types.ml 41.9 KB
Newer Older
1
open Ident
2
open Encodings
3

4
5
6
7
8
9
10
11
12
13
(*
To be sure not to use generic comparison ...
*)
let (=) : int -> int -> bool = (==)
let (<) : int -> int -> bool = (<)
let (<=) : int -> int -> bool = (<=)
let (<>) : int -> int -> bool = (<>)
let compare = 1


14
type const = 
15
16
17
  | Integer of Intervals.V.t
  | Atom of Atoms.V.t
  | Char of Chars.V.t
18

19
20
let compare_const c1 c2 =
  match (c1,c2) with
21
    | Integer x, Integer y -> Intervals.V.compare x y
22
23
    | Integer _, _ -> -1
    | _, Integer _ -> 1
24
    | Atom x, Atom y -> Atoms.V.compare x y
25
26
    | Atom _, _ -> -1
    | _, Atom _ -> 1
27
    | Char x, Char y -> Chars.V.compare x y
28
29

let hash_const = function
30
31
32
  | Integer x -> Intervals.V.hash x
  | Atom x -> Atoms.V.hash x
  | Char x -> Chars.V.hash x
33

34
35
let equal_const c1 c2 = compare_const c1 c2 = 0

36
37
type pair_kind = [ `Normal | `XML ]

38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
module rec Descr : 
sig
(*
  Want to write:
    type s = { ... }
    include Custom.T with type t = s
  but a  bug in OCaml 3.07+beta 2 makes it impossible
*)
  type t = {
    atoms : Atoms.t;
    ints  : Intervals.t;
    chars : Chars.t;
    times : BoolPair.t;
    xml   : BoolPair.t;
    arrow : BoolPair.t;
    record: BoolRec.t;
    absent: bool
  }
  val dump: Format.formatter -> t -> unit
  val check: t -> unit
  val equal: t -> t -> bool
  val hash: t -> int
  val compare:t -> t -> int
  val serialize: t Serialize.Put.f
  val deserialize: t Serialize.Get.f
end =
struct
  include Custom.Dummy
  type t = {
    atoms : Atoms.t;
    ints  : Intervals.t;
    chars : Chars.t;
    times : BoolPair.t;
    xml   : BoolPair.t;
    arrow : BoolPair.t;
    record: BoolRec.t;
    absent: bool
  }
  let equal a b =
    (Atoms.equal a.atoms b.atoms) &&
    (Chars.equal a.chars b.chars) &&
    (Intervals.equal a.ints  b.ints) &&
    (BoolPair.equal a.times b.times) &&
    (BoolPair.equal a.xml b.xml) &&
    (BoolPair.equal a.arrow b.arrow) &&
    (BoolRec.equal a.record b.record) &&
    (a.absent == b.absent)

  let compare a b =
    if a == b then 0 
    else let c = Atoms.compare a.atoms b.atoms in if c <> 0 then c
    else let c = Chars.compare a.chars b.chars in if c <> 0 then c
    else let c = Intervals.compare a.ints b.ints in if c <> 0 then c
    else let c = BoolPair.compare a.times b.times in if c <> 0 then c
    else let c = BoolPair.compare a.xml b.xml in if c <> 0 then c
    else let c = BoolPair.compare a.arrow b.arrow in if c <> 0 then c
    else let c = BoolRec.compare a.record b.record in if c <> 0 then c
    else if a.absent && not b.absent then -1
    else if b.absent && not a.absent then 1
    else 0
98
      
99
  let hash a =
100
101
102
    let accu = Chars.hash a.chars in
    let accu = 17 * accu + Intervals.hash a.ints in
    let accu = 17 * accu + Atoms.hash a.atoms in
103
104
105
106
107
108
    let accu = 17 * accu + BoolPair.hash a.times in
    let accu = 17 * accu + BoolPair.hash a.xml in
    let accu = 17 * accu + BoolPair.hash a.arrow in
    let accu = 17 * accu + BoolRec.hash a.record in
    let accu = if a.absent then accu+5 else accu in
    accu
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132

  let serialize t a =
    Chars.serialize t a.chars;
    Intervals.serialize t a.ints;
    Atoms.serialize t a.atoms;
    BoolPair.serialize t a.times;
    BoolPair.serialize t a.xml;
    BoolPair.serialize t a.arrow;
    BoolRec.serialize t a.record;
    Serialize.Put.bool t a.absent 

  let deserialize t =
    let chars = Chars.deserialize t in
    let ints = Intervals.deserialize t in
    let atoms = Atoms.deserialize t in
    let times = BoolPair.deserialize t in
    let xml = BoolPair.deserialize t in
    let arrow = BoolPair.deserialize t in
    let record = BoolRec.deserialize t in
    let absent = Serialize.Get.bool t in
    { chars = chars; ints = ints; atoms = atoms; times = times; xml = xml;
      arrow = arrow; record = record; absent = absent }
   
    
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
end
and Node :
sig
  type t = { id : int; mutable descr : Descr.t }
  val dump: Format.formatter -> t -> unit
  val check: t -> unit
  val equal: t -> t -> bool
  val hash: t -> int
  val compare:t -> t -> int
  val serialize: t Serialize.Put.f
  val deserialize: t Serialize.Get.f
end =
struct
  type t = { id : int; mutable descr : Descr.t }
  include Custom.Dummy
  let hash x = x.id
  let compare x y = Pervasives.compare x.id y.id
  let equal x y = x == y
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167

  let buffer = Serialize.Put.mk_property (fun t -> ref [])
  let serialize t n = 
    let l = Serialize.Put.get_property buffer t in
    Printf.eprintf "Node.serialize %i\n" n.id;
    Serialize.Put.int t n.id;
    if not (List.memq n.id !l) then
      (
	Printf.eprintf "Recurs\n";
	l := n.id :: !l;
	Descr.serialize t n.descr
      )

  let deserialize t = 
    let id = Serialize.Get.int t in
    Printf.eprintf "Node.deserialize %i\n" id;
    failwith "deserialize"
168
169
end

170
171
(* It is also possible to use Boolean insteand of Bool here;
   need to analyze when each one is more efficient *)
172
173
174
175
176
177
and BoolPair : Bool.S with type elem = Node.t * Node.t = 
Bool.Make(Custom.Pair(Node)(Node))

and BoolRec : Bool.S with type elem = bool * Node.t label_map =
Bool.Make(Custom.Pair(Custom.Bool)(LabelSet.MakeMap(Node)))

178

179
180
181
type descr = Descr.t
type node = Node.t
include Descr
182
	       
183
let empty = { 
184
185
186
  times = BoolPair.empty; 
  xml   = BoolPair.empty; 
  arrow = BoolPair.empty; 
187
  record= BoolRec.empty;
188
189
190
  ints  = Intervals.empty;
  atoms = Atoms.empty;
  chars = Chars.empty;
191
  absent= false;
192
193
194
}
	      
let any =  {
195
196
197
  times = BoolPair.full; 
  xml   = BoolPair.full; 
  arrow = BoolPair.full; 
198
  record= BoolRec.full; 
199
200
201
  ints  = Intervals.any;
  atoms = Atoms.any;
  chars = Chars.any;
202
  absent= false;
203
}
204
205
206
207

let non_constructed =
  { any with times = empty.times; xml = empty.xml; record = empty.record }
     
208
209
	     
let interval i = { empty with ints = i }
210
211
212
let times x y = { empty with times = BoolPair.atom (x,y) }
let xml x y = { empty with xml = BoolPair.atom (x,y) }
let arrow x y = { empty with arrow = BoolPair.atom (x,y) }
213
let record label t = 
214
215
216
  { empty with record = BoolRec.atom (true,LabelMap.singleton label t) }
let record' (x : bool * node Ident.label_map) =
  { empty with record = BoolRec.atom x }
217
218
219
220
221
222
let atom a = { empty with atoms = a }
let char c = { empty with chars = c }
let constant = function
  | Integer i -> interval (Intervals.atom i)
  | Atom a -> atom (Atoms.atom a)
  | Char c -> char (Chars.atom c)
223
      
224
225
let cup x y = 
  if x == y then x else {
226
227
228
    times = BoolPair.cup x.times y.times;
    xml   = BoolPair.cup x.xml y.xml;
    arrow = BoolPair.cup x.arrow y.arrow;
229
    record= BoolRec.cup x.record y.record;
230
231
232
    ints  = Intervals.cup x.ints  y.ints;
    atoms = Atoms.cup x.atoms y.atoms;
    chars = Chars.cup x.chars y.chars;
233
    absent= x.absent || y.absent;
234
235
236
237
  }
    
let cap x y = 
  if x == y then x else {
238
239
    times = BoolPair.cap x.times y.times;
    xml   = BoolPair.cap x.xml y.xml;
240
    record= BoolRec.cap x.record y.record;
241
    arrow = BoolPair.cap x.arrow y.arrow;
242
243
244
    ints  = Intervals.cap x.ints  y.ints;
    atoms = Atoms.cap x.atoms y.atoms;
    chars = Chars.cap x.chars y.chars;
245
    absent= x.absent && y.absent;
246
247
248
249
  }
    
let diff x y = 
  if x == y then empty else {
250
251
252
    times = BoolPair.diff x.times y.times;
    xml   = BoolPair.diff x.xml y.xml;
    arrow = BoolPair.diff x.arrow y.arrow;
253
    record= BoolRec.diff x.record y.record;
254
255
256
    ints  = Intervals.diff x.ints  y.ints;
    atoms = Atoms.diff x.atoms y.atoms;
    chars = Chars.diff x.chars y.chars;
257
    absent= x.absent && not y.absent;
258
259
  }
    
260

261

262

263
264
265
266
267
268
269
270
(* TODO: optimize disjoint check for boolean combinations *)
let trivially_disjoint a b =
  (Chars.disjoint a.chars b.chars) &&
  (Intervals.disjoint a.ints b.ints) &&
  (Atoms.disjoint a.atoms b.atoms) &&
  (BoolPair.trivially_disjoint a.times b.times) &&
  (BoolPair.trivially_disjoint a.xml b.xml) &&
  (BoolPair.trivially_disjoint a.arrow b.arrow) &&
271
272
  (BoolRec.trivially_disjoint a.record b.record) &&
  (not (a.absent && b.absent))
273

274

275
276
module DescrHash = Hashtbl.Make(Descr)
module DescrMap = Map.Make(Descr)
277
module DescrSet = Set.Make(Descr)
278
module DescrSList = SortedList.Make(Descr)
279

280
(* let hash_cons = DescrHash.create 17000 *)
281

282
let count = State.ref "Types.count" 0
283
let make () = incr count; { Node.id = !count; Node.descr = empty }
284
285
let define n d = 
(*  DescrHash.add hash_cons d n; *)
286
  n.Node.descr <- d
287
let cons d = 
288
(*   try DescrHash.find hash_cons d with Not_found ->
289
  incr count; let n = { id = !count; descr = d } in
290
  DescrHash.add hash_cons d n; n  *)
291
292
  incr count; { Node.id = !count; Node.descr = d }
let descr n = n.Node.descr
293
let internalize n = n
294
let id n = n.Node.id
295
296
297
298




299
300
let neg x = diff any x

301
302
let any_node = cons any

303
module LabelS = Set.Make(LabelPool)
304
305
306

let get_record r =
  let labs accu (_,r) = 
307
308
    List.fold_left 
      (fun accu (l,_) -> LabelS.add l accu) accu (LabelMap.get r) in
309
  let extend descrs labs (o,r) =
310
311
312
313
314
    let rec aux i labs r =
      match labs with
	| [] -> ()
	| l1::labs ->
	    match r with
315
	      | (l2,x)::r when l1 == l2 -> 
316
317
318
		  descrs.(i) <- cap descrs.(i) (descr x);
		  aux (i+1) labs r
	      | r ->
319
		  if not o then descrs.(i) <- 
320
		    cap descrs.(i) { empty with absent = true }; (* TODO:OPT *)
321
322
		  aux (i+1) labs r
    in
323
    aux 0 labs (LabelMap.get r);
324
325
326
327
    o
  in
  let line (p,n) =
    let labels = 
328
329
      List.fold_left labs (List.fold_left labs LabelS.empty p) n in
    let labels = LabelS.elements labels in
330
    let nlab = List.length labels in
331
    let mk () = Array.create nlab { any with absent = true } in
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346

    let pos = mk () in
    let opos = List.fold_left 
		 (fun accu x -> 
		    (extend pos labels x) && accu)
		 true p in
    let p = (opos, pos) in

    let n = List.map (fun x ->
			let neg = mk () in
			let o = extend neg labels x in
			(o,neg)
		     ) n in
    (labels,p,n)
  in
347
  List.map line (BoolRec.get r)
348
   
349

350

351
352
353
354
355
356
357


(* Subtyping algorithm *)

let diff_t d t = diff d (descr t)
let cap_t d t = cap d (descr t)
let cup_t d t = cup d (descr t)
358
let cap_product any_left any_right l =
359
360
  List.fold_left 
    (fun (d1,d2) (t1,t2) -> (cap_t d1 t1, cap_t d2 t2))
361
    (any_left,any_right)
362
    l
363
364
let any_pair = { empty with times = any.times }

365

366
367
368
let rec exists max f =
  (max > 0) && (f (max - 1) || exists (max - 1) f)

369
exception NotEmpty
370

371
372
373
374
375
376
377
378
379
380
381
382
type slot = { mutable status : status; 
	       mutable notify : notify;
	       mutable active : bool }
and status = Empty | NEmpty | Maybe
and notify = Nothing | Do of slot * (slot -> unit) * notify

let slot_empty = { status = Empty; active = false; notify = Nothing }
let slot_not_empty = { status = NEmpty; active = false; notify = Nothing }

let rec notify = function
  | Nothing -> ()
  | Do (n,f,rem) -> 
383
      if n.status == Maybe then (try f n with NotEmpty -> ());
384
385
386
387
388
389
390
391
392
393
      notify rem

let rec iter_s s f = function
  | [] -> ()
  | arg::rem -> f arg s; iter_s s f rem


let set s =
  s.status <- NEmpty;
  notify s.notify;
394
  s.notify <- Nothing; 
395
396
397
398
399
400
401
  raise NotEmpty

let rec big_conj f l n =
  match l with
    | [] -> set n
    | [arg] -> f arg n
    | arg::rem ->
402
403
404
	let s = 
	  { status = Maybe; active = false; 
	    notify = Do (n,(big_conj f rem), Nothing) } in
405
406
407
	try 
	  f arg s;
	  if s.active then n.active <- true
408
	with NotEmpty -> if n.status == NEmpty then raise NotEmpty
409

410
411
let guard a f n =
  match a with
412
    | { status = Empty } -> ()
413
414
415
    | { status = Maybe } as s -> 
	n.active <- true; 
	s.notify <- Do (n,f,s.notify)
416
    | { status = NEmpty } -> f n
417

418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
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
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
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

(* Fast approximation *)

module ClearlyEmpty = 
struct

let memo = DescrHash.create 33000
let marks = ref [] 

let rec slot d =
  if not ((Intervals.is_empty d.ints) && 
	  (Atoms.is_empty d.atoms) &&
	  (Chars.is_empty d.chars) &&
	  (not d.absent)) then slot_not_empty 
  else try DescrHash.find memo d
  with Not_found ->
    let s = { status = Maybe; active = false; notify = Nothing } in
    DescrHash.add memo d s;
    (try
       iter_s s check_times (BoolPair.get d.times);  
       iter_s s check_xml (BoolPair.get d.xml); 
       iter_s s check_arrow (BoolPair.get d.arrow);
       iter_s s check_record (get_record d.record);
       if s.active then marks := s :: !marks else s.status <- Empty;
     with
	 NotEmpty -> ());
    s

and check_times (left,right) s =
  let (accu1,accu2) = cap_product any any left in
  let single_right (t1,t2) s =
    let t1 = descr t1 and t2 = descr t2 in
    if trivially_disjoint accu1 t1 || trivially_disjoint accu2 t2 then set s 
    else
      let accu1 = diff accu1 t1 in guard (slot accu1) set s;
      let accu2 = diff accu2 t2 in guard (slot accu2) set s in
  guard (slot accu1) (guard (slot accu2) (big_conj single_right right)) s

and check_xml (left,right) s =
  let (accu1,accu2) = cap_product any any_pair left in
  let single_right (t1,t2) s =
    let t1 = descr t1 and t2 = descr t2 in
    if trivially_disjoint accu1 t1 || trivially_disjoint accu2 t2 then set s 
    else
      let accu1 = diff accu1 t1 in guard (slot accu1) set s;
      let accu2 = diff accu2 t2 in guard (slot accu2) set s in
  guard (slot accu1) (guard (slot accu2) (big_conj single_right right)) s

and check_arrow (left,right) s =
  let single_right (s1,s2) s =
    let accu1 = descr s1 and accu2 = neg (descr s2) in
    let single_left (t1,t2) s =
      let accu1 = diff_t accu1 t1 in guard (slot accu1) set s;
      let accu2 = cap_t  accu2 t2 in guard (slot accu2) set s
    in
    guard (slot accu1) (big_conj single_left left) s
  in
  big_conj single_right right s

and check_record (labels,(oleft,left),rights) s =
  let rec single_right (oright,right) s = 
    let next =
      (oleft && (not oright)) ||
      exists (Array.length left)
	(fun i -> trivially_disjoint left.(i) right.(i))
    in
    if next then set s
    else
      for i = 0 to Array.length left - 1 do
	let di = diff left.(i) right.(i) in guard (slot di) set s
      done
  in
  let rec start i s =
    if (i < 0) then big_conj single_right rights s
    else guard (slot left.(i)) (start (i - 1)) s
  in
  start (Array.length left - 1) s


let is_empty d =
  let s = slot d in
  List.iter 
    (fun s' -> 
       if s'.status == Maybe then s'.status <- Empty; s'.notify <- Nothing) 
    !marks;
  marks := [];
  s.status == Empty
end

let clearly_disjoint t1 t2 =
(*
  if trivially_disjoint t1 t2 then true
  else
    if ClearlyEmpty.is_empty (cap t1 t2) then
      (Printf.eprintf "!\n"; true) else false
*)
  trivially_disjoint t1 t2 || ClearlyEmpty.is_empty (cap t1 t2) 

let memo = DescrHash.create 33000
let marks = ref [] 

let rec slot d =
520
521
  if not ((Intervals.is_empty d.ints) && 
	  (Atoms.is_empty d.atoms) &&
522
523
	  (Chars.is_empty d.chars) &&
	  (not d.absent)) then slot_not_empty 
524
525
526
527
528
  else try DescrHash.find memo d
  with Not_found ->
    let s = { status = Maybe; active = false; notify = Nothing } in
    DescrHash.add memo d s;
    (try
529
       iter_s s check_times (BoolPair.get d.times);  
530
       iter_s s check_xml (BoolPair.get d.xml); 
531
       iter_s s check_arrow (BoolPair.get d.arrow);
532
533
534
535
536
537
538
539
540
       iter_s s check_record (get_record d.record);
       if s.active then marks := s :: !marks else s.status <- Empty;
     with
	 NotEmpty -> ());
    s

and check_times (left,right) s =
  let rec aux accu1 accu2 right s = match right with
    | (t1,t2)::right ->
541
542
543
	let t1 = descr t1 and t2 = descr t2 in
	if trivially_disjoint accu1 t1 || 
	   trivially_disjoint accu2 t2 then (
544
545
	     aux accu1 accu2 right s )
	else (
546
          let accu1' = diff accu1 t1 in 
547
	  guard (slot accu1') (aux accu1' accu2 right) s;
548
549

          let accu2' = diff accu2 t2 in 
550
	  guard (slot accu2') (aux accu1 accu2' right) s  
551
	)
552
553
    | [] -> set s
  in
554
  let (accu1,accu2) = cap_product any any left in
555
  guard (slot accu1) (guard (slot accu2) (aux accu1 accu2 right)) s
556
557
558
559
560

and check_xml (left,right) s =
  let rec aux accu1 accu2 right s = match right with
    | (t1,t2)::right ->
	let t1 = descr t1 and t2 = descr t2 in
561
	if clearly_disjoint accu1 t1 || 
562
563
564
565
	   trivially_disjoint accu2 t2 then (
	     aux accu1 accu2 right s )
	else (
          let accu1' = diff accu1 t1 in 
566
	  guard (slot accu1') (aux accu1' accu2 right) s;
567
568

          let accu2' = diff accu2 t2 in 
569
	  guard (slot accu2') (aux accu1 accu2' right) s  
570
571
572
573
	)
    | [] -> set s
  in
  let (accu1,accu2) = cap_product any any_pair left in
574
  guard (slot accu1) (guard (slot accu2) (aux accu1 accu2 right)) s
575

576
577
578
579
and check_arrow (left,right) s =
  let single_right (s1,s2) s =
    let rec aux accu1 accu2 left s = match left with
      | (t1,t2)::left ->
580
          let accu1' = diff_t accu1 t1 in 
581
	  guard (slot accu1') (aux accu1' accu2 left) s;
582
583

          let accu2' = cap_t  accu2 t2 in 
584
	  guard (slot accu2') (aux accu1 accu2' left) s
585
586
587
      | [] -> set s
    in
    let accu1 = descr s1 in
588
    guard (slot accu1) (aux accu1 (neg (descr s2)) left) s
589
590
  in
  big_conj single_right right s
591

592
and check_record (labels,(oleft,left),rights) s =
593
594
  let rec aux rights s = match rights with
    | [] -> set s
595
    | (oright,right)::rights ->
596
	let next =
597
	  (oleft && (not oright)) ||
598
	  exists (Array.length left)
599
	    (fun i -> trivially_disjoint left.(i) right.(i))
600
601
602
603
604
605
	in
	if next then aux rights s
	else
	  for i = 0 to Array.length left - 1 do
	    let back = left.(i) in
	    let di = diff back right.(i) in
606
607
	    guard (slot di) (fun s ->
			left.(i) <- di;
608
609
610
			aux rights s;
			left.(i) <- back;
		     ) s
611
(* TODO: are side effects correct ? *)
612
613
614
615
616
	  done
  in
  let rec start i s =
    if (i < 0) then aux rights s
    else
617
      guard (slot left.(i)) (start (i - 1)) s
618
619
620
621
622
623
624
  in
  start (Array.length left - 1) s


let is_empty d =
  let s = slot d in
  List.iter 
625
626
    (fun s' -> 
       if s'.status == Maybe then s'.status <- Empty; s'.notify <- Nothing) 
627
628
    !marks;
  marks := [];
629
  s.status == Empty
630

631
(*
632
let is_empty d =
633
634
635
636
637
638
639
(*  let b1 = ClearlyEmpty.is_empty d in
  let b2 = is_empty d in
  assert (b2 || not b1);
  Printf.eprintf "b1 = %b; b2 = %b\n" b1 b2;
  b2  *)
  if ClearlyEmpty.is_empty d then (Printf.eprintf "!\n"; true) else is_empty d
*)  
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659

let non_empty d = 
  not (is_empty d)

let subtype d1 d2 =
  is_empty (diff d1 d2)

module Product =
struct
  type t = (descr * descr) list

  let other ?(kind=`Normal) d = 
    match kind with
      | `Normal -> { d with times = empty.times }
      | `XML -> { d with xml = empty.xml }

  let is_product ?kind d = is_empty (other ?kind d)

  let need_second = function _::_::_ -> true | _ -> false

660
661
662
663
  let normal_aux = function
    | ([] | [ _ ]) as d -> d
    | d ->

664
665
666
667
668
669
670
    let res = ref [] in

    let add (t1,t2) =
      let rec loop t1 t2 = function
	| [] -> res := (ref (t1,t2)) :: !res
	| ({contents = (d1,d2)} as r)::l ->
	    (*OPT*) 
671
(*	    if equal_descr d1 t1 then r := (d1,cup d2 t2) else*)
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
	      
	      let i = cap t1 d1 in
	      if is_empty i then loop t1 t2 l
	      else (
		r := (i, cup t2 d2);
		let k = diff d1 t1 in 
		if non_empty k then res := (ref (k,d2)) :: !res;
		
		let j = diff t1 d1 in 
		if non_empty j then loop j t2 l
	      )
      in
      loop t1 t2 !res
    in
    List.iter add d;
    List.map (!) !res


(* Partitioning:

(t,s) - ((t1,s1) | (t2,s2) | ... | (tn,sn))
=
(t & t1, s - s1) | ... | (t & tn, s - sn) | (t - (t1|...|tn), s)

696
*)
697
  let get_aux any_right d =
698
699
    let accu = ref [] in
    let line (left,right) =
700
      let (d1,d2) = cap_product any any_right left in
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
      if (non_empty d1) && (non_empty d2) then
	let right = List.map (fun (t1,t2) -> descr t1, descr t2) right in
	let right = normal_aux right in
	let resid1 = ref d1 in
	let () = 
	  List.iter
	    (fun (t1,t2) ->
	       let t1 = cap d1 t1 in
	       if (non_empty t1) then
		 let () = resid1 := diff !resid1 t1 in
		 let t2 = diff d2 t2 in
		 if (non_empty t2) then accu := (t1,t2) :: !accu
	    ) right in
	if non_empty !resid1 then accu := (!resid1, d2) :: !accu 
    in
716
    List.iter line (BoolPair.get d);
717
    !accu
718
719
720
(* Maybe, can improve this function with:
     (t,s) \ (t1,s1) = (t&t',s\s') | (t\t',s),
   don't call normal_aux *)
721

722

723
724
  let get ?(kind=`Normal) d = 
    match kind with
725
726
      | `Normal -> get_aux any d.times
      | `XML -> get_aux any_pair d.xml
727
728
729

  let pi1 = List.fold_left (fun acc (t1,_) -> cup acc t1) empty
  let pi2 = List.fold_left (fun acc (_,t2) -> cup acc t2) empty
730
731
732
733
  let pi2_restricted restr = 
    List.fold_left (fun acc (t1,t2) -> 
		      if is_empty (cap t1 restr) then acc
		      else cup acc t2) empty
734
735

  let restrict_1 rects pi1 =
736
737
    let aux acc (t1,t2) = 
      let t1 = cap t1 pi1 in if is_empty t1 then acc else (t1,t2)::acc in
738
739
740
741
    List.fold_left aux [] rects
  
  type normal = t

742
  module Memo = Map.Make(BoolPair)
743

744
745
  (* TODO: try with an hashtable *)
  (* Also, avoid lookup for simple products (t1,t2) *)
746
  let memo = ref Memo.empty
747
  let normal_times d = 
748
749
750
    try Memo.find d !memo 
    with
	Not_found ->
751
	  let gd = get_aux any d in
752
	  let n = normal_aux gd in
753
754
(* Could optimize this call to normal_aux because one already
   know that each line is normalized ... *)
755
756
	  memo := Memo.add d n !memo;
	  n
757

758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
  let memo_xml = ref Memo.empty
  let normal_xml d = 
    try Memo.find d !memo_xml
    with
	Not_found ->
	  let gd = get_aux any_pair d in
	  let n = normal_aux gd in
	  memo_xml := Memo.add d n !memo_xml;
	  n

  let normal ?(kind=`Normal) d =
    match kind with 
      | `Normal -> normal_times d.times 
      | `XML -> normal_xml d.xml


774
775
776
777
778
779
780
781
782
783
  let merge_same_2 r =
    let r = 
      List.fold_left 
	(fun accu (t1,t2) ->
	   let t = try DescrMap.find t2 accu with Not_found -> empty in
	   DescrMap.add t2 (cup t t1) accu
	) DescrMap.empty r in
    DescrMap.fold (fun t2 t1 accu -> (t1,t2)::accu) r []
	 

784
785
786
787
788
789
790
  let constraint_on_2 n t1 =
    List.fold_left 
      (fun accu (d1,d2) ->
	 if is_empty (cap d1 t1) then accu else cap accu d2)
      any
      n

791
792
  let any = { empty with times = any.times }
  and any_xml = { empty with xml = any.xml }
793
  let is_empty d = d == []
794
end
795

796
module Record = 
797
struct
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
  let has_record d = not (is_empty { empty with record = d.record })
  let or_absent d = { d with absent = true }
  let any_or_absent = or_absent any
  let has_absent d = d.absent

  let only_absent = {empty with absent = true}
  let only_absent_node = cons only_absent

  module T = struct
    type t = descr
    let any = any_or_absent
    let cap = cap
    let cup = cup
    let diff = diff
    let is_empty = is_empty
    let empty = empty
  end
  module R = struct
    type t = descr
    let any = { empty with record = any.record }
    let cap = cap
    let cup = cup
    let diff = diff
    let is_empty = is_empty
    let empty = empty
  end
  module TR = Normal.Make(T)(R)

  let any_record = { empty with record = BoolRec.full }

  let atom o l = 
    if o && LabelMap.is_empty l then any_record else
    { empty with record = BoolRec.atom (o,l) }

  type zor = Pair of descr * descr | Any

  let aux_split d l=
    let f (o,r) =
      try
	let (lt,rem) = LabelMap.assoc_remove l r in
	Pair (descr lt, atom o rem)
      with Not_found -> 
	if o then
	  if LabelMap.is_empty r then Any else
	    Pair (any_or_absent, { empty with record = BoolRec.atom (o,r) })
	else
	  Pair (only_absent,
		{ empty with record = BoolRec.atom (o,r) })
    in
    List.fold_left 
      (fun b (p,n) ->
	 let rec aux_p accu = function
	   | x::p -> 
	       (match f x with
		  | Pair (t1,t2) -> aux_p ((t1,t2)::accu) p
		  | Any -> aux_p accu p)
	   | [] -> aux_n accu [] n
	 and aux_n p accu = function
	   | x::n -> 
	       (match f x with
		  | Pair (t1,t2) -> aux_n p ((t1,t2)::accu) n
		  | Any -> b)
	   | [] -> (p,accu) :: b in
	 aux_p [] p)
      []
      (BoolRec.get d.record)

  let split (d : descr) l =
    TR.boolean (aux_split d l)

  let split_normal d l =
    TR.boolean_normal (aux_split d l)


  let project d l =
    let t = TR.pi1 (split d l) in
    if t.absent then raise Not_found;
    t

  let project_opt d l =
    let t = TR.pi1 (split d l) in
    { t with absent = false }

  let condition d l t =
    TR.pi2_restricted t (split d l)
883

884
885
886
887
888
(* TODO: eliminate this cap ... (reord l only_absent_node) when
   not necessary. eg. {| ..... |} \ l *)

  let remove_field d l = 
    cap (TR.pi2 (split d l)) (record l only_absent_node)
889

890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
  let first_label d =
    let min = ref LabelPool.dummy_max in
    let aux (_,r) = 
      match LabelMap.get r with
	  (l,_)::_ -> if (l:int) < !min then min := l | _ -> () in
    BoolRec.iter aux d.record;
    !min

  let empty_cases d =
    let x = BoolRec.compute
	      ~empty:0 ~full:3 ~cup:(lor) ~cap:(land)
	      ~diff:(fun a b -> a land lnot b)
	      ~atom:(function (o,r) ->
		       assert (LabelMap.get r == []);
		       if o then 3 else 1
		    )
	      d.record in
    (x land 2 <> 0, x land 1 <> 0)

  let has_empty_record d =
    BoolRec.compute
      ~empty:false ~full:true ~cup:(||) ~cap:(&&)
      ~diff:(fun a b -> a && not b)
      ~atom:(function (o,r) ->
	       List.for_all 
	         (fun (l,t) -> (descr t).absent)
	         (LabelMap.get r)
	    )
      d.record
    

(*TODO: optimize merge
   - pre-compute the sequence of labels
   - remove empty or full { l = t }
*)

  let merge d1 d2 = 
    let res = ref empty in
    let rec aux accu d1 d2 =
      let l = min (first_label d1) (first_label d2) in
      if l = LabelPool.dummy_max then
	let (some1,none1) = empty_cases d1 
	and (some2,none2) = empty_cases d2 in
	let none = none1 && none2 and some = some1 || some2 in
	let accu = LabelMap.from_list (fun _ _ -> assert false) accu in
	(* approx for the case (some && not none) ... *)
	res := cup !res (record' (some, accu))
      else
	let l1 = split d1 l and l2 = split d2 l in
	let loop (t1,d1) (t2,d2) =
	  let t = 
	    if t2.absent 
	    then cup t1 { t2 with absent = false } 
	    else t2 
	  in
	  aux ((l,cons t)::accu) d1 d2
	in
	List.iter (fun x -> List.iter (loop x) l2) l1
	  
    in
    aux [] d1 d2;
    !res

  let any = { empty with record = any.record }

  let get d =
    let rec aux r accu d =
      let l = first_label d in
      if l == LabelPool.dummy_max then
	let (o1,o2) = empty_cases d in 
	if o1 || o2 then (LabelMap.from_list_disj r,o1,o2)::accu else accu
      else
	List.fold_left 
963
964
965
	  (fun accu (t1,t2) -> 
	     let x = (t1.absent, { t1 with absent = false }) in
	     aux ((l,x)::r) accu t2)
966
967
968
969
970
971
972
973
974
	  accu
	  (split d l)
    in
    aux [] [] d
end


module Print = 
struct
975
  let print_const ppf = function
976
977
978
    | Integer i -> Intervals.V.print ppf i
    | Atom a -> Atoms.V.print_quote ppf a
    | Char c -> Chars.V.print ppf c
979

980
  let nil_atom = Atoms.V.mk_ascii "nil"
981
982
983
984
985
986
987
988
989
  let nil_type = atom (Atoms.atom nil_atom)
  let (seqs_node,seqs_descr) = 
    let n = make () in
    let d = cup nil_type (times any_node n) in
    define n d;
    (n, d)

  let is_regexp t = subtype t seqs_descr

990
991
992
  module S = struct
  type t = { id : int; 
	     mutable def : d list; 
993
	     mutable state : [ `Expand | `None | `Marked | `Named of U.t ] }
994
  and  d =
995
    | Name of U.t
996
997
998
    | Regexp of t Pretty.regexp
    | Atomic of (Format.formatter -> unit)
    | Pair of t * t
999
    | Char of Chars.V.t
1000
    | Xml of [ `Tag of (Format.formatter -> unit) | `Type of t ] * t * t
1001
1002
    | Record of (bool * t) label_map * bool * bool
    | Arrows of (t * t) list * (t * t) list
1003
    | Neg of t
1004
1005
1006
1007
  let compare x y = x.id - y.id
  end
  module Decompile = Pretty.Decompile(DescrHash)(S)
  open S
1008

1009
  module DescrPairMap = Map.Make(Custom.Pair(Descr)(Descr))
1010
1011
1012

  let named = State.ref "Types.Print.named" DescrMap.empty
  let named_xml = State.ref "Types.Print.named_xml"  DescrPairMap.empty
1013
  let register_global (name : U.t) d = 
1014
    if equal { d with xml = BoolPair.empty } empty then 
1015
1016
1017
1018
      (let l = (*Product.merge_same_2*) (Product.get ~kind:`XML d) in
      match l with
	| [(t1,t2)] -> named_xml := DescrPairMap.add (t1,t2) name !named_xml
	| _ -> ());
1019
    named := DescrMap.add d name !named
1020

1021
  let memo = DescrHash.create 63
1022
1023
  let counter = ref 0
  let alloc def = { id = (incr counter; !counter); def = def; state = `None }
1024

1025
1026
1027
  let count_name = ref 0
  let name () =
    incr count_name;
1028
    U.mk ("X" ^ (string_of_int !count_name))
1029

1030
1031
  let to_print = ref []

1032
1033
1034
  let trivial_rec b = 
    b == BoolRec.empty || 
    (is_empty { empty with record = BoolRec.diff BoolRec.full b})
1035

1036
  let trivial_pair b = b == BoolPair.empty || b == BoolPair.full
1037
1038

  let worth_abbrev d = 
1039
1040
    not (trivial_pair d.times && trivial_pair d.xml && 
	 trivial_pair d.arrow && trivial_rec d.record) 
1041

1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
  let worth_complement d =
    let aux f x y = if f x y = 0 then 1 else 0 in
    let n = 
      aux Atoms.compare d.atoms any.atoms +
      aux Chars.compare d.chars any.chars +
      aux Intervals.compare d.ints any.ints +
      aux BoolPair.compare d.times any.times +
      aux BoolPair.compare d.xml any.xml +
      aux BoolPair.compare d.arrow any.arrow +
      aux BoolRec.compare d.record any.record in
    n >= 4

1054
  let rec prepare d =
1055
    try DescrHash.find memo d
1056
    with Not_found ->
1057
1058
      try 
	let n = DescrMap.find d !named in
1059
1060
	let s = alloc [] in
	s.state <- `Named n;
1061
1062
1063
	DescrHash.add memo d s;
	s
      with Not_found ->
1064
	if worth_complement d then 
1065
	  alloc [Neg (prepare (neg d))]
1066
	else
1067
1068
1069
	let slot = alloc [] in
	if not (worth_abbrev d) then slot.state <- `Expand;
	DescrHash.add memo d slot;
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
	let (seq,not_seq) =
	  if (subtype { empty with times = d.times } seqs_descr) then
	    (cap d seqs_descr, diff d seqs_descr)
	  else
	    (empty, d) in

	let add u = slot.def <- u :: slot.def in
	if (non_empty seq) then
	  add (Regexp (decompile seq));  
	List.iter
	  (fun (t1,t2) -> add (Pair (prepare t1, prepare t2)))
	  (Product.get not_seq);
	List.iter
	  (fun (t1,t2) ->
	     try 
	       let n = DescrPairMap.find (t1,t2) !named_xml in
	       add (Name n)
	     with
		 Not_found ->
	     let tag = 
1090
	       match Atoms.print_tag t1.atoms with
1091
		 | Some a when is_empty { t1 with atoms = Atoms.empty } -> `Tag a
1092
		 | _ -> `Type (prepare t1) in
1093
	     assert (equal { t2 with times = empty.times } empty);
1094
1095
1096
	     List.iter
	       (fun (ta,tb) -> add (Xml (tag, prepare ta, prepare tb)))
	       (Product.get t2)
1097
	  )
1098
1099
1100
	  ((*Product.merge_same_2*) (Product.get ~kind:`XML not_seq));
	List.iter
	  (fun (r,some,none) -> 
1101
	     let r = LabelMap.map (fun (o,t) -> (o, prepare t)) r in
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
	     add (Record (r,some,none)))
	  (Record.get not_seq);
	(match Chars.is_char not_seq.chars with
	  | Some c -> add (Char c)
	  | None ->
	      List.iter (fun x -> add (Atomic x)) (Chars.print not_seq.chars));
	List.iter (fun x -> add (Atomic x)) (Intervals.print not_seq.ints);
	List.iter (fun x -> add (Atomic x)) (Atoms.print not_seq.atoms);
	List.iter
	  (fun (p,n) ->
	     let aux (t,s) = prepare (descr t), prepare (descr s) in
	     let p = List.map aux p and n = List.map aux n in
	     add (Arrows (p,n)))
	  (BoolPair.get not_seq.arrow);
	slot.def <- List.rev slot.def;
	slot
	

  and decompile d =
    Decompile.decompile 
      (fun t -> 
	 let tr = Product.get t in
	 let tr = List.map (fun (l,t) -> prepare l, t) tr in
	 tr, Atoms.contains nil_atom t.atoms)
      d

1128
1129
  let gen = ref 0

1130
  let rec assign_name s =
1131
    incr gen;
1132
    match s.state with
1133
1134
1135
1136
1137
      | `None ->  
	  let g = !gen in
	  s.state <- `Marked; 
	  List.iter assign_name_rec s.def;
	  if (s.state == `Marked) && (!gen == g) then s.state <- `None
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
      | `Marked -> s.state <- `Named (name ()); to_print := s :: !to_print
      | _ -> ()
  and assign_name_rec = function
    | Neg t -> assign_name t
    | Name _ | Char _ | Atomic _ -> ()
    | Regexp r -> assign_name_regexp r
    | Pair (t1,t2) -> assign_name t1; assign_name t2
    | Xml (tag,t2,t3) -> 
	(match tag with `Type t -> assign_name t | _ -> ());
	assign_name t2;
	assign_name t3
    | Record (r,_,_) ->
	List.iter (fun (_,(_,t)) -> assign_name t) (LabelMap.get r)
    | Arrows (p,n) ->
	List.iter (fun (t1,t2) -> assign_name t1; assign_name t2) p;
	List.iter (fun (t1,t2) -> assign_name t1; assign_name t2) n
  and assign_name_regexp = function
    | Pretty.Epsilon | Pretty.Empty -> ()
    | Pretty.Alt (r1,r2) 
    | Pretty.Seq (r1,r2) -> assign_name_regexp r1; assign_name_regexp r2
    | Pretty.Star r | Pretty.Plus r -> assign_name_regexp r
    | Pretty.Trans t -> assign_name t

1161
  let rec do_print_slot pri ppf s =
1162
    match s.state with
1163
      | `Named n -> Format.fprintf ppf "%a" U.print n
1164
      | _ -> do_print_slot_real pri ppf s.def
1165
1166
1167
1168
1169
  and do_print_slot_real pri ppf def =
    let rec aux ppf = function
      | [] -> Format.fprintf ppf "Empty"
      | [ h ] -> do_print ppf h
      | h :: t -> Format.fprintf ppf "%a |@ %a" do_print h aux t
1170
    in
1171
1172
1173
1174
    if (pri >= 2) && (List.length def >= 2) 
    then Format.fprintf ppf "@[(%a)@]" aux def
    else aux ppf def
  and do_print ppf = function
1175
    | Neg t -> Format.fprintf ppf "Any \\ (@[%a@])" (do_print_slot 0) t
1176
    | Name n -> Format.fprintf ppf "%a" U.print n
1177
    | Char c -> Chars.V.print ppf c
1178
1179
1180
1181
1182
1183
    | Regexp r -> Format.fprintf ppf "@[[ %a ]@]" (do_print_regexp 0) r
    | Atomic a -> a ppf
    | Pair (t1,t2) -> 
	Format.fprintf ppf "@[(%a,%a)@]" 
	  (do_print_slot 0) t1 
	  (do_print_slot 0) t2
1184
    | Xml (tag,attr,t) -> 
1185
1186
	Format.fprintf ppf "<%a%a>%a" 
	  do_print_tag tag
1187
1188
	  do_print_attr attr
	  (do_print_slot 0) t
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
    | Record (r,some,none) ->
	if some then Format.fprintf ppf "@[{"
	else Format.fprintf ppf "@[{|";
	do_print_record ppf r;
	if not none then  Format.fprintf ppf ";@ ...";
	if some then Format.fprintf ppf " }@]"
	else Format.fprintf ppf " |}@]"
    | Arrows (p,n) ->
	(match p with
	   | [] -> Format.fprintf ppf "Arrow"
	   | (t,s)::l ->
	       Format.fprintf ppf "%a" do_print_arrow (t,s);
	       List.iter 
		 (fun (t,s) ->
		    Format.fprintf ppf " &@ %a" do_print_arrow (t,s)
		 ) l);
	List.iter 
	  (fun (t,s) ->
	     Format.fprintf ppf " \\@ %a" do_print_arrow (t,s)
	  ) n
  and do_print_arrow ppf (t,s) =
    Format.fprintf ppf "%a -> %a"
      (do_print_slot 0) t
      (do_print_slot 0) s
  and do_print_tag ppf = function
1214
    | `Tag s -> s ppf
1215
1216
    | `Type t -> Format.fprintf ppf "(%a)" (do_print_slot 0) t
  and do_print_attr ppf = function
1217
1218
    | { state = `Marked|`Expand; 
	def = [ Record (r,true,true) ] } -> do_print_record ppf r
Pietro Abate's avatar