    { () }

/* ------------------------------------------------------------------------- */
/* The "?", "+", and "*" modifiers are short-hands for applications of
   certain parameterized nonterminals, defined in the standard library. */

modifier:
  QUESTION
    { unknown_pos "option" }
| PLUS
    { unknown_pos "nonempty_list" }
| STAR
    { unknown_pos "list" }

/* ------------------------------------------------------------------------- */
/* A production group consists of a list of productions, followed by a
   semantic action and an optional precedence specification. */

production_groups:
  /* epsilon */
    { [] }
| production_groups BAR production_group
    { $3 :: $1 }

production_group:
  productions ACTION /* action is lexically delimited by braces */ optional_precedence
299
    {
300
      let productions, action, oprec2 = $1, $2, $3 in
301 302 303 304
      (* If multiple productions share a single semantic action, check
         that all of them bind the same names. *)
      ParserAux.check_production_group productions;
      (* Then, *)
305
      List.map (fun (producers, oprec1, level, pos) ->
306 307 308 309 310
        (* Replace [$i] with [_i]. *)
        let pr_producers = ParserAux.normalize_producers producers in
        (* Distribute the semantic action. Also, check that every [$i]
           is within bounds. *)
        let pr_action = action (ParserAux.producer_names producers) in
311 312 313 314 315 316 317
        {
          pr_producers;
          pr_action;
          pr_branch_prec_annotation   = ParserAux.override pos oprec1 oprec2;
          pr_branch_production_level  = level;
          pr_branch_position          = pos
        })
318
      productions
319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
    }

optional_precedence:
  /* epsilon */
    { None }
| PREC symbol
    { Some $2 }

/* ------------------------------------------------------------------------- */
/* A production is a list of producers, optionally followed by a
   precedence declaration. Lists of productions are nonempty and
   separated with bars. */

productions:
  production
    { [ $1 ] }
| production bar_productions
    { $1 :: $2 }

bar_productions:
  BAR production
    { [ $2 ] }
| BAR production bar_productions
    { $2 :: $3 }

production:
  producers optional_precedence
    { List.rev $1,
      $2,
348
      ParserAux.new_production_level(),
349
      Positions.import (symbol_start_pos(), symbol_end_pos())
350 351 352
    }

producers:
POTTIER Francois's avatar
POTTIER Francois committed
353
  /* epsilon */
354 355 356 357 358 359
    { [] }
| producers producer
    { $2 :: $1 }

/* ------------------------------------------------------------------------- */
/* A producer is an actual parameter, possibly preceded by a
360
   binding, and possibly followed with attributes. */
361 362

producer:
363
|           actual attributes
364
    { Positions.import (symbol_start_pos(), symbol_end_pos()),    None, $1, $2 }
365
| LID EQUAL actual attributes
366
    { Positions.import (symbol_start_pos(), symbol_end_pos()), Some $1, $3, $4 }
367 368

%%