8918aa0bf20c78733b533eb1a91e8b9e9b067e72
1 # This file is part of NIT (http://www.nitlanguage.org).
3 # Licensed under the Apache License, Version 2.0 (the "License");
4 # you may not use this file except in compliance with the License.
5 # You may obtain a copy of the License at
7 # http://www.apache.org/licenses/LICENSE-2.0
9 # Unless required by applicable law or agreed to in writing, software
10 # distributed under the License is distributed on an "AS IS" BASIS,
11 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 # See the License for the specific language governing permissions and
13 # limitations under the License.
15 # Tree-based representation of a String.
17 # Ropes are a data structure introduced in a 1995 paper from
18 # Hans J. Boehm, Russ Atkinson and Michael Plass.
19 # See : `Ropes : an Alternative to Strings`, `Software - Practice and Experience,
20 # Vol. 25(12), 1315-1330 (December 1995)`.
22 # The implementation developed here provides an automatic change
23 # of data structure depending on the length of the leaves.
25 # The length from which a `Rope` is built from a `flat` string
26 # if defined at top-level (see `maxlen`) and can be redefined by clients
27 # depending on their needs (e.g. if you need to bench the speed of
28 # the creation of concat nodes, lower the size of maxlen).
30 # A Rope as defined in the original paper is a Tree made of concatenation
31 # nodes and containing `Flat` (that is Array-based) strings as Leaves.
39 # `"My" " Name" " is" " Simon." `
41 # Note that the above example is not representative of the actual implementation
42 # of `Ropes`, since short leaves are merged to keep the rope at an acceptable
43 # height (hence, this rope here might actually be a `FlatString` !).
48 # Maxlen is the maximum length of a Leaf node
50 # When concatenating two leaves, if `new_length` > `maxlen`,
51 # A `Concat` node is created instead
53 # Its purpose is to limit the depth of the `Rope` (this
54 # improves performance when accessing/iterating).
55 fun maxlen
: Int do return 64
57 # String using a tree-based representation with leaves as `FlatStrings`
58 private abstract class Rope
62 private abstract class RopeString
66 redef fun chars
is cached
do return new RopeChars(self)
69 # Node that represents a concatenation between two `String`
75 redef fun substrings
do return new RopeSubstrings(self)
77 redef fun empty
do return ""
79 redef fun to_cstring
is cached
do
81 var ns
= new NativeString(len
+ 1)
84 for i
in substrings
do
86 i
.as(FlatString).items
.copy_to
(ns
, ilen
, i
.as(FlatString).index_from
, off
)
92 # Left child of the node
94 # Right child of the node
97 init(l
: String, r
: String) is old_style_init
do
100 length
= l
.length
+ r
.length
108 redef fun iterator
do return new RopeIter(self)
112 for j
in [1 .. i
[ do x
+= self
117 var llen
= left
.length
118 if i
>= llen
then return right
[i
- llen
]
122 redef fun substring
(from
, len
) do
123 var llen
= left
.length
125 if from
+ len
< llen
then return left
.substring
(from
,len
)
126 var lsublen
= llen
- from
127 return left
.substring_from
(from
) + right
.substring
(0, len
- lsublen
)
129 return right
.substring
(from
- llen
, len
)
133 redef fun reversed
do return new Concat(right
.reversed
, left
.reversed
)
135 redef fun insert_at
(s
, pos
) do
136 if pos
> left
.length
then
137 return left
+ right
.insert_at
(s
, pos
- left
.length
)
139 return left
.insert_at
(s
, pos
) + right
142 redef fun to_upper
do return new Concat(left
.to_upper
, right
.to_upper
)
144 redef fun to_lower
do return new Concat(left
.to_lower
, right
.to_lower
)
150 return new Concat(self, s
)
154 if rlen
+ slen
> maxlen
then return new Concat(left
, new Concat(r
, s
))
155 return new Concat(left
, r
+ s
)
160 # Mutable `Rope`, optimized for concatenation operations
162 # A `RopeBuffer` is an efficient way of building a `String` when
163 # concatenating small strings.
165 # It does concatenations in an optimized way by having a
166 # mutable part and an immutable part built by efficiently
167 # concatenating strings in chain.
169 # Every concatenation operation is done by copying a string to
170 # the mutable part and flushing it when full.
172 # However, when a long string is appended to the `Buffer`,
173 # the concatenation is done at it would be in a `Rope`.
178 redef fun chars
: Sequence[Char] is cached
do return new RopeBufferChars(self)
180 # The final string being built on the fly
181 private var str
: String is noinit
183 # Current concatenation buffer
184 private var ns
: NativeString is noinit
186 # Next available (e.g. unset) character in the `Buffer`
189 # Keeps track of the buffer's currently dumped part
191 # This might happen if for instance, a String was being
192 # built by concatenating small parts of string and suddenly
193 # a long string (length > maxlen) is appended.
194 private var dumped
: Int is noinit
196 # Length of the complete rope
199 # Length of the mutable part
201 # Is also used as base to compute the size of the next
202 # mutable native string (`ns`)
203 private var buf_size
: Int is noinit
205 redef fun substrings
: Iterator[String] do return new RopeBufSubstringIterator(self)
207 # Builds an empty `RopeBuffer`
210 ns
= new NativeString(maxlen
)
215 # Builds a new `RopeBuffer` with `str` in it.
216 init from
(str
: String) do
218 ns
= new NativeString(maxlen
)
224 # Resets the informations of the `Buffer`
226 # This is called when doing in-place modifications
227 # on a previously to_s'd `RopeBuffer`
229 var nns
= new NativeString(buf_size
)
230 var blen
= rpos
- dumped
231 ns
.copy_to
(nns
, blen
, dumped
, 0)
237 redef fun empty
do return new RopeBuffer
245 redef fun substring
(from
, count
) do
246 var strlen
= str
.length
250 if count
< 0 then count
= 0
254 if count
> length
then count
= length
- from
256 if count
== 0 then return empty
258 if from
< strlen
then
259 var subpos
= strlen
- from
260 if count
<= subpos
then
261 return new RopeBuffer.from
(str
.substring
(from
, count
))
263 var l
= str
.substring_from
(from
)
264 var rem
= count
- subpos
265 var nns
= new NativeString(rem
)
266 ns
.copy_to
(nns
, rem
, dumped
, 0)
267 return new RopeBuffer.from
(l
+ nns
.to_s_with_length
(rem
))
270 var nns
= new NativeString(count
)
271 ns
.copy_to
(nns
, count
, dumped
, 0)
272 return new RopeBuffer.from
(nns
.to_s_with_length
(count
))
276 redef fun append
(s
) do
281 if rp
> 0 and dumped
!= rp
then
282 str
+= new FlatString.with_infos
(ns
, rp
- dumped
, dumped
, rp
- 1)
288 if slen
> maxlen
then
289 if rp
> 0 and dumped
!= rp
then
290 str
+= new FlatString.with_infos
(ns
, rp
- dumped
, dumped
, rp
- 1)
296 var remsp
= buf_size
- rp
297 var sits
: NativeString
299 if s
isa FlatString then
302 else if s
isa FlatBuffer then
306 if slen
<= remsp
then
313 for i
in [0..remsp
[ do
327 if slen
<= remsp
then
328 sits
.copy_to
(ns
, slen
, begin
, rp
)
329 if rp
== buf_size
then
337 sits
.copy_to
(ns
, remsp
, begin
, rp
)
340 var nlen
= slen
- remsp
341 sits
.copy_to
(ns
, nlen
, begin
+ remsp
, 0)
351 if rp
== buf_size
then
359 # Converts the Buffer to a FlatString, appends it to
360 # the final String and re-allocates a new larger Buffer.
361 private fun dump_buffer
do
363 var nstr
= new FlatString.with_infos
(ns
, rpos
- dumped
, dumped
, rpos
- 1)
367 ns
= new NativeString(bs
)
374 new FlatString.with_infos
(ns
, rpos
- dumped
, dumped
, rpos
- 1).output
377 # Enlarge is useless here since the `Buffer`
378 # part is automatically dumped when necessary.
380 # Also, since the buffer can not be overused by a
381 # single string, there is no need for manual
384 # "You have no power here !"
385 redef fun enlarge
(i
) do end
389 var nnslen
= rpos
- dumped
390 if nnslen
== 0 then return str
391 return str
+ new FlatString.with_infos
(ns
, rpos
- dumped
, dumped
, rpos
- 1)
396 var nns
= new NativeString(buf_size
)
399 for i
in [0 .. rpos
[ do
407 if written
then reset
410 for i
in [0 .. rpos
[ do
411 mits
[i
] = mits
[i
].to_upper
416 if written
then reset
419 for i
in [0 .. rpos
[ do
420 mits
[i
] = mits
[i
].to_lower
425 redef class FlatString
427 redef fun insert_at
(s
, pos
) do
428 var l
= substring
(0, pos
)
429 var r
= substring_from
(pos
)
437 if slen
== 0 then return self
438 if mlen
== 0 then return s
439 var nlen
= slen
+ mlen
440 if s
isa FlatString then
441 if nlen
> maxlen
then return new Concat(self, s
)
443 var sifrom
= s
.index_from
444 var mifrom
= index_from
446 var ns
= new NativeString(nlen
+ 1)
447 mits
.copy_to
(ns
, mlen
, mifrom
, 0)
448 sits
.copy_to
(ns
, slen
, sifrom
, mlen
)
449 return ns
.to_s_with_length
(nlen
)
450 else if s
isa Concat then
452 var sllen
= sl
.length
453 if sllen
+ mlen
> maxlen
then return new Concat(self, s
)
454 return new Concat(self + sl
, s
.right
)
461 # A simple linked list for use with iterators
462 private class RopeIterPiece
463 # The encapsulated node of the `Rope`
465 # Was its left child (if any) visited ?
467 # Was its right child (if any) visited ?
469 # The previous node in the list.
470 var prev
: nullable RopeIterPiece
473 # A reverse iterator capable of working with `Rope` objects
474 private class RopeReviter
475 super IndexedIterator[Char]
477 # Current NativeString
479 # Current position in NativeString
481 # Position in the Rope (0-indexed)
483 # Iterator on the substrings, does the Postfix part of
484 # the Rope traversal.
485 var subs
: IndexedIterator[String]
487 init(root
: RopeString) is old_style_init
do
488 pos
= root
.length
- 1
489 subs
= new ReverseRopeSubstrings(root
)
494 init from
(root
: RopeString, pos
: Int) do
496 subs
= new ReverseRopeSubstrings.from
(root
, pos
)
498 pns
= pos
- subs
.index
501 redef fun index
do return pos
503 redef fun is_ok
do return pos
>= 0
505 redef fun item
do return ns
[pns
]
510 if pns
>= 0 then return
511 if not subs
.is_ok
then return
513 if not subs
.is_ok
then return
519 # Forward iterator on the chars of a `Rope`
520 private class RopeIter
521 super IndexedIterator[Char]
523 # Position in current `String`
525 # Current `String` being iterated on
527 # Substrings of the Rope
528 var subs
: IndexedIterator[String]
529 # Maximum position to iterate on (e.g. Rope.length)
531 # Position (char) in the Rope (0-indexed)
534 init(root
: RopeString) is old_style_init
do
535 subs
= new RopeSubstrings(root
)
538 max
= root
.length
- 1
542 init from
(root
: RopeString, pos
: Int) do
543 subs
= new RopeSubstrings.from
(root
, pos
)
544 pns
= pos
- subs
.index
547 max
= root
.length
- 1
550 redef fun item
do return str
[pns
]
552 redef fun is_ok
do return pos
<= max
554 redef fun index
do return pos
559 if pns
< subs
.item
.length
then return
560 if not subs
.is_ok
then return
562 if not subs
.is_ok
then return
568 # Substrings of a Rope (i.e. Reverse postfix iterator on leaves)
569 private class ReverseRopeSubstrings
570 super IndexedIterator[String]
573 var iter
: RopeIterPiece is noinit
575 var pos
: Int is noinit
578 var str
: String is noinit
580 init(root
: RopeString) is old_style_init
do
581 var r
= new RopeIterPiece(root
, false, true, null)
582 pos
= root
.length
- 1
583 var lnod
: String = root
585 if lnod
isa Concat then
587 r
= new RopeIterPiece(lnod
, false, true, r
)
596 init from
(root
: RopeString, pos
: Int) do
597 var r
= new RopeIterPiece(root
, false, true, null)
598 var rnod
: String = root
601 if rnod
isa Concat then
602 if off
>= rnod
.left
.length
then
603 off
-= rnod
.left
.length
605 r
= new RopeIterPiece(rnod
, false, true, r
)
609 r
= new RopeIterPiece(rnod
, false, true, r
)
621 redef fun item
do return str
623 redef fun index
do return pos
625 redef fun is_ok
do return pos
>= 0
628 if pos
< 0 then return
629 var curr
: nullable RopeIterPiece = iter
.prev
630 var currit
= curr
.node
631 while curr
!= null do
633 if not currit
isa Concat then
639 if not curr
.rdone
then
641 curr
= new RopeIterPiece(currit
.right
, false, false, curr
)
644 if not curr
.ldone
then
646 curr
= new RopeIterPiece(currit
.left
, false, false, curr
)
655 private class RopeBufSubstringIterator
656 super Iterator[String]
658 # Iterator on the substrings of the building string
659 var iter
: Iterator[String]
660 # Makes a String out of the buffered part of the Ropebuffer
662 # Did we attain the buffered part ?
663 var nsstr_done
= false
665 init(str
: RopeBuffer) is old_style_init
do
666 iter
= str
.str
.substrings
667 nsstr
= new FlatString.with_infos
(str
.ns
, str
.rpos
- str
.dumped
, str
.dumped
, str
.rpos
- 1)
668 if str
.length
== 0 then nsstr_done
= true
671 redef fun is_ok
do return iter
.is_ok
or not nsstr_done
675 if iter
.is_ok
then return iter
.item
688 # Substrings of a Rope (i.e. Postfix iterator on leaves)
689 private class RopeSubstrings
690 super IndexedIterator[String]
693 var iter
: RopeIterPiece is noinit
695 var pos
: Int is noinit
696 # Maximum position in `Rope` (i.e. length - 1)
697 var max
: Int is noinit
700 var str
: String is noinit
702 init(root
: RopeString) is old_style_init
do
703 var r
= new RopeIterPiece(root
, true, false, null)
705 max
= root
.length
- 1
706 var rnod
: String = root
708 if rnod
isa Concat then
710 r
= new RopeIterPiece(rnod
, true, false, r
)
720 init from
(root
: RopeString, pos
: Int) do
721 var r
= new RopeIterPiece(root
, true, false, null)
722 max
= root
.length
- 1
723 var rnod
: String = root
726 if rnod
isa Concat then
727 if off
>= rnod
.left
.length
then
729 off
-= rnod
.left
.length
731 r
= new RopeIterPiece(rnod
, true, false, r
)
734 r
= new RopeIterPiece(rnod
, true, false, r
)
746 redef fun item
do return str
748 redef fun is_ok
do return pos
<= max
750 redef fun index
do return pos
754 if pos
> max
then return
756 var rnod
: String = it
.node
758 if not rnod
isa Concat then
762 iter
= it
.as(not null)
768 it
= new RopeIterPiece(rnod
, false, false, it
)
769 else if not it
.rdone
then
772 it
= new RopeIterPiece(rnod
, false, false, it
)
782 # Implementation of a `StringCharView` for `RopeString` objects
783 private class RopeChars
788 init(s
: RopeString) is old_style_init
do tgt
= s
794 redef fun iterator_from
(i
) do return new RopeIter.from
(tgt
, i
)
796 redef fun reverse_iterator_from
(i
) do return new RopeReviter.from
(tgt
, i
)
801 super IndexedIterator[Char]
803 var sit
: IndexedIterator[Char]
813 init(t
: RopeBuffer) is old_style_init
do
816 sit
= t
.str
.chars
.iterator
821 init from
(t
: RopeBuffer, pos
: Int) do
824 sit
= t
.str
.chars
.iterator_from
(pos
)
825 pns
= pos
- t
.str
.length
829 redef fun is_ok
do return index
< maxpos
832 if sit
.is_ok
then return sit
.item
846 class RopeBufferReviter
847 super IndexedIterator[Char]
849 var sit
: IndexedIterator[Char]
857 init(tgt
: RopeBuffer) is old_style_init
do
858 sit
= tgt
.str
.chars
.reverse_iterator
860 index
= tgt
.length
- 1
864 init from
(tgt
: RopeBuffer, pos
: Int) do
865 sit
= tgt
.str
.chars
.reverse_iterator_from
(pos
- tgt
.rpos
- tgt
.dumped
)
866 pns
= pos
- tgt
.str
.length
871 redef fun is_ok
do return index
> 0
874 if pns
>= 0 then return ns
[pns
]
888 # View on the chars of a `RopeBuffer`
889 class RopeBufferChars
892 redef type SELFTYPE: RopeBuffer
895 if i
< target
.str
.length
then
898 return target
.ns
[i
- target
.str
.length
]
902 redef fun []=(i
,c
) do
903 if i
== target
.length
then target
.add c
904 if i
< target
.str
.length
then
906 var l
= s
.substring
(0, i
)
907 var r
= s
.substring_from
(i
+ 1)
908 target
.str
= l
+ c
.to_s
+ r
910 target
.ns
[i
- target
.str
.length
] = c
914 redef fun add
(c
) do target
.add c
916 redef fun push
(c
) do target
.add c
918 redef fun iterator_from
(i
) do return new RopeBufferIter.from
(target
, i
)
920 redef fun reverse_iterator_from
(i
) do return new RopeBufferReviter.from
(target
, i
)