1 | // |
---|
2 | // Cforall Version 1.0.0 Copyright (C) 2016 University of Waterloo |
---|
3 | // |
---|
4 | // The contents of this file are covered under the licence agreement in the |
---|
5 | // file "LICENCE" distributed with Cforall. |
---|
6 | // |
---|
7 | // string_res -- variable-length, mutable run of text, with resource semantics |
---|
8 | // |
---|
9 | // Author : Michael L. Brooks |
---|
10 | // Created On : Fri Sep 03 11:00:00 2021 |
---|
11 | // Last Modified By : Peter A. Buhr |
---|
12 | // Last Modified On : Mon Aug 14 18:06:01 2023 |
---|
13 | // Update Count : 12 |
---|
14 | // |
---|
15 | |
---|
16 | #include "string_res.hfa" |
---|
17 | #include "string_sharectx.hfa" |
---|
18 | #include "stdlib.hfa" |
---|
19 | #include <ctype.h> |
---|
20 | |
---|
21 | // Workaround for observed performance penalty from calling CFA's alloc. |
---|
22 | // Workaround is: EndVbyte = TEMP_ALLOC(char, CurrSize) |
---|
23 | // Should be: EndVbyte = alloc(CurrSize) |
---|
24 | #define TEMP_ALLOC(T, n) (( T* ) malloc( n * sizeof( T ) )) |
---|
25 | |
---|
26 | #include <assert.h> |
---|
27 | |
---|
28 | //######################### VbyteHeap "header" ######################### |
---|
29 | |
---|
30 | #ifdef VbyteDebug |
---|
31 | HandleNode *HeaderPtr; |
---|
32 | #endif // VbyteDebug |
---|
33 | |
---|
34 | struct VbyteHeap { |
---|
35 | |
---|
36 | int NoOfCompactions; // number of compactions of the byte area |
---|
37 | int NoOfExtensions; // number of extensions in the size of the byte area |
---|
38 | int NoOfReductions; // number of reductions in the size of the byte area |
---|
39 | |
---|
40 | int InitSize; // initial number of bytes in the byte-string area |
---|
41 | int CurrSize; // current number of bytes in the byte-string area |
---|
42 | char *StartVbyte; // pointer to the `st byte of the start of the byte-string area |
---|
43 | char *EndVbyte; // pointer to the next byte after the end of the currently used portion of byte-string area |
---|
44 | void *ExtVbyte; // pointer to the next byte after the end of the byte-string area |
---|
45 | |
---|
46 | HandleNode Header; // header node for handle list |
---|
47 | }; // VbyteHeap |
---|
48 | |
---|
49 | |
---|
50 | static void compaction( VbyteHeap & ); // compaction of the byte area |
---|
51 | static void garbage( VbyteHeap &, int ); // garbage collect the byte area |
---|
52 | static void extend( VbyteHeap &, int ); // extend the size of the byte area |
---|
53 | static void reduce( VbyteHeap &, int ); // reduce the size of the byte area |
---|
54 | |
---|
55 | static void ?{}( VbyteHeap &, size_t = 1000 ); |
---|
56 | static void ^?{}( VbyteHeap & ); |
---|
57 | |
---|
58 | static int ByteCmp( char *, int, int, char *, int, int ); // compare 2 blocks of bytes |
---|
59 | static char *VbyteAlloc( VbyteHeap &, int ); // allocate a block bytes in the heap |
---|
60 | static char *VbyteTryAdjustLast( VbyteHeap &, int ); |
---|
61 | |
---|
62 | static void AddThisAfter( HandleNode &, HandleNode & ); |
---|
63 | static void DeleteNode( HandleNode & ); |
---|
64 | static void MoveThisAfter( HandleNode &, const HandleNode & ); // move current handle after parameter handle |
---|
65 | |
---|
66 | |
---|
67 | // Allocate the storage for the variable sized area and intialize the heap variables. |
---|
68 | |
---|
69 | static void ?{}( VbyteHeap & this, size_t Size ) with(this) { |
---|
70 | #ifdef VbyteDebug |
---|
71 | serr | "enter:VbyteHeap::VbyteHeap, this:" | &this | " Size:" | Size; |
---|
72 | #endif // VbyteDebug |
---|
73 | NoOfCompactions = NoOfExtensions = NoOfReductions = 0; |
---|
74 | InitSize = CurrSize = Size; |
---|
75 | StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize); |
---|
76 | ExtVbyte = (void *)( StartVbyte + CurrSize ); |
---|
77 | Header.flink = Header.blink = &Header; |
---|
78 | Header.ulink = & this; |
---|
79 | #ifdef VbyteDebug |
---|
80 | HeaderPtr = &Header; |
---|
81 | serr | "exit:VbyteHeap::VbyteHeap, this:" | &this; |
---|
82 | #endif // VbyteDebug |
---|
83 | } // VbyteHeap |
---|
84 | |
---|
85 | |
---|
86 | // Release the dynamically allocated storage for the byte area. |
---|
87 | |
---|
88 | static void ^?{}( VbyteHeap & this ) with(this) { |
---|
89 | free( StartVbyte ); |
---|
90 | } // ~VbyteHeap |
---|
91 | |
---|
92 | |
---|
93 | //######################### HandleNode ######################### |
---|
94 | |
---|
95 | |
---|
96 | // Create a handle node. The handle is not linked into the handle list. This is the responsibilitiy of the handle |
---|
97 | // creator. |
---|
98 | |
---|
99 | static void ?{}( HandleNode & this ) with(this) { |
---|
100 | #ifdef VbyteDebug |
---|
101 | serr | "enter:HandleNode::HandleNode, this:" | &this; |
---|
102 | #endif // VbyteDebug |
---|
103 | s = 0; |
---|
104 | lnth = 0; |
---|
105 | #ifdef VbyteDebug |
---|
106 | serr | "exit:HandleNode::HandleNode, this:" | &this; |
---|
107 | #endif // VbyteDebug |
---|
108 | } // HandleNode |
---|
109 | |
---|
110 | // Create a handle node. The handle is linked into the handle list at the end. This means that this handle will NOT be |
---|
111 | // in order by string address, but this is not a problem because a string with length zero does nothing during garbage |
---|
112 | // collection. |
---|
113 | |
---|
114 | static void ?{}( HandleNode & this, VbyteHeap & vh ) with(this) { |
---|
115 | #ifdef VbyteDebug |
---|
116 | serr | "enter:HandleNode::HandleNode, this:" | &this; |
---|
117 | #endif // VbyteDebug |
---|
118 | s = 0; |
---|
119 | lnth = 0; |
---|
120 | ulink = &vh; |
---|
121 | AddThisAfter( this, *vh.Header.blink ); |
---|
122 | #ifdef VbyteDebug |
---|
123 | serr | "exit:HandleNode::HandleNode, this:" | &this; |
---|
124 | #endif // VbyteDebug |
---|
125 | } // HandleNode |
---|
126 | |
---|
127 | |
---|
128 | // Delete a node from the handle list by unchaining it from the list. If the handle node was allocated dynamically, it |
---|
129 | // is the responsibility of the creator to destroy it. |
---|
130 | |
---|
131 | static void ^?{}( HandleNode & this ) with(this) { |
---|
132 | #ifdef VbyteDebug |
---|
133 | serr | "enter:HandleNode::~HandleNode, this:" | & this; |
---|
134 | { |
---|
135 | serr | nlOff; |
---|
136 | serr | " lnth:" | lnth | " s:" | (void *)s | ",\""; |
---|
137 | for ( i; lnth ) { |
---|
138 | serr | s[i]; |
---|
139 | } // for |
---|
140 | serr | "\" flink:" | flink | " blink:" | blink | nl; |
---|
141 | serr | nlOn; |
---|
142 | } |
---|
143 | #endif // VbyteDebug |
---|
144 | DeleteNode( this ); |
---|
145 | } // ~HandleNode |
---|
146 | |
---|
147 | |
---|
148 | //######################### String Sharing Context ######################### |
---|
149 | |
---|
150 | static string_sharectx * ambient_string_sharectx; // fickle top of stack |
---|
151 | static string_sharectx default_string_sharectx = {NEW_SHARING}; // stable bottom of stack |
---|
152 | |
---|
153 | void ?{}( string_sharectx & this, StringSharectx_Mode mode ) with( this ) { |
---|
154 | (older){ ambient_string_sharectx }; |
---|
155 | if ( mode == NEW_SHARING ) { |
---|
156 | (activeHeap){ new( (size_t) 1000 ) }; |
---|
157 | } else { |
---|
158 | verify( mode == NO_SHARING ); |
---|
159 | (activeHeap){ 0p }; |
---|
160 | } |
---|
161 | ambient_string_sharectx = & this; |
---|
162 | } |
---|
163 | |
---|
164 | void ^?{}( string_sharectx & this ) with( this ) { |
---|
165 | if ( activeHeap ) delete( activeHeap ); |
---|
166 | |
---|
167 | // unlink this from older-list starting from ambient_string_sharectx |
---|
168 | // usually, this==ambient_string_sharectx and the loop runs zero times |
---|
169 | string_sharectx *& c = ambient_string_sharectx; |
---|
170 | while ( c != &this ) &c = &c->older; // find this |
---|
171 | c = this.older; // unlink |
---|
172 | } |
---|
173 | |
---|
174 | //######################### String Resource ######################### |
---|
175 | |
---|
176 | |
---|
177 | VbyteHeap * DEBUG_string_heap() { |
---|
178 | assert( ambient_string_sharectx->activeHeap && "No sharing context is active" ); |
---|
179 | return ambient_string_sharectx->activeHeap; |
---|
180 | } |
---|
181 | |
---|
182 | size_t DEBUG_string_bytes_avail_until_gc( VbyteHeap * heap ) { |
---|
183 | return ((char*)heap->ExtVbyte) - heap->EndVbyte; |
---|
184 | } |
---|
185 | |
---|
186 | size_t DEBUG_string_bytes_in_heap( VbyteHeap * heap ) { |
---|
187 | return heap->CurrSize; |
---|
188 | } |
---|
189 | |
---|
190 | const char * DEBUG_string_heap_start( VbyteHeap * heap ) { |
---|
191 | return heap->StartVbyte; |
---|
192 | } |
---|
193 | |
---|
194 | // Returns the size of the string in bytes |
---|
195 | size_t size(const string_res &s) with(s) { |
---|
196 | return Handle.lnth; |
---|
197 | } |
---|
198 | |
---|
199 | // Output operator |
---|
200 | ofstream & ?|?(ofstream &out, const string_res &s) { |
---|
201 | // CFA string is NOT null terminated, so print exactly lnth characters in a minimum width of 0. |
---|
202 | out | wd( 0, s.Handle.lnth, s.Handle.s ) | nonl; |
---|
203 | return out; |
---|
204 | } |
---|
205 | |
---|
206 | void ?|?(ofstream &out, const string_res &s) { |
---|
207 | (ofstream &)(out | s); ends( out ); |
---|
208 | } |
---|
209 | |
---|
210 | // Input operator |
---|
211 | ifstream & ?|?(ifstream &in, string_res &s) { |
---|
212 | |
---|
213 | // Reading into a temp before assigning to s is near zero overhead in typical cases because of sharing. |
---|
214 | // If s is a substring of something larger, simple assignment takes care of that case correctly. |
---|
215 | // But directly reading a variable amount of text into the middle of a larger context is not practical. |
---|
216 | string_res temp; |
---|
217 | |
---|
218 | // Read in chunks. Often, one chunk is enough. Keep the string that accumulates chunks last in the heap, |
---|
219 | // so available room is rest of heap. When a chunk fills the heap, force growth then take the next chunk. |
---|
220 | for (;;) { |
---|
221 | // Append dummy content to temp, forcing expansion when applicable (occurs always on subsequent loops) |
---|
222 | // length 2 ensures room for at least one real char, plus scanf/pipe-cstr's null terminator |
---|
223 | temp += "--"; |
---|
224 | assert( temp.Handle.ulink->EndVbyte == temp.Handle.s + temp.Handle.lnth ); // last in heap |
---|
225 | |
---|
226 | // reset, to overwrite the appended "--" |
---|
227 | temp.Handle.lnth -= 2; |
---|
228 | temp.Handle.ulink->EndVbyte -= 2; |
---|
229 | |
---|
230 | // rest of heap, less 1 byte for null terminator, is available to read into |
---|
231 | int lenReadable = (char*)temp.Handle.ulink->ExtVbyte - temp.Handle.ulink->EndVbyte - 1; |
---|
232 | assert (lenReadable >= 1); |
---|
233 | |
---|
234 | // get bytes |
---|
235 | in | wdi( lenReadable + 1, lenReadable, temp.Handle.ulink->EndVbyte ); |
---|
236 | int lenWasRead = strlen(temp.Handle.ulink->EndVbyte); |
---|
237 | |
---|
238 | // update metadata |
---|
239 | temp.Handle.lnth += lenWasRead; |
---|
240 | temp.Handle.ulink->EndVbyte += lenWasRead; |
---|
241 | |
---|
242 | if (lenWasRead < lenReadable) break; |
---|
243 | } |
---|
244 | |
---|
245 | s = temp; |
---|
246 | return in; |
---|
247 | } |
---|
248 | |
---|
249 | |
---|
250 | // Empty constructor |
---|
251 | void ?{}(string_res &s) with(s) { |
---|
252 | if( ambient_string_sharectx->activeHeap ) { |
---|
253 | (Handle){ * ambient_string_sharectx->activeHeap }; |
---|
254 | (shareEditSet_owns_ulink){ false }; |
---|
255 | verify( Handle.s == 0p && Handle.lnth == 0 ); |
---|
256 | } else { |
---|
257 | (Handle){ * new( (size_t) 10 ) }; // TODO: can I lazily avoid allocating for empty string |
---|
258 | (shareEditSet_owns_ulink){ true }; |
---|
259 | Handle.s = Handle.ulink->StartVbyte; |
---|
260 | verify( Handle.lnth == 0 ); |
---|
261 | } |
---|
262 | s.shareEditSet_prev = &s; |
---|
263 | s.shareEditSet_next = &s; |
---|
264 | } |
---|
265 | |
---|
266 | static void eagerCopyCtorHelper(string_res &s, const char* rhs, size_t rhslnth) with(s) { |
---|
267 | if( ambient_string_sharectx->activeHeap ) { |
---|
268 | (Handle){ * ambient_string_sharectx->activeHeap }; |
---|
269 | (shareEditSet_owns_ulink){ false }; |
---|
270 | } else { |
---|
271 | (Handle){ * new( rhslnth ) }; |
---|
272 | (shareEditSet_owns_ulink){ true }; |
---|
273 | } |
---|
274 | Handle.s = VbyteAlloc(*Handle.ulink, rhslnth); |
---|
275 | Handle.lnth = rhslnth; |
---|
276 | memmove( Handle.s, rhs, rhslnth ); |
---|
277 | s.shareEditSet_prev = &s; |
---|
278 | s.shareEditSet_next = &s; |
---|
279 | } |
---|
280 | |
---|
281 | // Constructor from a raw buffer and size |
---|
282 | void ?{}(string_res &s, const char* rhs, size_t rhslnth) with(s) { |
---|
283 | eagerCopyCtorHelper(s, rhs, rhslnth); |
---|
284 | } |
---|
285 | |
---|
286 | // private ctor (not in header): use specified heap (ignore ambient) and copy chars in |
---|
287 | void ?{}( string_res &s, VbyteHeap & heap, const char* rhs, size_t rhslnth ) with(s) { |
---|
288 | (Handle){ heap }; |
---|
289 | Handle.s = VbyteAlloc(*Handle.ulink, rhslnth); |
---|
290 | Handle.lnth = rhslnth; |
---|
291 | (s.shareEditSet_owns_ulink){ false }; |
---|
292 | memmove( Handle.s, rhs, rhslnth ); |
---|
293 | s.shareEditSet_prev = &s; |
---|
294 | s.shareEditSet_next = &s; |
---|
295 | } |
---|
296 | |
---|
297 | // General copy constructor |
---|
298 | void ?{}(string_res &s, const string_res & s2, StrResInitMode mode, size_t start, size_t end ) { |
---|
299 | |
---|
300 | verify( start <= end && end <= s2.Handle.lnth ); |
---|
301 | |
---|
302 | if (s2.Handle.ulink != ambient_string_sharectx->activeHeap && mode == COPY_VALUE) { |
---|
303 | // crossing heaps (including private): copy eagerly |
---|
304 | eagerCopyCtorHelper(s, s2.Handle.s + start, end - start); |
---|
305 | verify(s.shareEditSet_prev == &s); |
---|
306 | verify(s.shareEditSet_next == &s); |
---|
307 | } else { |
---|
308 | (s.Handle){}; |
---|
309 | s.Handle.s = s2.Handle.s + start; |
---|
310 | s.Handle.lnth = end - start; |
---|
311 | s.Handle.ulink = s2.Handle.ulink; |
---|
312 | |
---|
313 | AddThisAfter(s.Handle, s2.Handle ); // insert this handle after rhs handle |
---|
314 | // ^ bug? skip others at early point in string |
---|
315 | |
---|
316 | if (mode == COPY_VALUE) { |
---|
317 | verify(s2.Handle.ulink == ambient_string_sharectx->activeHeap); |
---|
318 | // requested logical copy in same heap: defer copy until write |
---|
319 | |
---|
320 | (s.shareEditSet_owns_ulink){ false }; |
---|
321 | |
---|
322 | // make s alone in its shareEditSet |
---|
323 | s.shareEditSet_prev = &s; |
---|
324 | s.shareEditSet_next = &s; |
---|
325 | } else { |
---|
326 | verify( mode == SHARE_EDITS ); |
---|
327 | // sharing edits with source forces same heap as source (ignore context) |
---|
328 | |
---|
329 | (s.shareEditSet_owns_ulink){ s2.shareEditSet_owns_ulink }; |
---|
330 | |
---|
331 | // s2 is logically const but not implementation const |
---|
332 | string_res & s2mod = (string_res &) s2; |
---|
333 | |
---|
334 | // insert s after s2 on shareEditSet |
---|
335 | s.shareEditSet_next = s2mod.shareEditSet_next; |
---|
336 | s.shareEditSet_prev = &s2mod; |
---|
337 | s.shareEditSet_next->shareEditSet_prev = &s; |
---|
338 | s.shareEditSet_prev->shareEditSet_next = &s; |
---|
339 | } |
---|
340 | } |
---|
341 | } |
---|
342 | |
---|
343 | static void assignEditSet(string_res & this, string_res * shareEditSetStartPeer, string_res * shareEditSetEndPeer, |
---|
344 | char * resultSesStart, |
---|
345 | size_t resultSesLnth, |
---|
346 | HandleNode * resultPadPosition, size_t bsize ) { |
---|
347 | |
---|
348 | char * beforeBegin = shareEditSetStartPeer->Handle.s; |
---|
349 | size_t beforeLen = this.Handle.s - beforeBegin; |
---|
350 | |
---|
351 | char * afterBegin = this.Handle.s + this.Handle.lnth; |
---|
352 | size_t afterLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - afterBegin; |
---|
353 | |
---|
354 | size_t oldLnth = this.Handle.lnth; |
---|
355 | |
---|
356 | this.Handle.s = resultSesStart + beforeLen; |
---|
357 | this.Handle.lnth = bsize; |
---|
358 | if (resultPadPosition) |
---|
359 | MoveThisAfter( this.Handle, *resultPadPosition ); |
---|
360 | |
---|
361 | // adjust all substring string and handle locations, and check if any substring strings are outside the new base string |
---|
362 | char *limit = resultSesStart + resultSesLnth; |
---|
363 | for ( string_res * p = this.shareEditSet_next; p != &this; p = p->shareEditSet_next ) { |
---|
364 | verify (p->Handle.s >= beforeBegin); |
---|
365 | if ( p->Handle.s >= afterBegin ) { |
---|
366 | verify ( p->Handle.s <= afterBegin + afterLen ); |
---|
367 | verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen ); |
---|
368 | // p starts after the edit |
---|
369 | // take start and end as end-anchored |
---|
370 | size_t startOffsetFromEnd = afterBegin + afterLen - p->Handle.s; |
---|
371 | p->Handle.s = limit - startOffsetFromEnd; |
---|
372 | // p->Handle.lnth unaffected |
---|
373 | } else if ( p->Handle.s <= beforeBegin + beforeLen ) { |
---|
374 | // p starts before, or at the start of, the edit |
---|
375 | if ( p->Handle.s + p->Handle.lnth <= beforeBegin + beforeLen ) { |
---|
376 | // p ends before the edit |
---|
377 | // take end as start-anchored too |
---|
378 | // p->Handle.lnth unaffected |
---|
379 | } else if ( p->Handle.s + p->Handle.lnth < afterBegin ) { |
---|
380 | // p ends during the edit; p does not include the last character replaced |
---|
381 | // clip end of p to end at start of edit |
---|
382 | p->Handle.lnth = beforeLen - ( p->Handle.s - beforeBegin ); |
---|
383 | } else { |
---|
384 | // p ends after the edit |
---|
385 | verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen ); |
---|
386 | // take end as end-anchored |
---|
387 | // stretch-shrink p according to the edit |
---|
388 | p->Handle.lnth += this.Handle.lnth; |
---|
389 | p->Handle.lnth -= oldLnth; |
---|
390 | } |
---|
391 | // take start as start-anchored |
---|
392 | size_t startOffsetFromStart = p->Handle.s - beforeBegin; |
---|
393 | p->Handle.s = resultSesStart + startOffsetFromStart; |
---|
394 | } else { |
---|
395 | verify ( p->Handle.s < afterBegin ); |
---|
396 | // p starts during the edit |
---|
397 | verify( p->Handle.s + p->Handle.lnth >= beforeBegin + beforeLen ); |
---|
398 | if ( p->Handle.s + p->Handle.lnth < afterBegin ) { |
---|
399 | // p ends during the edit; p does not include the last character replaced |
---|
400 | // set p to empty string at start of edit |
---|
401 | p->Handle.s = this.Handle.s; |
---|
402 | p->Handle.lnth = 0; |
---|
403 | } else { |
---|
404 | // p includes the end of the edit |
---|
405 | // clip start of p to start at end of edit |
---|
406 | int charsToClip = afterBegin - p->Handle.s; |
---|
407 | p->Handle.s = this.Handle.s + this.Handle.lnth; |
---|
408 | p->Handle.lnth -= charsToClip; |
---|
409 | } |
---|
410 | } |
---|
411 | if (resultPadPosition) |
---|
412 | MoveThisAfter( p->Handle, *resultPadPosition ); // move substring handle to maintain sorted order by string position |
---|
413 | } |
---|
414 | } |
---|
415 | |
---|
416 | // traverse the share-edit set (SES) to recover the range of a base string to which `this` belongs |
---|
417 | static void locateInShareEditSet( string_res &this, string_res *&shareEditSetStartPeer, string_res *&shareEditSetEndPeer ) { |
---|
418 | shareEditSetStartPeer = & this; |
---|
419 | shareEditSetEndPeer = & this; |
---|
420 | for (string_res * editPeer = this.shareEditSet_next; editPeer != &this; editPeer = editPeer->shareEditSet_next) { |
---|
421 | if ( editPeer->Handle.s < shareEditSetStartPeer->Handle.s ) { |
---|
422 | shareEditSetStartPeer = editPeer; |
---|
423 | } |
---|
424 | if ( shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth < editPeer->Handle.s + editPeer->Handle.lnth) { |
---|
425 | shareEditSetEndPeer = editPeer; |
---|
426 | } |
---|
427 | } |
---|
428 | } |
---|
429 | |
---|
430 | static string_res & assign_(string_res &this, const char* buffer, size_t bsize, const string_res & valSrc) { |
---|
431 | |
---|
432 | string_res * shareEditSetStartPeer; |
---|
433 | string_res * shareEditSetEndPeer; |
---|
434 | locateInShareEditSet( this, shareEditSetStartPeer, shareEditSetEndPeer ); |
---|
435 | |
---|
436 | verify( shareEditSetEndPeer->Handle.s >= shareEditSetStartPeer->Handle.s ); |
---|
437 | size_t origEditSetLength = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - shareEditSetStartPeer->Handle.s; |
---|
438 | verify( origEditSetLength >= this.Handle.lnth ); |
---|
439 | |
---|
440 | if ( this.shareEditSet_owns_ulink ) { // assigning to private context |
---|
441 | // ok to overwrite old value within LHS |
---|
442 | char * prefixStartOrig = shareEditSetStartPeer->Handle.s; |
---|
443 | int prefixLen = this.Handle.s - prefixStartOrig; |
---|
444 | char * suffixStartOrig = this.Handle.s + this.Handle.lnth; |
---|
445 | int suffixLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - suffixStartOrig; |
---|
446 | |
---|
447 | int delta = bsize - this.Handle.lnth; |
---|
448 | if ( char * oldBytes = VbyteTryAdjustLast( *this.Handle.ulink, delta ) ) { |
---|
449 | // growing: copy from old to new |
---|
450 | char * dest = VbyteAlloc( *this.Handle.ulink, origEditSetLength + delta ); |
---|
451 | char *destCursor = dest; memcpy(destCursor, prefixStartOrig, prefixLen); |
---|
452 | destCursor += prefixLen; memcpy(destCursor, buffer , bsize ); |
---|
453 | destCursor += bsize; memcpy(destCursor, suffixStartOrig, suffixLen); |
---|
454 | assignEditSet(this, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
455 | dest, |
---|
456 | origEditSetLength + delta, |
---|
457 | 0p, bsize); |
---|
458 | free( oldBytes ); |
---|
459 | } else { |
---|
460 | // room is already allocated in-place: bubble suffix and overwite middle |
---|
461 | memmove( suffixStartOrig + delta, suffixStartOrig, suffixLen ); |
---|
462 | memcpy( this.Handle.s, buffer, bsize ); |
---|
463 | |
---|
464 | assignEditSet(this, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
465 | shareEditSetStartPeer->Handle.s, |
---|
466 | origEditSetLength + delta, |
---|
467 | 0p, bsize); |
---|
468 | } |
---|
469 | |
---|
470 | } else if ( // assigning to shared context |
---|
471 | this.Handle.lnth == origEditSetLength && // overwriting entire run of SES |
---|
472 | & valSrc && // sourcing from a managed string |
---|
473 | valSrc.Handle.ulink == this.Handle.ulink ) { // sourcing from same heap |
---|
474 | |
---|
475 | // SES's result will only use characters from the source string => reuse source |
---|
476 | assignEditSet(this, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
477 | valSrc.Handle.s, |
---|
478 | valSrc.Handle.lnth, |
---|
479 | &((string_res&)valSrc).Handle, bsize); |
---|
480 | |
---|
481 | } else { |
---|
482 | // overwriting a proper substring of some string: mash characters from old and new together (copy on write) |
---|
483 | // OR we are importing characters: need to copy eagerly (can't refer to source) |
---|
484 | |
---|
485 | // full string is from start of shareEditSetStartPeer thru end of shareEditSetEndPeer |
---|
486 | // `this` occurs in the middle of it, to be replaced |
---|
487 | // build up the new text in `pasting` |
---|
488 | |
---|
489 | string_res pasting = { |
---|
490 | * this.Handle.ulink, // maintain same heap, regardless of context |
---|
491 | shareEditSetStartPeer->Handle.s, // start of SES |
---|
492 | this.Handle.s - shareEditSetStartPeer->Handle.s }; // length of SES, before this |
---|
493 | append( pasting, |
---|
494 | buffer, // start of replacement for this |
---|
495 | bsize ); // length of replacement for this |
---|
496 | append( pasting, |
---|
497 | this.Handle.s + this.Handle.lnth, // start of SES after this |
---|
498 | shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - |
---|
499 | (this.Handle.s + this.Handle.lnth) ); // length of SES, after this |
---|
500 | |
---|
501 | // The above string building can trigger compaction. |
---|
502 | // The reference points (that are arguments of the string building) may move during that building. |
---|
503 | // From this point on, they are stable. |
---|
504 | |
---|
505 | assignEditSet(this, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
506 | pasting.Handle.s, |
---|
507 | pasting.Handle.lnth, |
---|
508 | &pasting.Handle, bsize); |
---|
509 | } |
---|
510 | |
---|
511 | return this; |
---|
512 | } |
---|
513 | |
---|
514 | string_res & assign(string_res &this, const char* buffer, size_t bsize) { |
---|
515 | return assign_(this, buffer, bsize, *0p); |
---|
516 | } |
---|
517 | |
---|
518 | string_res & ?=?(string_res &s, char other) { |
---|
519 | return assign(s, &other, 1); |
---|
520 | } |
---|
521 | |
---|
522 | // Copy assignment operator |
---|
523 | string_res & ?=?(string_res & this, const string_res & rhs) with( this ) { |
---|
524 | return assign_(this, rhs.Handle.s, rhs.Handle.lnth, rhs); |
---|
525 | } |
---|
526 | |
---|
527 | string_res & ?=?(string_res & this, string_res & rhs) with( this ) { |
---|
528 | const string_res & rhs2 = rhs; |
---|
529 | return this = rhs2; |
---|
530 | } |
---|
531 | |
---|
532 | |
---|
533 | // Destructor |
---|
534 | void ^?{}(string_res &s) with(s) { |
---|
535 | // much delegated to implied ^VbyteSM |
---|
536 | |
---|
537 | // sever s from its share-edit peers, if any (four no-ops when already solo) |
---|
538 | s.shareEditSet_prev->shareEditSet_next = s.shareEditSet_next; |
---|
539 | s.shareEditSet_next->shareEditSet_prev = s.shareEditSet_prev; |
---|
540 | // s.shareEditSet_next = &s; |
---|
541 | // s.shareEditSet_prev = &s; |
---|
542 | |
---|
543 | if (shareEditSet_owns_ulink && s.shareEditSet_next == &s) { // last one out |
---|
544 | delete( s.Handle.ulink ); |
---|
545 | } |
---|
546 | } |
---|
547 | |
---|
548 | |
---|
549 | // Returns the character at the given index |
---|
550 | // With unicode support, this may be different from just the byte at the given |
---|
551 | // offset from the start of the string. |
---|
552 | char ?[?](const string_res &s, size_t index) with(s) { |
---|
553 | //TODO: Check if index is valid (no exceptions yet) |
---|
554 | return Handle.s[index]; |
---|
555 | } |
---|
556 | |
---|
557 | void assignAt(const string_res &s, size_t index, char val) { |
---|
558 | string_res editZone = { s, SHARE_EDITS, index, index+1 }; |
---|
559 | assign(editZone, &val, 1); |
---|
560 | } |
---|
561 | |
---|
562 | |
---|
563 | /////////////////////////////////////////////////////////////////// |
---|
564 | // Concatenation |
---|
565 | |
---|
566 | void append(string_res &str1, const char * buffer, size_t bsize) { |
---|
567 | size_t clnth = str1.Handle.lnth + bsize; |
---|
568 | if ( str1.Handle.s + str1.Handle.lnth == buffer ) { // already juxtapose ? |
---|
569 | // no-op |
---|
570 | } else { // must copy some text |
---|
571 | if ( str1.Handle.s + str1.Handle.lnth == VbyteAlloc(*str1.Handle.ulink, 0) ) { // str1 at end of string area ? |
---|
572 | VbyteAlloc( *str1.Handle.ulink, bsize ); // create room for 2nd part at the end of string area |
---|
573 | } else { // copy the two parts |
---|
574 | char * str1newBuf = VbyteAlloc( *str1.Handle.ulink, clnth ); |
---|
575 | char * str1oldBuf = str1.Handle.s; // must read after VbyteAlloc call in case it gs's |
---|
576 | str1.Handle.s = str1newBuf; |
---|
577 | memcpy( str1.Handle.s, str1oldBuf, str1.Handle.lnth ); |
---|
578 | } // if |
---|
579 | memcpy( str1.Handle.s + str1.Handle.lnth, buffer, bsize ); |
---|
580 | } // if |
---|
581 | str1.Handle.lnth = clnth; |
---|
582 | } |
---|
583 | |
---|
584 | void ?+=?(string_res &str1, const string_res &str2) { |
---|
585 | append( str1, str2.Handle.s, str2.Handle.lnth ); |
---|
586 | } |
---|
587 | |
---|
588 | void ?+=?(string_res &s, char other) { |
---|
589 | append( s, &other, 1 ); |
---|
590 | } |
---|
591 | |
---|
592 | |
---|
593 | |
---|
594 | |
---|
595 | |
---|
596 | ////////////////////////////////////////////////////////// |
---|
597 | // Comparisons |
---|
598 | |
---|
599 | |
---|
600 | bool ?==?(const string_res &s1, const string_res &s2) { |
---|
601 | return ByteCmp( s1.Handle.s, 0, s1.Handle.lnth, s2.Handle.s, 0, s2.Handle.lnth) == 0; |
---|
602 | } |
---|
603 | |
---|
604 | bool ?!=?(const string_res &s1, const string_res &s2) { |
---|
605 | return !(s1 == s2); |
---|
606 | } |
---|
607 | bool ?==?(const string_res &s, const char* other) { |
---|
608 | string_res sother = other; |
---|
609 | return s == sother; |
---|
610 | } |
---|
611 | bool ?!=?(const string_res &s, const char* other) { |
---|
612 | return !(s == other); |
---|
613 | } |
---|
614 | |
---|
615 | |
---|
616 | ////////////////////////////////////////////////////////// |
---|
617 | // Search |
---|
618 | |
---|
619 | bool contains(const string_res &s, char ch) { |
---|
620 | for ( i; size(s) ) { |
---|
621 | if (s[i] == ch) return true; |
---|
622 | } |
---|
623 | return false; |
---|
624 | } |
---|
625 | |
---|
626 | int find(const string_res &s, char search) { |
---|
627 | return findFrom(s, 0, search); |
---|
628 | } |
---|
629 | |
---|
630 | int findFrom(const string_res &s, size_t fromPos, char search) { |
---|
631 | // FIXME: This paricular overload (find of single char) is optimized to use memchr. |
---|
632 | // The general overload (find of string, memchr applying to its first character) and `contains` should be adjusted to match. |
---|
633 | char * searchFrom = s.Handle.s + fromPos; |
---|
634 | size_t searchLnth = s.Handle.lnth - fromPos; |
---|
635 | int searchVal = search; |
---|
636 | char * foundAt = (char *) memchr(searchFrom, searchVal, searchLnth); |
---|
637 | if (foundAt == 0p) return s.Handle.lnth; |
---|
638 | else return foundAt - s.Handle.s; |
---|
639 | } |
---|
640 | |
---|
641 | int find(const string_res &s, const string_res &search) { |
---|
642 | return findFrom(s, 0, search); |
---|
643 | } |
---|
644 | |
---|
645 | int findFrom(const string_res &s, size_t fromPos, const string_res &search) { |
---|
646 | return findFrom(s, fromPos, search.Handle.s, search.Handle.lnth); |
---|
647 | } |
---|
648 | |
---|
649 | int find(const string_res &s, const char* search) { |
---|
650 | return findFrom(s, 0, search); |
---|
651 | } |
---|
652 | int findFrom(const string_res &s, size_t fromPos, const char* search) { |
---|
653 | return findFrom(s, fromPos, search, strlen(search)); |
---|
654 | } |
---|
655 | |
---|
656 | int find(const string_res &s, const char* search, size_t searchsize) { |
---|
657 | return findFrom(s, 0, search, searchsize); |
---|
658 | } |
---|
659 | |
---|
660 | int findFrom(const string_res &s, size_t fromPos, const char* search, size_t searchsize) { |
---|
661 | |
---|
662 | /* Remaining implementations essentially ported from Sunjay's work */ |
---|
663 | |
---|
664 | |
---|
665 | // FIXME: This is a naive algorithm. We probably want to switch to someting |
---|
666 | // like Boyer-Moore in the future. |
---|
667 | // https://en.wikipedia.org/wiki/String_searching_algorithm |
---|
668 | |
---|
669 | // Always find the empty string |
---|
670 | if (searchsize == 0) { |
---|
671 | return 0; |
---|
672 | } |
---|
673 | |
---|
674 | for ( i; fromPos ~ s.Handle.lnth ) { |
---|
675 | size_t remaining = s.Handle.lnth - i; |
---|
676 | // Never going to find the search string if the remaining string is |
---|
677 | // smaller than search |
---|
678 | if (remaining < searchsize) { |
---|
679 | break; |
---|
680 | } |
---|
681 | |
---|
682 | bool matched = true; |
---|
683 | for ( j; searchsize ) { |
---|
684 | if (search[j] != s.Handle.s[i + j]) { |
---|
685 | matched = false; |
---|
686 | break; |
---|
687 | } |
---|
688 | } |
---|
689 | if (matched) { |
---|
690 | return i; |
---|
691 | } |
---|
692 | } |
---|
693 | |
---|
694 | return s.Handle.lnth; |
---|
695 | } |
---|
696 | |
---|
697 | bool includes(const string_res &s, const string_res &search) { |
---|
698 | return includes(s, search.Handle.s, search.Handle.lnth); |
---|
699 | } |
---|
700 | |
---|
701 | bool includes(const string_res &s, const char* search) { |
---|
702 | return includes(s, search, strlen(search)); |
---|
703 | } |
---|
704 | |
---|
705 | bool includes(const string_res &s, const char* search, size_t searchsize) { |
---|
706 | return find(s, search, searchsize) < s.Handle.lnth; |
---|
707 | } |
---|
708 | |
---|
709 | bool startsWith(const string_res &s, const string_res &prefix) { |
---|
710 | return startsWith(s, prefix.Handle.s, prefix.Handle.lnth); |
---|
711 | } |
---|
712 | |
---|
713 | bool startsWith(const string_res &s, const char* prefix) { |
---|
714 | return startsWith(s, prefix, strlen(prefix)); |
---|
715 | } |
---|
716 | |
---|
717 | bool startsWith(const string_res &s, const char* prefix, size_t prefixsize) { |
---|
718 | if (s.Handle.lnth < prefixsize) { |
---|
719 | return false; |
---|
720 | } |
---|
721 | return memcmp(s.Handle.s, prefix, prefixsize) == 0; |
---|
722 | } |
---|
723 | |
---|
724 | bool endsWith(const string_res &s, const string_res &suffix) { |
---|
725 | return endsWith(s, suffix.Handle.s, suffix.Handle.lnth); |
---|
726 | } |
---|
727 | |
---|
728 | bool endsWith(const string_res &s, const char* suffix) { |
---|
729 | return endsWith(s, suffix, strlen(suffix)); |
---|
730 | } |
---|
731 | |
---|
732 | bool endsWith(const string_res &s, const char* suffix, size_t suffixsize) { |
---|
733 | if (s.Handle.lnth < suffixsize) { |
---|
734 | return false; |
---|
735 | } |
---|
736 | // Amount to offset the bytes pointer so that we are comparing the end of s |
---|
737 | // to suffix. s.bytes + offset should be the first byte to compare against suffix |
---|
738 | size_t offset = s.Handle.lnth - suffixsize; |
---|
739 | return memcmp(s.Handle.s + offset, suffix, suffixsize) == 0; |
---|
740 | } |
---|
741 | |
---|
742 | /* Back to Mike's work */ |
---|
743 | |
---|
744 | |
---|
745 | /////////////////////////////////////////////////////////////////////////// |
---|
746 | // charclass, include, exclude |
---|
747 | |
---|
748 | void ?{}( charclass_res & this, const string_res & chars) { |
---|
749 | (this){ chars.Handle.s, chars.Handle.lnth }; |
---|
750 | } |
---|
751 | |
---|
752 | void ?{}( charclass_res & this, const char * chars ) { |
---|
753 | (this){ chars, strlen(chars) }; |
---|
754 | } |
---|
755 | |
---|
756 | void ?{}( charclass_res & this, const char * chars, size_t charssize ) { |
---|
757 | (this.chars){ chars, charssize }; |
---|
758 | // now sort it ? |
---|
759 | } |
---|
760 | |
---|
761 | void ^?{}( charclass_res & this ) { |
---|
762 | ^(this.chars){}; |
---|
763 | } |
---|
764 | |
---|
765 | static bool test( const charclass_res & mask, char c ) { |
---|
766 | // instead, use sorted char list? |
---|
767 | return contains( mask.chars, c ); |
---|
768 | } |
---|
769 | |
---|
770 | int exclude(const string_res &s, const charclass_res &mask) { |
---|
771 | for ( i; size(s) ) { |
---|
772 | if ( test(mask, s[i]) ) return i; |
---|
773 | } |
---|
774 | return size(s); |
---|
775 | } |
---|
776 | |
---|
777 | int include(const string_res &s, const charclass_res &mask) { |
---|
778 | for ( i; size(s) ) { |
---|
779 | if ( ! test(mask, s[i]) ) return i; |
---|
780 | } |
---|
781 | return size(s); |
---|
782 | } |
---|
783 | |
---|
784 | //######################### VbyteHeap "implementation" ######################### |
---|
785 | |
---|
786 | |
---|
787 | // Add a new HandleNode node n after the current HandleNode node. |
---|
788 | |
---|
789 | static void AddThisAfter( HandleNode & this, HandleNode & n ) with(this) { |
---|
790 | #ifdef VbyteDebug |
---|
791 | serr | "enter:AddThisAfter, this:" | &this | " n:" | &n; |
---|
792 | #endif // VbyteDebug |
---|
793 | // Performance note: we are on the critical path here. MB has ensured that the verifies don't contribute to runtime (are compiled away, like they're supposed to be). |
---|
794 | verify( n.ulink != 0p ); |
---|
795 | verify( this.ulink == n.ulink ); |
---|
796 | flink = n.flink; |
---|
797 | blink = &n; |
---|
798 | n.flink->blink = &this; |
---|
799 | n.flink = &this; |
---|
800 | #ifdef VbyteDebug |
---|
801 | { |
---|
802 | serr | "HandleList:"; |
---|
803 | serr | nlOff; |
---|
804 | for ( HandleNode *ni = HeaderPtr->flink; ni != HeaderPtr; ni = ni->flink ) { |
---|
805 | serr | "\tnode:" | ni | " lnth:" | ni->lnth | " s:" | (void *)ni->s | ",\""; |
---|
806 | for ( i; ni->lnth ) { |
---|
807 | serr | ni->s[i]; |
---|
808 | } // for |
---|
809 | serr | "\" flink:" | ni->flink | " blink:" | ni->blink | nl; |
---|
810 | } // for |
---|
811 | serr | nlOn; |
---|
812 | } |
---|
813 | serr | "exit:AddThisAfter"; |
---|
814 | #endif // VbyteDebug |
---|
815 | } // AddThisAfter |
---|
816 | |
---|
817 | |
---|
818 | // Delete the current HandleNode node. |
---|
819 | |
---|
820 | static void DeleteNode( HandleNode & this ) with(this) { |
---|
821 | #ifdef VbyteDebug |
---|
822 | serr | "enter:DeleteNode, this:" | &this; |
---|
823 | #endif // VbyteDebug |
---|
824 | flink->blink = blink; |
---|
825 | blink->flink = flink; |
---|
826 | #ifdef VbyteDebug |
---|
827 | serr | "exit:DeleteNode"; |
---|
828 | #endif // VbyteDebug |
---|
829 | } // DeleteNode |
---|
830 | |
---|
831 | |
---|
832 | |
---|
833 | // Allocates specified storage for a string from byte-string area. If not enough space remains to perform the |
---|
834 | // allocation, the garbage collection routine is called. |
---|
835 | |
---|
836 | static char * VbyteAlloc( VbyteHeap & this, int size ) with(this) { |
---|
837 | #ifdef VbyteDebug |
---|
838 | serr | "enter:VbyteAlloc, size:" | size; |
---|
839 | #endif // VbyteDebug |
---|
840 | uintptr_t NoBytes; |
---|
841 | char *r; |
---|
842 | |
---|
843 | NoBytes = ( uintptr_t )EndVbyte + size; |
---|
844 | if ( NoBytes > ( uintptr_t )ExtVbyte ) { // enough room for new byte-string ? |
---|
845 | garbage( this, size ); // firer up the garbage collector |
---|
846 | verify( (( uintptr_t )EndVbyte + size) <= ( uintptr_t )ExtVbyte && "garbage run did not free up required space" ); |
---|
847 | } // if |
---|
848 | r = EndVbyte; |
---|
849 | EndVbyte += size; |
---|
850 | #ifdef VbyteDebug |
---|
851 | serr | "exit:VbyteAlloc, r:" | (void *)r | " EndVbyte:" | (void *)EndVbyte | " ExtVbyte:" | ExtVbyte; |
---|
852 | #endif // VbyteDebug |
---|
853 | return r; |
---|
854 | } // VbyteAlloc |
---|
855 | |
---|
856 | |
---|
857 | // Adjusts the last allocation in this heap by delta bytes, or resets this heap to be able to offer |
---|
858 | // new allocations of its original size + delta bytes. Positive delta means bigger; |
---|
859 | // negative means smaller. A null return indicates that the original heap location has room for |
---|
860 | // the requested growth. A non-null return indicates that copying to a new location is required |
---|
861 | // but has not been done; the returned value is the old heap storage location; `this` heap is |
---|
862 | // modified to reference the new location. In the copy-requred case, the caller should use |
---|
863 | // VbyteAlloc to claim the new space, while doing optimal copying from old to new, then free old. |
---|
864 | |
---|
865 | static char * VbyteTryAdjustLast( VbyteHeap & this, int delta ) with(this) { |
---|
866 | |
---|
867 | if ( ( uintptr_t )EndVbyte + delta <= ( uintptr_t )ExtVbyte ) { |
---|
868 | // room available |
---|
869 | EndVbyte += delta; |
---|
870 | return 0p; |
---|
871 | } |
---|
872 | |
---|
873 | char *oldBytes = StartVbyte; |
---|
874 | |
---|
875 | NoOfExtensions += 1; |
---|
876 | CurrSize *= 2; |
---|
877 | StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize); |
---|
878 | ExtVbyte = StartVbyte + CurrSize; |
---|
879 | |
---|
880 | return oldBytes; |
---|
881 | } |
---|
882 | |
---|
883 | |
---|
884 | // Move an existing HandleNode node h somewhere after the current HandleNode node so that it is in ascending order by |
---|
885 | // the address in the byte string area. |
---|
886 | |
---|
887 | static void MoveThisAfter( HandleNode & this, const HandleNode & h ) with(this) { |
---|
888 | #ifdef VbyteDebug |
---|
889 | serr | "enter:MoveThisAfter, this:" | & this | " h:" | & h; |
---|
890 | #endif // VbyteDebug |
---|
891 | verify( h.ulink != 0p ); |
---|
892 | verify( this.ulink == h.ulink ); |
---|
893 | if ( s < h.s ) { // check argument values |
---|
894 | // serr | "VbyteSM: Error - Cannot move byte string starting at:" | s | " after byte string starting at:" |
---|
895 | // | ( h->s ) | " and keep handles in ascending order"; |
---|
896 | // exit(-1 ); |
---|
897 | verify( 0 && "VbyteSM: Error - Cannot move byte strings as requested and keep handles in ascending order"); |
---|
898 | } // if |
---|
899 | |
---|
900 | HandleNode *i; |
---|
901 | for ( i = h.flink; i->s != 0 && s > ( i->s ); i = i->flink ); // find the position for this node after h |
---|
902 | if ( & this != i->blink ) { |
---|
903 | DeleteNode( this ); |
---|
904 | AddThisAfter( this, *i->blink ); |
---|
905 | } // if |
---|
906 | #ifdef VbyteDebug |
---|
907 | { |
---|
908 | serr | "HandleList:"; |
---|
909 | serr | nlOff; |
---|
910 | for ( HandleNode *n = HeaderPtr->flink; n != HeaderPtr; n = n->flink ) { |
---|
911 | serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; |
---|
912 | for ( i; n->lnth ) { |
---|
913 | serr | n->s[i]; |
---|
914 | } // for |
---|
915 | serr | "\" flink:" | n->flink | " blink:" | n->blink | nl; |
---|
916 | } // for |
---|
917 | serr | nlOn; |
---|
918 | } |
---|
919 | serr | "exit:MoveThisAfter"; |
---|
920 | #endif // VbyteDebug |
---|
921 | } // MoveThisAfter |
---|
922 | |
---|
923 | |
---|
924 | |
---|
925 | |
---|
926 | |
---|
927 | //######################### VbyteHeap ######################### |
---|
928 | |
---|
929 | // Compare two byte strings in the byte-string area. The routine returns the following values: |
---|
930 | // |
---|
931 | // 1 => Src1-byte-string > Src2-byte-string |
---|
932 | // 0 => Src1-byte-string = Src2-byte-string |
---|
933 | // -1 => Src1-byte-string < Src2-byte-string |
---|
934 | |
---|
935 | int ByteCmp( char *Src1, int Src1Start, int Src1Lnth, char *Src2, int Src2Start, int Src2Lnth ) { |
---|
936 | #ifdef VbyteDebug |
---|
937 | serr | "enter:ByteCmp, Src1Start:" | Src1Start | " Src1Lnth:" | Src1Lnth | " Src2Start:" | Src2Start | " Src2Lnth:" | Src2Lnth; |
---|
938 | #endif // VbyteDebug |
---|
939 | int cmp; |
---|
940 | |
---|
941 | CharZip: for ( int i = 0; ; i += 1 ) { |
---|
942 | if ( i == Src2Lnth - 1 ) { |
---|
943 | for ( ; ; i += 1 ) { |
---|
944 | if ( i == Src1Lnth - 1 ) { |
---|
945 | cmp = 0; |
---|
946 | break CharZip; |
---|
947 | } // exit |
---|
948 | if ( Src1[Src1Start + i] != ' ') { |
---|
949 | // SUSPECTED BUG: this could be be why Peter got the bug report about == " " (why is this case here at all?) |
---|
950 | cmp = 1; |
---|
951 | break CharZip; |
---|
952 | } // exit |
---|
953 | } // for |
---|
954 | } // exit |
---|
955 | if ( i == Src1Lnth - 1 ) { |
---|
956 | for ( ; ; i += 1 ) { |
---|
957 | if ( i == Src2Lnth - 1 ) { |
---|
958 | cmp = 0; |
---|
959 | break CharZip; |
---|
960 | } // exit |
---|
961 | if ( Src2[Src2Start + i] != ' ') { |
---|
962 | cmp = -1; |
---|
963 | break CharZip; |
---|
964 | } // exit |
---|
965 | } // for |
---|
966 | } // exit |
---|
967 | if ( Src2[Src2Start + i] != Src1[Src1Start+ i]) { |
---|
968 | cmp = Src1[Src1Start + i] > Src2[Src2Start + i] ? 1 : -1; |
---|
969 | break CharZip; |
---|
970 | } // exit |
---|
971 | } // for |
---|
972 | #ifdef VbyteDebug |
---|
973 | serr | "exit:ByteCmp, cmp:" | cmp; |
---|
974 | #endif // VbyteDebug |
---|
975 | return cmp; |
---|
976 | } // ByteCmp |
---|
977 | |
---|
978 | |
---|
979 | // The compaction moves all of the byte strings currently in use to the beginning of the byte-string area and modifies |
---|
980 | // the handles to reflect the new positions of the byte strings. Compaction assumes that the handle list is in ascending |
---|
981 | // order by pointers into the byte-string area. The strings associated with substrings do not have to be moved because |
---|
982 | // the containing string has been moved. Hence, they only require that their string pointers be adjusted. |
---|
983 | |
---|
984 | void compaction(VbyteHeap & this) with(this) { |
---|
985 | HandleNode *h; |
---|
986 | char *obase, *nbase, *limit; |
---|
987 | |
---|
988 | NoOfCompactions += 1; |
---|
989 | EndVbyte = StartVbyte; |
---|
990 | h = Header.flink; // ignore header node |
---|
991 | for () { |
---|
992 | memmove( EndVbyte, h->s, h->lnth ); |
---|
993 | obase = h->s; |
---|
994 | h->s = EndVbyte; |
---|
995 | nbase = h->s; |
---|
996 | EndVbyte += h->lnth; |
---|
997 | limit = obase + h->lnth; |
---|
998 | h = h->flink; |
---|
999 | |
---|
1000 | // check if any substrings are allocated within a string |
---|
1001 | |
---|
1002 | for () { |
---|
1003 | if ( h == &Header ) break; // end of header list ? |
---|
1004 | if ( h->s >= limit ) break; // outside of current string ? |
---|
1005 | h->s = nbase + (( uintptr_t )h->s - ( uintptr_t )obase ); |
---|
1006 | h = h->flink; |
---|
1007 | } // for |
---|
1008 | if ( h == &Header ) break; // end of header list ? |
---|
1009 | } // for |
---|
1010 | } // compaction |
---|
1011 | |
---|
1012 | |
---|
1013 | static double heap_expansion_freespace_threshold = 0.1; // default inherited from prior work: expand heap when less than 10% "free" (i.e. garbage) |
---|
1014 | // probably an unreasonable default, but need to assess early-round tests on changing it |
---|
1015 | |
---|
1016 | void TUNING_set_string_heap_liveness_threshold( double val ) { |
---|
1017 | heap_expansion_freespace_threshold = 1.0 - val; |
---|
1018 | } |
---|
1019 | |
---|
1020 | |
---|
1021 | // Garbage determines the amount of free space left in the heap and then reduces, leave the same, or extends the size of |
---|
1022 | // the heap. The heap is then compacted in the existing heap or into the newly allocated heap. |
---|
1023 | |
---|
1024 | void garbage(VbyteHeap & this, int minreq ) with(this) { |
---|
1025 | #ifdef VbyteDebug |
---|
1026 | serr | "enter:garbage"; |
---|
1027 | { |
---|
1028 | serr | "HandleList:"; |
---|
1029 | for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) { |
---|
1030 | serr | nlOff; |
---|
1031 | serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; |
---|
1032 | for ( i; n->lnth ) { |
---|
1033 | serr | n->s[i]; |
---|
1034 | } // for |
---|
1035 | serr | nlOn; |
---|
1036 | serr | "\" flink:" | n->flink | " blink:" | n->blink; |
---|
1037 | } // for |
---|
1038 | } |
---|
1039 | #endif // VbyteDebug |
---|
1040 | int AmountUsed, AmountFree; |
---|
1041 | |
---|
1042 | AmountUsed = 0; |
---|
1043 | for ( HandleNode *i = Header.flink; i != &Header; i = i->flink ) { // calculate amount of byte area used |
---|
1044 | AmountUsed += i->lnth; |
---|
1045 | } // for |
---|
1046 | AmountFree = ( uintptr_t )ExtVbyte - ( uintptr_t )StartVbyte - AmountUsed; |
---|
1047 | |
---|
1048 | if ( ( double ) AmountFree < ( CurrSize * heap_expansion_freespace_threshold ) || AmountFree < minreq ) { // free space less than threshold or not enough to serve cur request |
---|
1049 | |
---|
1050 | extend( this, max( CurrSize, minreq ) ); // extend the heap |
---|
1051 | |
---|
1052 | // Peter says, "This needs work before it should be used." |
---|
1053 | // } else if ( AmountFree > CurrSize / 2 ) { // free space greater than 3 times the initial allocation ? |
---|
1054 | // reduce(( AmountFree / CurrSize - 3 ) * CurrSize ); // reduce the memory |
---|
1055 | |
---|
1056 | // `extend` implies a `compaction` during the copy |
---|
1057 | |
---|
1058 | } else { |
---|
1059 | compaction(this); // in-place |
---|
1060 | }// if |
---|
1061 | #ifdef VbyteDebug |
---|
1062 | { |
---|
1063 | serr | "HandleList:"; |
---|
1064 | for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) { |
---|
1065 | serr | nlOff; |
---|
1066 | serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; |
---|
1067 | for ( i; n->lnth ) { |
---|
1068 | serr | n->s[i]; |
---|
1069 | } // for |
---|
1070 | serr | nlOn; |
---|
1071 | serr | "\" flink:" | n->flink | " blink:" | n->blink; |
---|
1072 | } // for |
---|
1073 | } |
---|
1074 | serr | "exit:garbage"; |
---|
1075 | #endif // VbyteDebug |
---|
1076 | } // garbage |
---|
1077 | |
---|
1078 | #undef VbyteDebug |
---|
1079 | |
---|
1080 | |
---|
1081 | |
---|
1082 | // Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string |
---|
1083 | // area is deleted. |
---|
1084 | |
---|
1085 | void extend( VbyteHeap & this, int size ) with (this) { |
---|
1086 | #ifdef VbyteDebug |
---|
1087 | serr | "enter:extend, size:" | size; |
---|
1088 | #endif // VbyteDebug |
---|
1089 | char *OldStartVbyte; |
---|
1090 | |
---|
1091 | NoOfExtensions += 1; |
---|
1092 | OldStartVbyte = StartVbyte; // save previous byte area |
---|
1093 | |
---|
1094 | CurrSize += size > InitSize ? size : InitSize; // minimum extension, initial size |
---|
1095 | StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize); |
---|
1096 | ExtVbyte = (void *)( StartVbyte + CurrSize ); |
---|
1097 | compaction(this); // copy from old heap to new & adjust pointers to new heap |
---|
1098 | free( OldStartVbyte ); // release old heap |
---|
1099 | #ifdef VbyteDebug |
---|
1100 | serr | "exit:extend, CurrSize:" | CurrSize; |
---|
1101 | #endif // VbyteDebug |
---|
1102 | } // extend |
---|
1103 | |
---|
1104 | //WIP |
---|
1105 | #if 0 |
---|
1106 | |
---|
1107 | // Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string |
---|
1108 | // area is deleted. |
---|
1109 | |
---|
1110 | void VbyteHeap::reduce( int size ) { |
---|
1111 | #ifdef VbyteDebug |
---|
1112 | serr | "enter:reduce, size:" | size; |
---|
1113 | #endif // VbyteDebug |
---|
1114 | char *OldStartVbyte; |
---|
1115 | |
---|
1116 | NoOfReductions += 1; |
---|
1117 | OldStartVbyte = StartVbyte; // save previous byte area |
---|
1118 | |
---|
1119 | CurrSize -= size; |
---|
1120 | StartVbyte = EndVbyte = new char[CurrSize]; |
---|
1121 | ExtVbyte = (void *)( StartVbyte + CurrSize ); |
---|
1122 | compaction(); // copy from old heap to new & adjust pointers to new heap |
---|
1123 | delete OldStartVbyte; // release old heap |
---|
1124 | #ifdef VbyteDebug |
---|
1125 | serr | "exit:reduce, CurrSize:" | CurrSize; |
---|
1126 | #endif // VbyteDebug |
---|
1127 | } // reduce |
---|
1128 | |
---|
1129 | |
---|
1130 | #endif |
---|