typer.ml 50.7 KB
Newer Older
1
(* TODO:
2
 - rewrite type-checking of operators to propagate constraint
3
4
 - optimize computation of pattern free variables
 - check whether it is worth using recursive hash-consing internally
5
6
*)

7
8
9
open Location
open Ast
open Ident
10

11
12
let debug_schema = false

13
let warning loc msg =
14
  Format.fprintf !Location.warning_ppf "Warning %a:@\n%a%s@." 
15
16
    Location.print_loc (loc,`Full)
    Location.html_hilight (loc,`Full)
17
18
    msg

19
20
type item =
  | Type of Types.t
21
  | Val of Types.t
22

23
type t = {
24
  ids : item Env.t;
25
26
  ns: Ns.table;
  cu: Types.CompUnit.t Env.t;
27
}
28

29
30
31
32
33
let hash _ = failwith "Typer.hash"
let compare _ _ = failwith "Typer.compare"
let dump ppf _ = failwith "Typer.dump"
let equal _ _ = failwith "Typer.equal"
let check _ = failwith "Typer.check"
34
35

(* TODO: filter out builtin defs ? *)
36
37
38
39
let serialize_item s = function
  | Type t -> Serialize.Put.bits 1 s 0; Types.serialize s t
  | Val t -> Serialize.Put.bits 1 s 1; Types.serialize s t

40
let serialize s env =
41
  Serialize.Put.env Id.serialize serialize_item Env.iter s env.ids;
42
  Ns.serialize_table s env.ns
43

44
45
46
47
48
let deserialize_item s = match Serialize.Get.bits 1 s with
  | 0 -> Type (Types.deserialize s)
  | 1 -> Val (Types.deserialize s)
  | _ -> assert false

49
let deserialize s =
50
51
  let ids = 
    Serialize.Get.env Id.deserialize deserialize_item Env.add Env.empty s in
52
  let ns = Ns.deserialize_table s in
53
  { ids = ids; ns = ns; cu = Env.empty }
54
55


56
57
let empty_env = {
  ids = Env.empty;
58
59
  ns = Ns.empty_table;
  cu = Env.empty;
60
61
}

62
63
let from_comp_unit = ref (fun cu -> assert false)

64
65
66
67
68
69
70
71
72
73
let enter_cu x cu env =
  { env with cu = Env.add (ident x) cu env.cu }

let find_cu loc x env =
  try Env.find x env.cu
  with Not_found -> 
    raise_loc_generic loc 
      ("Unbound compunit prefix " ^ (Ident.to_string x))


74
75
76
77
78
79
80
81
let enter_type id t env =
  { env with ids = Env.add id (Type t) env.ids }
let enter_types l env =
  { env with ids = 
      List.fold_left (fun accu (id,t) -> Env.add id (Type t) accu) env.ids l }
let find_type id env =
  match Env.find id env.ids with
    | Type t -> t
82
    | Val _ -> raise Not_found
83

84
85
let find_type_global loc cu id env =
  let cu = find_cu loc cu env in
86
87
88
  let env = !from_comp_unit cu in
  find_type id env

89
let enter_value id t env = 
90
  { env with ids = Env.add id (Val t) env.ids }
91
92
let enter_values l env =
  { env with ids = 
93
      List.fold_left (fun accu (id,t) -> Env.add id (Val t) accu) env.ids l }
94
95
let find_value id env =
  match Env.find id env.ids with
96
    | Val t -> t
97
    | _ -> raise Not_found
98
99
100
let find_value_global cu id env =
  let env = !from_comp_unit cu in
  find_value id env
101
	
102
103
104
105
106
107
let value_name_ok id env =
  try match Env.find id env.ids with
    | Val t -> true
    | _ -> false
  with Not_found -> true

108
109
110
111
let iter_values env f =
  Env.iter (fun x ->
	      function Val t -> f x t;
		| _ -> ()) env.ids
112

113

114

115
(* Namespaces *)
116

117
let set_ns_table_for_printer env = 
118
  Ns.InternalPrinter.set_table env.ns
119

120
let get_ns_table tenv = tenv.ns
121

122
let enter_ns p ns env =
123
  { env with ns = Ns.add_prefix p ns env.ns }
124

125
126
127
128
129
let protect_error_ns loc f x =
  try f x
  with Ns.UnknownPrefix ns ->
    raise_loc_generic loc 
    ("Undefined namespace prefix " ^ (U.to_string ns))
130

131
let parse_atom env loc t =
132
  let (ns,l) = protect_error_ns loc (Ns.map_tag env.ns) t in
133
134
135
  Atoms.V.mk ns l
 
let parse_ns env loc ns =
136
  protect_error_ns loc (Ns.map_prefix env.ns) ns
137

138
let parse_label env loc t =
139
  let (ns,l) = protect_error_ns loc (Ns.map_attr env.ns) t in
140
  LabelPool.mk (ns,l)
141

142
143
144
145
146
147
148
149
150
151
152
153
154
let parse_record env loc f r =
  let r = List.map (fun (l,x) -> (parse_label env loc l, f x)) r in
  LabelMap.from_list (fun _ _ -> raise_loc_generic loc "Duplicated record field") r

let rec const env loc = function
  | LocatedExpr (loc,e) -> const env loc e
  | Pair (x,y) -> Types.Pair (const env loc x, const env loc y)
  | Xml (x,y) -> Types.Xml (const env loc x, const env loc y)
  | RecordLitt x -> Types.Record (parse_record env loc (const env loc) x)
  | String (i,j,s,c) -> Types.String (i,j,s,const env loc c)
  | Atom t -> Types.Atom (parse_atom env loc t)
  | Integer i -> Types.Integer i
  | Char c -> Types.Char c
155
  | Const c -> c
156
157
158
159
  | _ -> raise_loc_generic loc "This should be a scalar or structured constant"

(* I. Transform the abstract syntax of types and patterns into
      the internal form *)
160

161
exception NonExhaustive of Types.descr
162
exception Constraint of Types.descr * Types.descr
163
exception ShouldHave of Types.descr * string
164
exception ShouldHave2 of Types.descr * string * Types.descr
165
exception WrongLabel of Types.descr * label
166
exception UnboundId of id * bool
167
exception Error of string
168

169
170
let raise_loc loc exn = raise (Location (loc,`Full,exn))
let raise_loc_str loc ofs exn = raise (Location (loc,`Char ofs,exn))
171
let error loc msg = raise_loc loc (Error msg)
172

173
  (* just to remember imported schemas *)
174
let schemas = State.ref "Typer.schemas" (Hashtbl.create 3)
175
176
177

let schema_types = State.ref "Typer.schema_types" (Hashtbl.create 51)
let schema_elements = State.ref "Typer.schema_elements" (Hashtbl.create 51)
178
let schema_attributes = State.ref "Typer.schema_attributes" (Hashtbl.create 51)
179
180
181
182
let schema_attribute_groups =
  State.ref "Typer.schema_attribute_groups" (Hashtbl.create 51)
let schema_model_groups =
  State.ref "Typer.schema_model_groups" (Hashtbl.create 51)
183

184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
  (* raise Not_found *)
let find_schema_descr kind schema name =
  let elt () = Hashtbl.find !schema_elements (schema, name) in
  let typ () = Hashtbl.find !schema_types (schema, name) in
  let att () = Hashtbl.find !schema_attributes (schema, name) in
  let att_group () = Hashtbl.find !schema_attribute_groups (schema, name) in
  let mod_group () = Hashtbl.find !schema_model_groups (schema, name) in
  let rec do_try n = function
    | [] -> raise Not_found
    | f :: rem -> (try f () with Not_found -> do_try n rem)
  in
  match kind with
    | Some `Element -> do_try "element" [ elt ]
    | Some `Type -> do_try "type" [ typ ]
    | Some `Attribute -> do_try "atttribute" [ att ]
    | Some `Attribute_group -> do_try "attribute group" [ att_group ]
    | Some `Model_group -> do_try "model group" [ mod_group ]
    | None ->
        (* policy for unqualified schema component resolution. This order should
         * be consistent with Schema_component.get_component *)
        do_try "component" [ elt; typ; att; att_group; mod_group ]

  (* as above, but raise Error *)
let find_schema_descr' k s n =
  try
    find_schema_descr k s n
  with Not_found ->
211
212
    raise (Error (Printf.sprintf "No %s named '%s' found in schema '%s'"
      (Schema_common.string_of_component_kind k) (U.get_str n) (U.get_str s)))
213

214
215
216
217
218
219
220
221
(* Eliminate Recursion, propagate Sequence Capture Variables *)

let rec seq_vars accu = function
  | Epsilon | Elem _ -> accu
  | Seq (r1,r2) | Alt (r1,r2) -> seq_vars (seq_vars accu r1) r2
  | Star r | WeakStar r -> seq_vars accu r
  | SeqCapture (v,r) -> seq_vars (IdSet.add v accu) r

222
223
224
225
226
227
228
229
230
231
232
233
234
235
(* We use two intermediate representation from AST types/patterns
   to internal ones:

      AST -(1)-> derecurs -(2)-> slot -(3)-> internal

   (1) eliminate recursion, schema, 
       propagate sequence capture variables, keep regexps

   (2) stratify, detect ill-formed recursion, compile regexps

   (3) check additional constraints on types / patterns;
       deep (recursive) hash-consing
*)     

236
237
238
239
type derecurs_slot = {
  ploc : Location.loc;
  pid  : int;
  mutable ploop : bool;
240
  mutable pdescr : derecurs;
241
} and derecurs =
242
  | PDummy
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
  | PAlias of derecurs_slot
  | PType of Types.descr
  | POr of derecurs * derecurs
  | PAnd of derecurs * derecurs
  | PDiff of derecurs * derecurs
  | PTimes of derecurs * derecurs
  | PXml of derecurs * derecurs
  | PArrow of derecurs * derecurs
  | POptional of derecurs
  | PRecord of bool * derecurs label_map
  | PCapture of id
  | PConstant of id * Types.const
  | PRegexp of derecurs_regexp * derecurs
and derecurs_regexp =
  | PEpsilon
  | PElem of derecurs
  | PSeq of derecurs_regexp * derecurs_regexp
  | PAlt of derecurs_regexp * derecurs_regexp
  | PStar of derecurs_regexp
  | PWeakStar of derecurs_regexp

264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283

type descr = 
  | IDummy
  | IType of Types.descr
  | IOr of descr * descr
  | IAnd of descr * descr
  | IDiff of descr * descr
  | ITimes of slot * slot
  | IXml of slot * slot
  | IArrow of slot * slot
  | IOptional of descr
  | IRecord of bool * slot label_map
  | ICapture of id
  | IConstant of id * Types.const
and slot = {
  mutable fv : fv option;
  mutable hash : int option;
  mutable rank1: int; mutable rank2: int;
  mutable gen1 : int; mutable gen2: int;
  mutable d    : descr;
284
}
285
286
287
288
289
290
291
292
293
294
295
296
297
    

let counter = ref 0
let mk_derecurs_slot loc = 
  incr counter; 
  { ploop = false; ploc = loc; pid = !counter; pdescr = PDummy }
	  
let mk_slot () = 
  { d=IDummy; fv=None; hash=None; rank1=0; rank2=0; gen1=0; gen2=0 } 


(* This environment is used in phase (1) to eliminate recursion *)
type penv = {
298
  penv_tenv : t;
299
300
301
302
  penv_derec : derecurs_slot Env.t;
}

let penv tenv = { penv_tenv = tenv; penv_derec = Env.empty }
303

304
let rec hash_derecurs = function
305
  | PDummy -> assert false
306
307
308
  | PAlias s -> 
      s.pid
  | PType t -> 
309
      1 + 17 * (Types.hash t)
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
  | POr (p1,p2) -> 
      2 + 17 * (hash_derecurs p1) + 257 * (hash_derecurs p2)
  | PAnd (p1,p2) -> 
      3 + 17 * (hash_derecurs p1) + 257 * (hash_derecurs p2)
  | PDiff (p1,p2) -> 
      4 + 17 * (hash_derecurs p1) + 257 * (hash_derecurs p2)
  | PTimes (p1,p2) -> 
      5 + 17 * (hash_derecurs p1) + 257 * (hash_derecurs p2)
  | PXml (p1,p2) -> 
      6 + 17 * (hash_derecurs p1) + 257 * (hash_derecurs p2)
  | PArrow (p1,p2) -> 
      7 + 17 * (hash_derecurs p1) + 257 * (hash_derecurs p2)
  | POptional p -> 
      8 + 17 * (hash_derecurs p)
  | PRecord (o,r) -> 
      (if o then 9 else 10) + 17 * (LabelMap.hash hash_derecurs r)
  | PCapture x -> 
      11 + 17 * (Id.hash x)
  | PConstant (x,c) -> 
329
      12 + 17 * (Id.hash x) + 257 * (Types.Const.hash c)
330
331
  | PRegexp (p,q) -> 
      13 + 17 * (hash_derecurs_regexp p) + 257 * (hash_derecurs q)
332
and hash_derecurs_regexp = function
333
334
335
336
337
338
339
340
341
342
343
344
  | PEpsilon -> 
      1
  | PElem p -> 
      2 + 17 * (hash_derecurs p)
  | PSeq (p1,p2) -> 
      3 + 17 * (hash_derecurs_regexp p1) + 257 * (hash_derecurs_regexp p2)
  | PAlt (p1,p2) -> 
      4 + 17 * (hash_derecurs_regexp p1) + 257 * (hash_derecurs_regexp p2)
  | PStar p -> 
      5 + 17 * (hash_derecurs_regexp p)
  | PWeakStar p -> 
      6 + 17 * (hash_derecurs_regexp p)
345
346

let rec equal_derecurs p1 p2 = (p1 == p2) || match p1,p2 with
347
348
349
  | PAlias s1, PAlias s2 -> 
      s1 == s2
  | PType t1, PType t2 -> 
350
      Types.equal t1 t2
351
352
353
354
355
  | POr (p1,q1), POr (p2,q2)
  | PAnd (p1,q1), PAnd (p2,q2)
  | PDiff (p1,q1), PDiff (p2,q2)
  | PTimes (p1,q1), PTimes (p2,q2)
  | PXml (p1,q1), PXml (p2,q2)
356
357
358
359
360
361
362
363
364
  | PArrow (p1,q1), PArrow (p2,q2) -> 
      (equal_derecurs p1 p2) && (equal_derecurs q1 q2)
  | POptional p1, POptional p2 -> 
      equal_derecurs p1 p2
  | PRecord (o1,r1), PRecord (o2,r2) -> 
      (o1 == o2) && (LabelMap.equal equal_derecurs r1 r2)
  | PCapture x1, PCapture x2 -> 
      Id.equal x1 x2
  | PConstant (x1,c1), PConstant (x2,c2) -> 
365
      (Id.equal x1 x2) && (Types.Const.equal c1 c2)
366
367
  | PRegexp (p1,q1), PRegexp (p2,q2) -> 
      (equal_derecurs_regexp p1 p2) && (equal_derecurs q1 q2)
368
369
  | _ -> false
and equal_derecurs_regexp r1 r2 = match r1,r2 with
370
371
372
373
  | PEpsilon, PEpsilon -> 
      true
  | PElem p1, PElem p2 -> 
      equal_derecurs p1 p2
374
  | PSeq (p1,q1), PSeq (p2,q2) 
375
376
  | PAlt (p1,q1), PAlt (p2,q2) -> 
      (equal_derecurs_regexp p1 p2) && (equal_derecurs_regexp q1 q2)
377
  | PStar p1, PStar p2
378
379
  | PWeakStar p1, PWeakStar p2 -> 
      equal_derecurs_regexp p1 p2
380
  | _ -> false
381

382
383
384
385
386
387
388
389
390
391
392
module DerecursTable = Hashtbl.Make(
  struct 
    type t = derecurs 
    let hash = hash_derecurs
    let equal = equal_derecurs
  end
)

module RE = Hashtbl.Make(
  struct 
    type t = derecurs_regexp * derecurs 
393
394
395
396
    let hash (p,q) = 
      (hash_derecurs_regexp p) + 17 * (hash_derecurs q)
    let equal (p1,q1) (p2,q2) = 
      (equal_derecurs_regexp p1 p2) && (equal_derecurs q1 q2)
397
398
  end
)
399

400
401
402
403
let gen = ref 0
let rank = ref 0
	     
let rec hash_descr = function
404
  | IDummy -> assert false
405
  | IType x -> Types.hash x
406
407
408
409
410
411
412
413
414
  | IOr (d1,d2) -> 1 + 17 * (hash_descr d1) + 257 * (hash_descr d2)
  | IAnd (d1,d2) -> 2 + 17 * (hash_descr d1) + 257 * (hash_descr d2)
  | IDiff (d1,d2) -> 3 + 17 * (hash_descr d1) + 257 * (hash_descr d2)
  | IOptional d -> 4 + 17 * (hash_descr d)
  | ITimes (s1,s2) -> 5 + 17 * (hash_slot s1) + 257 * (hash_slot s2)
  | IXml (s1,s2) -> 6 + 17 * (hash_slot s1) + 257 * (hash_slot s2)
  | IArrow (s1,s2) -> 7 + 17 * (hash_slot s1) + 257 * (hash_slot s2)
  | IRecord (o,r) -> (if o then 8 else 9) + 17 * (LabelMap.hash hash_slot r)
  | ICapture x -> 10 + 17 * (Id.hash x)
415
  | IConstant (x,y) -> 11 + 17 * (Id.hash x) + 257 * (Types.Const.hash y)
416
417
418
419
420
and hash_slot s =
  if s.gen1 = !gen then 13 * s.rank1
  else (
    incr rank;
    s.rank1 <- !rank; s.gen1 <- !gen;
421
    hash_descr s.d
422
423
424
425
  )
    
let rec equal_descr d1 d2 = 
  match (d1,d2) with
426
  | IType x1, IType x2 -> Types.equal x1 x2
427
428
429
430
431
432
433
  | IOr (x1,y1), IOr (x2,y2) 
  | IAnd (x1,y1), IAnd (x2,y2) 
  | IDiff (x1,y1), IDiff (x2,y2) -> (equal_descr x1 x2) && (equal_descr y1 y2)
  | IOptional x1, IOptional x2 -> equal_descr x1 x2
  | ITimes (x1,y1), ITimes (x2,y2) 
  | IXml (x1,y1), IXml (x2,y2) 
  | IArrow (x1,y1), IArrow (x2,y2) -> (equal_slot x1 x2) && (equal_slot y1 y2)
434
435
  | IRecord (o1,r1), IRecord (o2,r2) -> 
      (o1 = o2) && (LabelMap.equal equal_slot r1 r2)
436
  | ICapture x1, ICapture x2 -> Id.equal x1 x2
437
  | IConstant (x1,y1), IConstant (x2,y2) -> 
438
      (Id.equal x1 x2) && (Types.Const.equal y1 y2)
439
440
441
442
443
444
445
446
  | _ -> false
and equal_slot s1 s2 =
  ((s1.gen1 = !gen) && (s2.gen2 = !gen) && (s1.rank1 = s2.rank2))
  ||
  ((s1.gen1 <> !gen) && (s2.gen2 <> !gen) && (
     incr rank;
     s1.rank1 <- !rank; s1.gen1 <- !gen;
     s2.rank2 <- !rank; s2.gen2 <- !gen;
447
     equal_descr s1.d s2.d
448
449
   ))
  
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
module SlotTable = Hashtbl.Make(
  struct
    type t = slot
	
    let hash s =
      match s.hash with
	| Some h -> h
	| None ->
	    incr gen; rank := 0; 
	    let h = hash_slot s in
	    s.hash <- Some h;
	    h
	      
    let equal s1 s2 = 
      (s1 == s2) || 
      (incr gen; rank := 0; 
       let e = equal_slot s1 s2 in
       (*     if e then Printf.eprintf "Recursive hash-consing: Equal\n";  *)
       e)
  end)


let rec derecurs env p = match p.descr with
  | PatVar v ->
474
475
476
477
478
479
480
481
482
483
      (match Ns.split_qname v with
	 | "", v ->
	     let v = ident v in
	     (try PAlias (Env.find v env.penv_derec)
	      with Not_found -> 
		try PType (find_type v env.penv_tenv)
		with Not_found -> PCapture v)
	 | cu, v -> 
	     try 
	       let cu = ident (U.mk cu) in
484
	       PType (find_type_global p.loc cu (ident v) env.penv_tenv)
485
	     with Not_found ->
486
487
	       raise_loc_generic p.loc 
	       ("Unbound external type " ^ cu ^ ":" ^ (U.to_string v)))
488
489
  | SchemaVar (kind, schema_name, component_name) ->
      PType (derecurs_schema env kind schema_name component_name)
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
  | Recurs (p,b) -> derecurs (derecurs_def env b) p
  | Internal t -> PType t
  | NsT ns -> PType (Types.atom (Atoms.any_in_ns (parse_ns env.penv_tenv p.loc ns)))
  | Or (p1,p2) -> POr (derecurs env p1, derecurs env p2)
  | And (p1,p2) -> PAnd (derecurs env p1, derecurs env p2)
  | Diff (p1,p2) -> PDiff (derecurs env p1, derecurs env p2)
  | Prod (p1,p2) -> PTimes (derecurs env p1, derecurs env p2)
  | XmlT (p1,p2) -> PXml (derecurs env p1, derecurs env p2)
  | Arrow (p1,p2) -> PArrow (derecurs env p1, derecurs env p2)
  | Optional p -> POptional (derecurs env p)
  | Record (o,r) -> PRecord (o, parse_record env.penv_tenv p.loc (derecurs env) r)
  | Constant (x,c) -> PConstant (x,const env.penv_tenv p.loc c)
  | Cst c -> PType (Types.constant (const env.penv_tenv p.loc c))
  | Regexp (r,q) -> 
      let constant_nil t v = 
	PAnd (t, PConstant (v, Types.Atom Sequence.nil_atom)) in
      let vars = seq_vars IdSet.empty r in
      let q = IdSet.fold constant_nil (derecurs env q) vars in
      let r = derecurs_regexp (fun p -> p) env r in
      PRegexp (r, q)
and derecurs_regexp vars env = function
  | Epsilon -> 
      PEpsilon
  | Elem p -> 
      PElem (vars (derecurs env p))
  | Seq (p1,p2) -> 
      PSeq (derecurs_regexp vars env p1, derecurs_regexp vars env p2)
  | Alt (p1,p2) -> 
      PAlt (derecurs_regexp vars env p1, derecurs_regexp vars env p2)
  | Star p -> 
      PStar (derecurs_regexp vars env p)
  | WeakStar p -> 
      PWeakStar (derecurs_regexp vars env p)
  | SeqCapture (x,p) -> 
      derecurs_regexp (fun p -> PAnd (vars p, PCapture x)) env p


and derecurs_def env b =
  let b = List.map (fun (v,p) -> (v,p,mk_derecurs_slot p.loc)) b in
  let n = 
    List.fold_left (fun env (v,p,s) -> Env.add v s env) env.penv_derec b in
  let env = { env with penv_derec = n } in
  List.iter (fun (v,p,s) -> s.pdescr <- derecurs env p) b;
  env

535
and derecurs_schema env = find_schema_descr
536
    
537
538
539
540
541
let rec fv_slot s =
  match s.fv with
    | Some x -> x
    | None ->
	if s.gen1 = !gen then IdSet.empty 
542
	else (s.gen1 <- !gen; fv_descr s.d)
543
and fv_descr = function
544
  | IDummy -> assert false
545
  | IType _ -> IdSet.empty
546
547
548
549
550
551
552
  | IOr (d1,d2)
  | IAnd (d1,d2)  
  | IDiff (d1,d2) -> IdSet.cup (fv_descr d1) (fv_descr d2)
  | IOptional d -> fv_descr d
  | ITimes (s1,s2)  
  | IXml (s1,s2)  
  | IArrow (s1,s2) -> IdSet.cup (fv_slot s1) (fv_slot s2)
553
554
  | IRecord (o,r) -> 
      List.fold_left IdSet.cup IdSet.empty (LabelMap.map_to_list fv_slot r)
555
  | ICapture x | IConstant (x,_) -> IdSet.singleton x
556

557
558
559
560
561
562
563
564
let compute_fv s =
  match s.fv with
    | Some x -> ()
    | None ->
	incr gen;
	let x = fv_slot s in
	s.fv <- Some x
	  
565
566
567
let check_no_capture loc s =
  match IdSet.pick s with
    | Some x ->  
568
	raise_loc_generic loc ("Capture variable not allowed: " ^ (Ident.to_string x))
569
    | None -> ()
570
    
571
572
573
let compile_slot_hash = DerecursTable.create 67
let compile_hash = DerecursTable.create 67

574
575
let todo_defs = ref []
let todo_fv = ref []
576
577
578
579
580
581
582
583

let rec compile p =
  try DerecursTable.find compile_hash p
  with Not_found ->
    let c = real_compile p in
    DerecursTable.replace compile_hash p c;
    c
and real_compile = function
584
  | PDummy -> assert false
585
586
587
588
  | PAlias v ->
      if v.ploop then
	raise_loc_generic v.ploc ("Unguarded recursion on type/pattern");
      v.ploop <- true;
589
      let r = compile v.pdescr in
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
      v.ploop <- false;
      r
  | PType t -> IType t
  | POr (t1,t2) -> IOr (compile t1, compile t2)
  | PAnd (t1,t2) -> IAnd (compile t1, compile t2)
  | PDiff (t1,t2) -> IDiff (compile t1, compile t2)
  | PTimes (t1,t2) -> ITimes (compile_slot t1, compile_slot t2)
  | PXml (t1,t2) -> IXml (compile_slot t1, compile_slot t2)
  | PArrow (t1,t2) -> IArrow (compile_slot t1, compile_slot t2)
  | POptional t -> IOptional (compile t)
  | PRecord (o,r) ->  IRecord (o, LabelMap.map compile_slot r)
  | PConstant (x,v) -> IConstant (x,v)
  | PCapture x -> ICapture x
  | PRegexp (r,q) -> compile_regexp r q
and compile_regexp r q =
  let memo = RE.create 17 in
606
607
608
  let add accu i = 
    match accu with None -> Some i | Some j -> Some (IOr (j,i)) in
  let get = function Some x -> x | None -> assert false in
609
  let rec queue accu = function
610
611
612
    | PRegexp (r,q) -> aux accu r q 
    | _ -> add accu (compile q)
  and aux accu r q =
613
614
615
616
    if RE.mem memo (r,q) then accu
    else (
      RE.add memo (r,q) ();
      match r with
617
	| PEpsilon -> queue accu q
618
619
620
621
622
623
624
625
626
627
628
	| PElem p ->
(* Be careful not to create pairs with same second component *)
	    let rec extract = function
	      | PConstant (x,v) -> `Const (x,v)
	      | POr (x,y) ->
		  (match extract x, extract y with
		    | `Pat x, `Pat y -> `Pat (POr (x,y))
		    | x, y -> `Or (x,y))
	      | p -> `Pat p
	    in
	    let rec mk accu = function
629
630
631
632
	      | `Const (x,v) -> 
		  (match queue None q with 
		    | Some q -> add accu (IAnd (IConstant (x,v), q))
		    | None -> accu)
633
	      | `Or (x,y) -> mk (mk accu x) y
634
635
	      | `Pat p -> 
		  add accu (ITimes (compile_slot p, compile_slot q))
636
637
	    in
	    mk accu (extract p)
638
639
640
641
642
643
	| PSeq (r1,r2) -> aux accu r1 (PRegexp (r2,q))
	| PAlt (r1,r2) -> aux (aux accu r1 q) r2 q
	| PStar r1 -> aux (aux accu r1 (PRegexp (r,q))) PEpsilon q
	| PWeakStar r1 -> aux (aux accu PEpsilon q) r1 (PRegexp (r,q))
    )
  in
644
  get (aux None r q)
645
646
and compile_slot p =
  try DerecursTable.find compile_slot_hash p
647
  with Not_found ->
648
649
650
    let s = mk_slot () in
    todo_defs := (s,p) :: !todo_defs;
    todo_fv := s :: !todo_fv;
651
    DerecursTable.add compile_slot_hash p s;
652
    s
653

654
      
655
let timer_fv = Stats.Timer.create "Typer.fv"
656
let rec flush_defs () = 
657
658
659
660
  match !todo_defs with
    | [] -> 
	Stats.Timer.start timer_fv;
	List.iter compute_fv !todo_fv;
661
662
	todo_fv := [];
	Stats.Timer.stop timer_fv ()
663
664
665
666
    | (s,p)::t -> 
	todo_defs := t; 
	s.d <- compile p; 
	flush_defs ()
667
668
669
670
671
672
673
674
675
676
677
678
679
680
	
let typ_nodes = SlotTable.create 67
let pat_nodes = SlotTable.create 67
		  
let rec typ = function
  | IType t -> t
  | IOr (s1,s2) -> Types.cup (typ s1) (typ s2)
  | IAnd (s1,s2) ->  Types.cap (typ s1) (typ s2)
  | IDiff (s1,s2) -> Types.diff (typ s1) (typ s2)
  | ITimes (s1,s2) -> Types.times (typ_node s1) (typ_node s2)
  | IXml (s1,s2) -> Types.xml (typ_node s1) (typ_node s2)
  | IArrow (s1,s2) -> Types.arrow (typ_node s1) (typ_node s2)
  | IOptional s -> Types.Record.or_absent (typ s)
  | IRecord (o,r) -> Types.record' (o, LabelMap.map typ_node r)
681
  | IDummy | ICapture _ | IConstant (_,_) -> assert false
682
      
683
and typ_node s : Types.Node.t =
684
685
686
687
  try SlotTable.find typ_nodes s
  with Not_found ->
    let x = Types.make () in
    SlotTable.add typ_nodes s x;
688
    Types.define x (typ s.d);
689
690
691
692
693
694
695
696
    x
      
let rec pat d : Patterns.descr =
  if IdSet.is_empty (fv_descr d)
  then Patterns.constr (typ d)
  else pat_aux d
    
and pat_aux = function
697
  | IDummy -> assert false
698
699
700
701
702
703
  | IOr (s1,s2) -> Patterns.cup (pat s1) (pat s2)
  | IAnd (s1,s2) -> Patterns.cap (pat s1) (pat s2)
  | IDiff (s1,s2) when IdSet.is_empty (fv_descr s2) ->
      let s2 = Types.neg (typ s2) in
      Patterns.cap (pat s1) (Patterns.constr s2)
  | IDiff _ ->
704
      raise (Patterns.Error "Differences are not allowed in patterns")
705
706
707
  | ITimes (s1,s2) -> Patterns.times (pat_node s1) (pat_node s2)
  | IXml (s1,s2) -> Patterns.xml (pat_node s1) (pat_node s2)
  | IOptional _ -> 
708
      raise (Patterns.Error "Optional fields are not allowed in record patterns")
709
710
711
712
713
714
715
716
717
718
719
720
721
722
  | IRecord (o,r) ->
      let pats = ref [] in
      let aux l s = 
	if IdSet.is_empty (fv_slot s) then typ_node s
	else
	  ( pats := Patterns.record l (pat_node s) :: !pats;
	    Types.any_node )
      in
      let constr = Types.record' (o,LabelMap.mapi aux r) in
      List.fold_left Patterns.cap (Patterns.constr constr) !pats
	(* TODO: can avoid constr when o=true, and all fields have fv *)
  | ICapture x -> Patterns.capture x
  | IConstant (x,c) -> Patterns.constant x c
  | IArrow _ ->
723
      raise (Patterns.Error "Arrows are not allowed in patterns")
724
725
726
727
728
729
  | IType _ -> assert false
      
and pat_node s : Patterns.node =
  try SlotTable.find pat_nodes s
  with Not_found ->
    let x = Patterns.make (fv_slot s) in
730
731
    try
      SlotTable.add pat_nodes s x;
732
      Patterns.define x (pat s.d);
733
734
735
      x
    with exn -> SlotTable.remove pat_nodes s; raise exn
      (* For the toplevel ... *)
736

737

738
let type_defs env b =
739
740
  List.iter 
    (fun (v,p) ->
741
742
       if Env.mem v env.ids
       then raise_loc_generic p.loc ("Identifier " ^ (Ident.to_string v) ^ " is already bound")
743
    ) b;
744
745
  let penv = derecurs_def (penv env) b in
  let b = List.map (fun (v,p) -> (v,p,compile (derecurs penv p))) b in
746
747
748
749
  flush_defs ();
  let b = 
    List.map 
      (fun (v,p,s) -> 
750
	 check_no_capture p.loc (fv_descr s);
751
752
753
	 let t = typ s in
	 if (p.loc <> noloc) && (Types.is_empty t) then
	   warning p.loc 
754
	     ("This definition yields an empty type for " ^ (Ident.to_string v));
755
	 (v,t)) b in
756
  List.iter (fun (v,t) -> Types.Print.register_global (Id.value v) t) b;
757
  b
758
759


760
761
762
763
764
let dump_types ppf env =
  Env.iter (fun v -> 
	      function 
		  (Type _) -> Format.fprintf ppf " %a" Ident.print v
		| _ -> ()) env.ids
765
766
let dump_type ppf env name =
  try
767
    (match Env.find (Ident.ident name) env.ids with
768
769
    | Type t -> Types.Print.print ppf t
    | _ -> raise Not_found)
770
771
  with Not_found ->
    raise (Error (Printf.sprintf "Type %s not found" (U.get_str name)))
772
773
774
775

let dump_schema_type ppf (k, s, n) =
  let descr = find_schema_descr' k s n in
  Types.Print.print ppf descr
776

777
let dump_ns ppf env =
778
  Ns.dump_table ppf env.ns
779

780

781
782
let do_typ loc r = 
  let s = compile_slot r in
783
  flush_defs ();
784
785
  check_no_capture loc (fv_slot s);
  typ_node s
786
   
787
788
let typ env p =
  do_typ p.loc (derecurs (penv env) p)
789
    
790
791
let pat env p = 
  let s = compile_slot (derecurs (penv env) p) in
792
793
794
  flush_defs ();
  try pat_node s
  with Patterns.Error e -> raise_loc_generic p.loc e
795
    | Location (loc,_,exn) when loc = noloc -> raise (Location (p.loc, `Full, exn))
796
797


798
799
(* II. Build skeleton *)

800

801
802
803
804
805
type type_fun = Types.t -> bool -> Types.t
let mk_unary_op = ref (fun _ _ -> assert false)
let typ_unary_op = ref (fun _ _ _ -> assert false)
let mk_binary_op = ref (fun _ _ -> assert false)
let typ_binary_op = ref (fun _ _ _ _ -> assert false)
806
807


808
module Fv = IdSet
809

810
811
812
type branch = Branch of Typed.branch * branch list

let cur_branch : branch list ref = ref []
813

814
let exp loc fv e =
815
816
  fv,
  { Typed.exp_loc = loc;
817
    Typed.exp_typ = Types.empty;
818
    Typed.exp_descr = e;
819
  }
820
821


822
823
let rec expr env loc = function
  | LocatedExpr (loc,e) -> expr env loc e
824
  | Forget (e,t) ->
825
      let (fv,e) = expr env loc e and t = typ env t in
826
827
      exp loc fv (Typed.Forget (e,t))
  | Var s -> 
828
829
830
831
      (match Ns.split_qname s with
	| "", id -> let id = ident id in
	  exp loc (Fv.singleton id) (Typed.Var id)
	| cu, id -> 
832
	    let cu = find_cu loc (ident (U.mk cu)) env in
833
	    exp loc Fv.empty (Typed.ExtVar (cu, ident id)))
834
  | Apply (e1,e2) -> 
835
      let (fv1,e1) = expr env loc e1 and (fv2,e2) = expr env loc e2 in
836
837
      exp loc (Fv.cup fv1 fv2) (Typed.Apply (e1,e2))
  | Abstraction a ->
838
      let iface = List.map (fun (t1,t2) -> (typ env t1, typ env t2)) 
839
840
841
842
843
844
845
		    a.fun_iface in
      let t = List.fold_left 
		(fun accu (t1,t2) -> Types.cap accu (Types.arrow t1 t2)) 
		Types.any iface in
      let iface = List.map 
		    (fun (t1,t2) -> (Types.descr t1, Types.descr t2)) 
		    iface in
846
      let (fv0,body) = branches env a.fun_body in
847
848
849
850
851
852
853
854
855
856
857
      let fv = match a.fun_name with
	| None -> fv0
	| Some f -> Fv.remove f fv0 in
      let e = Typed.Abstraction 
		{ Typed.fun_name = a.fun_name;
		  Typed.fun_iface = iface;
		  Typed.fun_body = body;
		  Typed.fun_typ = t;
		  Typed.fun_fv = fv
		} in
      exp loc fv e
858
  | (Integer _ | Char _ | Atom _ | Const _) as c -> 
859
      exp loc Fv.empty (Typed.Cst (const env loc c))
860
  | Pair (e1,e2) ->
861
      let (fv1,e1) = expr env loc e1 and (fv2,e2) = expr env loc e2 in
862
863
      exp loc (Fv.cup fv1 fv2) (Typed.Pair (e1,e2))
  | Xml (e1,e2) ->
864
      let (fv1,e1) = expr env loc e1 and (fv2,e2) = expr env loc e2 in
865
866
      exp loc (Fv.cup fv1 fv2) (Typed.Xml (e1,e2))
  | Dot (e,l) ->
867
868
      let (fv,e) = expr env loc e in
      exp loc fv (Typed.Dot (e,parse_label env loc l))
869
  | RemoveField (e,l) ->
870
871
      let (fv,e) = expr env loc e in
      exp loc fv (Typed.RemoveField (e,parse_label env loc l))
872
873
  | RecordLitt r -> 
      let fv = ref Fv.empty in
874
      let r = parse_record env loc
875
		(fun e -> 
876
		   let (fv2,e) = expr env loc e 
877
878
879
		   in fv := Fv.cup !fv fv2; e)
		r in
      exp loc !fv (Typed.RecordLitt r)
880
  | String (i,j,s,e) ->
881
      let (fv,e) = expr env loc e in
882
      exp loc fv (Typed.String (i,j,s,e))
883
  | Op (op,le) ->
884
      let (fvs,ltes) = List.split (List.map (expr env loc) le) in
885
      let fv = List.fold_left Fv.cup Fv.empty fvs in
886
      (try
887
888
889
	 (match ltes with
	    | [e] -> exp loc fv (Typed.UnaryOp (!mk_unary_op op env, e))
	    | [e1;e2] -> exp loc fv (Typed.BinaryOp (!mk_binary_op op env, e1,e2))
890
891
892
	    | _ -> assert false)
       with Not_found -> assert false)

893
  | Match (e,b) -> 
894
895
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
896
      exp loc (Fv.cup fv1 fv2) (Typed.Match (e, b))
897
  | Map (e,b) ->
898
899
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
900
901
      exp loc (Fv.cup fv1 fv2) (Typed.Map (e, b))
  | Transform (e,b) ->
902
903
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
904
      exp loc (Fv.cup fv1 fv2) (Typed.Transform (e, b))
905
  | Xtrans (e,b) ->
906
907
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
908
      exp loc (Fv.cup fv1 fv2) (Typed.Xtrans (e, b))
909
  | Validate (e,kind,schema,elt) ->
910
      let (fv,e) = expr env loc e in
911
      exp loc fv (Typed.Validate (e, kind, schema, elt))
912
  | Try (e,b) ->
913
914
      let (fv1,e) = expr env loc e
      and (fv2,b) = branches env b in
915
      exp loc (Fv.cup fv1 fv2) (Typed.Try (e, b))
916
  | NamespaceIn (pr,ns,e) ->
917
918
      let env = enter_ns pr ns env in
      expr env loc e
919
  | Ref (e,t) ->
920
      let (fv,e) = expr env loc e and t = typ env t in
921
      exp loc fv (Typed.Ref (e,t))
922
	      
923
  and branches env b = 
924
    let fv = ref Fv.empty in
925
    let accept = ref Types.empty in
926
    let branch (p,e) = 
927
928
      let cur_br = !cur_branch in
      cur_branch := [];
929
      let (fv2,e) = expr env noloc e in
930
      let br_loc = merge_loc p.loc e.Typed.exp_loc in
931
      let p = pat env p in
932
933
934
935
936
937
      (match Fv.pick (Fv.diff (Patterns.fv p) fv2) with
	| None -> ()
	| Some x ->
	    let x = U.to_string (Id.value x) in
	    warning br_loc 
	      ("The capture variable " ^ x ^ 
938
	       " is declared in the pattern but not used in the body of this branch. It might be a misspelled or undeclared type or name (if it isn't, use _ instead)."));
939
940
941
942
943
944
945
946
947
      let fv2 = Fv.diff fv2 (Patterns.fv p) in
      fv := Fv.cup !fv fv2;
      accept := Types.cup !accept (Types.descr (Patterns.accept p));
      let br = 
	{ 
	  Typed.br_loc = br_loc;
	  Typed.br_used = br_loc = noloc;
	  Typed.br_pat = p;
	  Typed.br_body = e } in
948
      cur_branch := Branch (br, !cur_branch) :: cur_br;
949
950
      br in
    let b = List.map branch b in
951
952
953
954
    (!fv, 
     { 
       Typed.br_typ = Types.empty; 
       Typed.br_branches = b; 
955
956
       Typed.br_accept = !accept;
       Typed.br_compiled = None;
957
958
     } 
    )
959

960
let expr env e = snd (expr env noloc e)
961

962
963
let let_decl env p e =
  { Typed.let_pat = pat env p;
964
    Typed.let_body = expr env e;
965
966
    Typed.let_compiled = None }

967
968
969

(* Hide global "typing/parsing" environment *)

970

971
972
(* III. Type-checks *)

973
974
open Typed

975
976
let require loc t s = 
  if not (Types.subtype t s) then raise_loc loc (Constraint (t, s))
977

978
let verify loc t s = 
979
980
  require loc t s; t