Context Navigation

string_res.cfa @ f678c53b

Last change on this file since f678c53b was fbe3f03, checked in by Peter A. Buhr <pabuhr@…>, 7 months ago
formatting
Property mode set to `100644`
File size: 42.4 KB

Line
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 Apr 15 21:56:27 2024
13	// Update Count : 85
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	#include <complex.h> // creal, cimag
28
29	//######################### VbyteHeap "header" #########################
30
31	#ifdef VbyteDebug
32	HandleNode *HeaderPtr;
33	#endif // VbyteDebug
34
35	struct VbyteHeap {
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 & s, size_t Size ) with(s) {
70	#ifdef VbyteDebug
71	serr \| "enter:VbyteHeap::VbyteHeap, s:" \| &s \| " 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 = &s;
79	#ifdef VbyteDebug
80	HeaderPtr = &Header;
81	serr \| "exit:VbyteHeap::VbyteHeap, s:" \| &s;
82	#endif // VbyteDebug
83	} // VbyteHeap
84
85
86	// Release the dynamically allocated storage for the byte area.
87
88	static void ^?{}( VbyteHeap & s ) with(s) {
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 & s ) with(s) {
100	#ifdef VbyteDebug
101	serr \| "enter:HandleNode::HandleNode, s:" \| &s;
102	#endif // VbyteDebug
103	s = 0;
104	lnth = 0;
105	#ifdef VbyteDebug
106	serr \| "exit:HandleNode::HandleNode, s:" \| &s;
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 & s, VbyteHeap & vh ) with(s) {
115	#ifdef VbyteDebug
116	serr \| "enter:HandleNode::HandleNode, s:" \| &s;
117	#endif // VbyteDebug
118	s = 0;
119	lnth = 0;
120	ulink = &vh;
121	AddThisAfter( s, *vh.Header.blink );
122	#ifdef VbyteDebug
123	serr \| "exit:HandleNode::HandleNode, s:" \| &s;
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 & s ) with(s) {
132	#ifdef VbyteDebug
133	serr \| "enter:HandleNode::~HandleNode, s:" \| & s;
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( s );
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 & s, StringSharectx_Mode mode ) with( s ) {
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 = & s;
162	}
163
164	void ^?{}( string_sharectx & s ) with( s ) {
165	if ( activeHeap ) delete( activeHeap );
166
167	// unlink s from older-list starting from ambient_string_sharectx
168	// usually, s==ambient_string_sharectx and the loop runs zero times
169	string_sharectx *& c = ambient_string_sharectx;
170	while ( c != &s ) &c = &c->older; // find s
171	c = s.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	return out \| wd( 0, s.Handle.lnth, s.Handle.s ) \| nonl;
203	}
204
205	void ?\|?(ofstream & out, const string_res & s) {
206	(ofstream &)(out \| s); ends( out );
207	}
208
209	// Input operator
210	ifstream & ?\|?(ifstream & in, string_res & s) {
211	// Reading into a temp before assigning to s is near zero overhead in typical cases because of sharing.
212	// If s is a substring of something larger, simple assignment takes care of that case correctly.
213	// But directly reading a variable amount of text into the middle of a larger context is not practical.
214	string_res temp;
215
216	// Read in chunks. Often, one chunk is enough. Keep the string that accumulates chunks last in the heap,
217	// so available room is rest of heap. When a chunk fills the heap, force growth then take the next chunk.
218	for (bool cont = true; cont; ) {
219	cont = false;
220
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 is available to read into
231	int lenReadable = (char *)temp.Handle.ulink->ExtVbyte - temp.Handle.ulink->EndVbyte;
232	assert (lenReadable >= 2);
233
234	// get bytes
235	try {
236	*(temp.Handle.ulink->EndVbyte) = '\0'; // pre-assign empty cstring
237	in \| wdi( lenReadable, temp.Handle.ulink->EndVbyte );
238	} catch (cstring_length *) {
239	cont = true;
240	}
241	int lenWasRead = strlen(temp.Handle.ulink->EndVbyte);
242
243	// update metadata
244	temp.Handle.lnth += lenWasRead;
245	temp.Handle.ulink->EndVbyte += lenWasRead;
246	}
247
248	if ( temp.Handle.lnth > 0 ) s = temp;
249	return in;
250	}
251
252	ifstream & ?\|?( ifstream & is, _Istream_Rquoted f ) with( f.rstr ) {
253	if ( eof( is ) ) throwResume ExceptionInst( missing_data );
254	int args;
255	fini: {
256	char rfmt[5] = { ' ', delimiters[0], '%', 'n', '\0' };
257	int len = -1; // may not be set in fmt
258	args = fmt( is, rfmt, &len ); // remove leading whitespace and quote
259	if ( eof( is ) \|\| len == -1 ) break fini;
260
261	// Change the remainder of the read into a getline by reseting the closing delimiter.
262	if ( delimiters[1] != '\0' ) {
263	delimiters[0] = delimiters[1];
264	delimiters[1] = '\0';
265	} // if
266	flags.delimiter = true;
267	return is \| (_Istream_Rstr )&f;
268	} // fini
269	// read failed => no pattern match => set string to null
270	if ( ! flags.ignore && s != 0p && args == 0 ) s[0] = '\0';
271	if ( args == 1 && eof( is ) ) { // data but scan ended at EOF
272	clear( is ); // => reset EOF => detect again on next read
273	} // if
274	return is;
275	}
276
277	ifstream & ?\|?( ifstream & is, _Istream_Rstr f ) {
278	// .---------------,
279	// \| \| \| \| \|...\|0\|0\| null terminator and guard if missing
280	// `---------------'
281	enum { gwd = 128 + 1, wd = gwd - 1 }; // guard and unguard width
282	char cstr[gwd]; // read in chunks
283	bool cont = false;
284
285	_Istream_Cwidth cf = { cstr, (_Istream_str_base)f };
286	if ( ! cf.flags.rwd ) cf.wd = wd;
287
288	cstr[wd] = '\0'; // guard null terminate string
289	try {
290	cstr[0] = '\0'; // pre-assign as empty cstring
291	is \| cf;
292	} catch( cstring_length * ) {
293	cont = true;
294	} finally {
295	if ( ! cf.flags.ignore // ok to initialize string
296	// && cstr[0] != '\0' // something was read
297	) {
298	*(f.s) = cstr;
299	}
300	} // try
301	for ( ; cont; ) { // overflow read ?
302	cont = false;
303	try {
304	cstr[0] = '\0'; // pre-assign as empty cstring
305	is \| cf;
306	} catch( cstring_length * ) {
307	cont = true; // continue not allowed
308	} finally {
309	if ( ! cf.flags.ignore && cstr[0] != '\0' ) { // something was read
310	*(f.s) += cstr; // build string chunk at a time
311	}
312	} // try
313	} // for
314	return is;
315	} // ?\|?
316
317	// Empty constructor
318	void ?{}(string_res & s) with(s) {
319	if( ambient_string_sharectx->activeHeap ) {
320	(Handle){ * ambient_string_sharectx->activeHeap };
321	(shareEditSet_owns_ulink){ false };
322	verify( Handle.s == 0p && Handle.lnth == 0 );
323	} else {
324	(Handle){ * new( (size_t) 10 ) }; // TODO: can I lazily avoid allocating for empty string
325	(shareEditSet_owns_ulink){ true };
326	Handle.s = Handle.ulink->StartVbyte;
327	verify( Handle.lnth == 0 );
328	}
329	s.shareEditSet_prev = &s;
330	s.shareEditSet_next = &s;
331	}
332
333	static void eagerCopyCtorHelper(string_res & s, const char * rhs, size_t rhslnth) with(s) {
334	if( ambient_string_sharectx->activeHeap ) {
335	(Handle){ * ambient_string_sharectx->activeHeap };
336	(shareEditSet_owns_ulink){ false };
337	} else {
338	(Handle){ * new( rhslnth ) };
339	(shareEditSet_owns_ulink){ true };
340	}
341	Handle.s = VbyteAlloc(*Handle.ulink, rhslnth);
342	Handle.lnth = rhslnth;
343	memmove( Handle.s, rhs, rhslnth );
344	s.shareEditSet_prev = &s;
345	s.shareEditSet_next = &s;
346	}
347
348	// Constructor from a raw buffer and size
349	void ?{}(string_res & s, const char * rhs, size_t rhslnth) with(s) {
350	eagerCopyCtorHelper(s, rhs, rhslnth);
351	}
352
353	void ?{}( string_res & s, ssize_t rhs ) {
354	char buf[64];
355	int len;
356	snprintf( buf, sizeof(buf)-1, "%zd%n", rhs, &len );
357	( s ){ buf, len };
358	}
359	void ?{}( string_res & s, size_t rhs ) {
360	char buf[64];
361	int len;
362	snprintf( buf, sizeof(buf)-1, "%zu%n", rhs, &len );
363	( s ){ buf, len };
364	}
365	void ?{}( string_res & s, double rhs ) {
366	char buf[64];
367	int len;
368	snprintf( buf, sizeof(buf)-1, "%g%n", rhs, &len );
369	( s ){ buf, len };
370	}
371	void ?{}( string_res & s, long double rhs ) {
372	char buf[64];
373	int len;
374	snprintf( buf, sizeof(buf)-1, "%Lg%n", rhs, &len );
375	( s ){ buf, len };
376	}
377	void ?{}( string_res & s, double _Complex rhs ) {
378	char buf[64];
379	int len;
380	snprintf( buf, sizeof(buf)-1, "%g+%gi%n", creal( rhs ), cimag( rhs ), &len );
381	( s ){ buf, len };
382	}
383	void ?{}( string_res & s, long double _Complex rhs ) {
384	char buf[64];
385	int len;
386	snprintf( buf, sizeof(buf)-1, "%Lg+%Lgi%n", creall( rhs ), cimagl( rhs ), &len );
387	( s ){ buf, len };
388	}
389
390	// private ctor (not in header): use specified heap (ignore ambient) and copy chars in
391	void ?{}( string_res & s, VbyteHeap & heap, const char * rhs, size_t rhslnth ) with(s) {
392	(Handle){ heap };
393	Handle.s = VbyteAlloc(*Handle.ulink, rhslnth);
394	Handle.lnth = rhslnth;
395	(s.shareEditSet_owns_ulink){ false };
396	memmove( Handle.s, rhs, rhslnth );
397	s.shareEditSet_prev = &s;
398	s.shareEditSet_next = &s;
399	}
400
401
402	// General copy constructor
403	void ?{}(string_res & s, const string_res & s2, StrResInitMode mode, size_t start, size_t len ) {
404	size_t end = start + len;
405	verify( start <= end && end <= s2.Handle.lnth );
406
407	if (s2.Handle.ulink != ambient_string_sharectx->activeHeap && mode == COPY_VALUE) {
408	// crossing heaps (including private): copy eagerly
409	eagerCopyCtorHelper(s, s2.Handle.s + start, end - start);
410	verify(s.shareEditSet_prev == &s);
411	verify(s.shareEditSet_next == &s);
412	} else {
413	(s.Handle){};
414	s.Handle.s = s2.Handle.s + start;
415	s.Handle.lnth = end - start;
416	s.Handle.ulink = s2.Handle.ulink;
417
418	AddThisAfter(s.Handle, s2.Handle ); // insert this handle after rhs handle
419	// ^ bug? skip others at early point in string
420
421	if (mode == COPY_VALUE) {
422	verify(s2.Handle.ulink == ambient_string_sharectx->activeHeap);
423	// requested logical copy in same heap: defer copy until write
424
425	(s.shareEditSet_owns_ulink){ false };
426
427	// make s alone in its shareEditSet
428	s.shareEditSet_prev = &s;
429	s.shareEditSet_next = &s;
430	} else {
431	verify( mode == SHARE_EDITS );
432	// sharing edits with source forces same heap as source (ignore context)
433
434	(s.shareEditSet_owns_ulink){ s2.shareEditSet_owns_ulink };
435
436	// s2 is logically const but not implementation const
437	string_res & s2mod = (string_res &) s2;
438
439	// insert s after s2 on shareEditSet
440	s.shareEditSet_next = s2mod.shareEditSet_next;
441	s.shareEditSet_prev = &s2mod;
442	s.shareEditSet_next->shareEditSet_prev = &s;
443	s.shareEditSet_prev->shareEditSet_next = &s;
444	}
445	}
446	}
447
448	static void assignEditSet(string_res & s, string_res * shareEditSetStartPeer, string_res * shareEditSetEndPeer,
449	char * resultSesStart,
450	size_t resultSesLnth,
451	HandleNode * resultPadPosition, size_t bsize ) {
452
453	char * beforeBegin = shareEditSetStartPeer->Handle.s;
454	size_t beforeLen = s.Handle.s - beforeBegin;
455
456	char * afterBegin = s.Handle.s + s.Handle.lnth;
457	size_t afterLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - afterBegin;
458
459	size_t oldLnth = s.Handle.lnth;
460
461	s.Handle.s = resultSesStart + beforeLen;
462	s.Handle.lnth = bsize;
463	if (resultPadPosition)
464	MoveThisAfter( s.Handle, *resultPadPosition );
465
466	// adjust all substring string and handle locations, and check if any substring strings are outside the new base string
467	char *limit = resultSesStart + resultSesLnth;
468	for ( string_res * p = s.shareEditSet_next; p != &s; p = p->shareEditSet_next ) {
469	verify (p->Handle.s >= beforeBegin);
470	if ( p->Handle.s >= afterBegin ) {
471	verify ( p->Handle.s <= afterBegin + afterLen );
472	verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen );
473	// p starts after the edit
474	// take start and end as end-anchored
475	size_t startOffsetFromEnd = afterBegin + afterLen - p->Handle.s;
476	p->Handle.s = limit - startOffsetFromEnd;
477	// p->Handle.lnth unaffected
478	} else if ( p->Handle.s <= beforeBegin + beforeLen ) {
479	// p starts before, or at the start of, the edit
480	if ( p->Handle.s + p->Handle.lnth <= beforeBegin + beforeLen ) {
481	// p ends before the edit
482	// take end as start-anchored too
483	// p->Handle.lnth unaffected
484	} else if ( p->Handle.s + p->Handle.lnth < afterBegin ) {
485	// p ends during the edit; p does not include the last character replaced
486	// clip end of p to end at start of edit
487	p->Handle.lnth = beforeLen - ( p->Handle.s - beforeBegin );
488	} else {
489	// p ends after the edit
490	verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen );
491	// take end as end-anchored
492	// stretch-shrink p according to the edit
493	p->Handle.lnth += s.Handle.lnth;
494	p->Handle.lnth -= oldLnth;
495	}
496	// take start as start-anchored
497	size_t startOffsetFromStart = p->Handle.s - beforeBegin;
498	p->Handle.s = resultSesStart + startOffsetFromStart;
499	} else {
500	verify ( p->Handle.s < afterBegin );
501	// p starts during the edit
502	verify( p->Handle.s + p->Handle.lnth >= beforeBegin + beforeLen );
503	if ( p->Handle.s + p->Handle.lnth < afterBegin ) {
504	// p ends during the edit; p does not include the last character replaced
505	// set p to empty string at start of edit
506	p->Handle.s = s.Handle.s;
507	p->Handle.lnth = 0;
508	} else {
509	// p includes the end of the edit
510	// clip start of p to start at end of edit
511	int charsToClip = afterBegin - p->Handle.s;
512	p->Handle.s = s.Handle.s + s.Handle.lnth;
513	p->Handle.lnth -= charsToClip;
514	}
515	}
516	if (resultPadPosition)
517	MoveThisAfter( p->Handle, *resultPadPosition ); // move substring handle to maintain sorted order by string position
518	}
519	}
520
521	// traverse the share-edit set (SES) to recover the range of a base string to which `s` belongs
522	static void locateInShareEditSet( string_res & s, string_res & shareEditSetStartPeer, string_res & shareEditSetEndPeer ) {
523	shareEditSetStartPeer = & s;
524	shareEditSetEndPeer = & s;
525	for (string_res * editPeer = s.shareEditSet_next; editPeer != &s; editPeer = editPeer->shareEditSet_next) {
526	if ( editPeer->Handle.s < shareEditSetStartPeer->Handle.s ) {
527	shareEditSetStartPeer = editPeer;
528	}
529	if ( shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth < editPeer->Handle.s + editPeer->Handle.lnth) {
530	shareEditSetEndPeer = editPeer;
531	}
532	}
533	}
534
535	static string_res & assign_(string_res & s, const char * buffer, size_t bsize, const string_res & valSrc) {
536	string_res * shareEditSetStartPeer;
537	string_res * shareEditSetEndPeer;
538	locateInShareEditSet( s, shareEditSetStartPeer, shareEditSetEndPeer );
539
540	verify( shareEditSetEndPeer->Handle.s >= shareEditSetStartPeer->Handle.s );
541	size_t origEditSetLength = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - shareEditSetStartPeer->Handle.s;
542	verify( origEditSetLength >= s.Handle.lnth );
543
544	if ( s.shareEditSet_owns_ulink ) { // assigning to private context
545	// ok to overwrite old value within LHS
546	char * prefixStartOrig = shareEditSetStartPeer->Handle.s;
547	int prefixLen = s.Handle.s - prefixStartOrig;
548	char * suffixStartOrig = s.Handle.s + s.Handle.lnth;
549	int suffixLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - suffixStartOrig;
550
551	int delta = bsize - s.Handle.lnth;
552	if ( char * oldBytes = VbyteTryAdjustLast( *s.Handle.ulink, delta ) ) {
553	// growing: copy from old to new
554	char * dest = VbyteAlloc( *s.Handle.ulink, origEditSetLength + delta );
555	char *destCursor = dest; memcpy(destCursor, prefixStartOrig, prefixLen);
556	destCursor += prefixLen; memcpy(destCursor, buffer , bsize );
557	destCursor += bsize; memcpy(destCursor, suffixStartOrig, suffixLen);
558	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
559	dest,
560	origEditSetLength + delta,
561	0p, bsize);
562	free( oldBytes );
563	} else {
564	// room is already allocated in-place: bubble suffix and overwite middle
565	memmove( suffixStartOrig + delta, suffixStartOrig, suffixLen );
566	memcpy( s.Handle.s, buffer, bsize );
567
568	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
569	shareEditSetStartPeer->Handle.s,
570	origEditSetLength + delta,
571	0p, bsize);
572	}
573
574	} else if ( // assigning to shared context
575	s.Handle.lnth == origEditSetLength && // overwriting entire run of SES
576	& valSrc && // sourcing from a managed string
577	valSrc.Handle.ulink == s.Handle.ulink ) { // sourcing from same heap
578
579	// SES's result will only use characters from the source string => reuse source
580	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
581	valSrc.Handle.s,
582	valSrc.Handle.lnth,
583	&((string_res&)valSrc).Handle, bsize);
584
585	} else {
586	// overwriting a proper substring of some string: mash characters from old and new together (copy on write)
587	// OR we are importing characters: need to copy eagerly (can't refer to source)
588
589	// full string is from start of shareEditSetStartPeer thru end of shareEditSetEndPeer
590	// `s` occurs in the middle of it, to be replaced
591	// build up the new text in `pasting`
592
593	string_res pasting = {
594	* s.Handle.ulink, // maintain same heap, regardless of context
595	shareEditSetStartPeer->Handle.s, // start of SES
596	s.Handle.s - shareEditSetStartPeer->Handle.s }; // length of SES, before s
597	append( pasting,
598	buffer, // start of replacement for s
599	bsize ); // length of replacement for s
600	append( pasting,
601	s.Handle.s + s.Handle.lnth, // start of SES after s
602	shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth -
603	(s.Handle.s + s.Handle.lnth) ); // length of SES, after s
604
605	// The above string building can trigger compaction.
606	// The reference points (that are arguments of the string building) may move during that building.
607	// From s point on, they are stable.
608
609	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
610	pasting.Handle.s,
611	pasting.Handle.lnth,
612	&pasting.Handle, bsize);
613	}
614
615	return s;
616	}
617
618	string_res & assign(string_res & s, const string_res & src, size_t maxlen) {
619	return assign_(s, src.Handle.s, min(src.Handle.lnth, maxlen), *0p);
620	}
621
622	string_res & assign(string_res & s, const char * buffer, size_t bsize) {
623	return assign_(s, buffer, bsize, *0p);
624	}
625
626	string_res & ?=?(string_res & s, char c) {
627	return assign(s, &c, 1);
628	}
629
630	string_res & ?=?( string_res & s, ssize_t rhs ) {
631	string_res rhs2 = rhs;
632	s = rhs2;
633	return s;
634	}
635	string_res & ?=?( string_res & s, size_t rhs ) {
636	string_res rhs2 = rhs;
637	s = rhs2;
638	return s;
639	}
640	string_res & ?=?( string_res & s, double rhs ) {
641	string_res rhs2 = rhs;
642	s = rhs2;
643	return s;
644	}
645	string_res & ?=?( string_res & s, long double rhs ) {
646	string_res rhs2 = rhs;
647	s = rhs2;
648	return s;
649	}
650	string_res & ?=?( string_res & s, double _Complex rhs ) {
651	string_res rhs2 = rhs;
652	s = rhs2;
653	return s;
654	}
655	string_res & ?=?( string_res & s, long double _Complex rhs ) {
656	string_res rhs2 = rhs;
657	s = rhs2;
658	return s;
659	}
660
661	// Copy assignment operator
662	string_res & ?=?(string_res & s, const string_res & rhs) with( s ) {
663	return assign_(s, rhs.Handle.s, rhs.Handle.lnth, rhs);
664	}
665
666	string_res & ?=?(string_res & s, string_res & rhs) with( s ) {
667	const string_res & rhs2 = rhs;
668	return s = rhs2;
669	}
670
671
672	// Destructor
673	void ^?{}(string_res & s) with(s) {
674	// much delegated to implied ^VbyteSM
675
676	// sever s from its share-edit peers, if any (four no-ops when already solo)
677	s.shareEditSet_prev->shareEditSet_next = s.shareEditSet_next;
678	s.shareEditSet_next->shareEditSet_prev = s.shareEditSet_prev;
679	// s.shareEditSet_next = &s;
680	// s.shareEditSet_prev = &s;
681
682	if (shareEditSet_owns_ulink && s.shareEditSet_next == &s) { // last one out
683	delete( s.Handle.ulink );
684	}
685	}
686
687
688	// Returns the character at the given index
689	// With unicode support, this may be different from just the byte at the given
690	// offset from the start of the string.
691	char ?[?](const string_res & s, size_t index) with(s) {
692	//TODO: Check if index is valid (no exceptions yet)
693	return Handle.s[index];
694	}
695
696	void assignAt(const string_res & s, size_t index, char val) {
697	// caution: not tested (not reachable by string-api-coverage interface)
698	// equivalent form at string level is `s[index] = val`,
699	// which uses the overload that returns a length-1 string
700	string_res editZone = { s, SHARE_EDITS, index, 1 };
701	assign(editZone, &val, 1);
702	}
703
704
705	///////////////////////////////////////////////////////////////////
706	// Concatenation
707
708	void append(string_res & str1, const char * buffer, size_t bsize) {
709	size_t clnth = str1.Handle.lnth + bsize;
710	if ( str1.Handle.s + str1.Handle.lnth == buffer ) { // already juxtapose ?
711	// no-op
712	} else { // must copy some text
713	if ( str1.Handle.s + str1.Handle.lnth == VbyteAlloc(*str1.Handle.ulink, 0) ) { // str1 at end of string area ?
714	VbyteAlloc( *str1.Handle.ulink, bsize ); // create room for 2nd part at the end of string area
715	} else { // copy the two parts
716	char * str1newBuf = VbyteAlloc( *str1.Handle.ulink, clnth );
717	char * str1oldBuf = str1.Handle.s; // must read after VbyteAlloc call in case it gs's
718	str1.Handle.s = str1newBuf;
719	memcpy( str1.Handle.s, str1oldBuf, str1.Handle.lnth );
720	} // if
721	memcpy( str1.Handle.s + str1.Handle.lnth, buffer, bsize );
722	} // if
723	str1.Handle.lnth = clnth;
724	}
725
726	void ?+=?(string_res & str1, const string_res & str2) {
727	append( str1, str2.Handle.s, str2.Handle.lnth );
728	}
729
730	void append(string_res & str1, const string_res & str2, size_t maxlen) {
731	append( str1, str2.Handle.s, min(str2.Handle.lnth, maxlen) );
732	}
733
734	void ?+=?(string_res & s, char c) {
735	append( s, & c, 1 );
736	}
737	void ?+=?(string_res & s, const char * c) {
738	append( s, c, strlen(c) );
739	}
740
741	///////////////////////////////////////////////////////////////////
742	// Repetition
743
744	void ?*=?(string_res & s, size_t factor) {
745	string_res s2 = { s, COPY_VALUE };
746	s = "";
747	for (factor) s += s2;
748	}
749
750	//////////////////////////////////////////////////////////
751	// Comparisons
752
753	int strcmp(const string_res & s1, const string_res & s2) {
754	// return 0;
755	int ans1 = memcmp(s1.Handle.s, s2.Handle.s, min(s1.Handle.lnth, s2.Handle.lnth));
756	if (ans1 != 0) return ans1;
757	return s1.Handle.lnth - s2.Handle.lnth;
758	}
759
760	bool ?==?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) == 0; }
761	bool ?!=?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) != 0; }
762	bool ?>? (const string_res & s1, const string_res & s2) { return strcmp(s1, s2) > 0; }
763	bool ?>=?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) >= 0; }
764	bool ?<=?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) <= 0; }
765	bool ?<? (const string_res & s1, const string_res & s2) { return strcmp(s1, s2) < 0; }
766
767	int strcmp (const string_res & s1, const char * s2) {
768	string_res s2x = s2;
769	return strcmp(s1, s2x);
770	}
771
772	bool ?==?(const string_res & s1, const char * s2) { return strcmp(s1, s2) == 0; }
773	bool ?!=?(const string_res & s1, const char * s2) { return strcmp(s1, s2) != 0; }
774	bool ?>? (const string_res & s1, const char * s2) { return strcmp(s1, s2) > 0; }
775	bool ?>=?(const string_res & s1, const char * s2) { return strcmp(s1, s2) >= 0; }
776	bool ?<=?(const string_res & s1, const char * s2) { return strcmp(s1, s2) <= 0; }
777	bool ?<? (const string_res & s1, const char * s2) { return strcmp(s1, s2) < 0; }
778
779	int strcmp (const char * s1, const string_res & s2) {
780	string_res s1x = s1;
781	return strcmp(s1x, s2);
782	}
783
784	bool ?==?(const char * s1, const string_res & s2) { return strcmp(s1, s2) == 0; }
785	bool ?!=?(const char * s1, const string_res & s2) { return strcmp(s1, s2) != 0; }
786	bool ?>? (const char * s1, const string_res & s2) { return strcmp(s1, s2) > 0; }
787	bool ?>=?(const char * s1, const string_res & s2) { return strcmp(s1, s2) >= 0; }
788	bool ?<=?(const char * s1, const string_res & s2) { return strcmp(s1, s2) <= 0; }
789	bool ?<? (const char * s1, const string_res & s2) { return strcmp(s1, s2) < 0; }
790
791
792	//////////////////////////////////////////////////////////
793	// Search
794
795	bool contains(const string_res & s, char ch) {
796	for ( i; size(s) ) {
797	if (s[i] == ch) return true;
798	}
799	return false;
800	}
801
802	int find(const string_res & s, char search) {
803	return findFrom(s, 0, search);
804	}
805
806	int findFrom(const string_res & s, size_t fromPos, char search) {
807	// FIXME: This paricular overload (find of single char) is optimized to use memchr.
808	// The general overload (find of string, memchr applying to its first character) and `contains` should be adjusted to match.
809	char * searchFrom = s.Handle.s + fromPos;
810	size_t searchLnth = s.Handle.lnth - fromPos;
811	int searchVal = search;
812	char * foundAt = (char *) memchr(searchFrom, searchVal, searchLnth);
813	if (foundAt == 0p) return s.Handle.lnth;
814	else return foundAt - s.Handle.s;
815	}
816
817	int find(const string_res & s, const string_res & search) {
818	return findFrom(s, 0, search);
819	}
820
821	int findFrom(const string_res & s, size_t fromPos, const string_res & search) {
822	return findFrom(s, fromPos, search.Handle.s, search.Handle.lnth);
823	}
824
825	int find(const string_res & s, const char * search) {
826	return findFrom(s, 0, search);
827	}
828	int findFrom(const string_res & s, size_t fromPos, const char * search) {
829	return findFrom(s, fromPos, search, strlen(search));
830	}
831
832	int find(const string_res & s, const char * search, size_t searchsize) {
833	return findFrom(s, 0, search, searchsize);
834	}
835
836	int findFrom(const string_res & s, size_t fromPos, const char * search, size_t searchsize) {
837	/* Remaining implementations essentially ported from Sunjay's work */
838
839	// FIXME: This is a naive algorithm. We probably want to switch to someting
840	// like Boyer-Moore in the future.
841	// https://en.wikipedia.org/wiki/String_searching_algorithm
842
843	// Always find the empty string
844	if (searchsize == 0) {
845	return 0;
846	}
847
848	for ( i; fromPos ~ s.Handle.lnth ) {
849	size_t remaining = s.Handle.lnth - i;
850	// Never going to find the search string if the remaining string is
851	// smaller than search
852	if (remaining < searchsize) {
853	break;
854	}
855
856	bool matched = true;
857	for ( j; searchsize ) {
858	if (search[j] != s.Handle.s[i + j]) {
859	matched = false;
860	break;
861	}
862	}
863	if (matched) {
864	return i;
865	}
866	}
867	return s.Handle.lnth;
868	}
869
870	bool includes(const string_res & s, const string_res & search) {
871	return includes(s, search.Handle.s, search.Handle.lnth);
872	}
873
874	bool includes(const string_res & s, const char * search) {
875	return includes(s, search, strlen(search));
876	}
877
878	bool includes(const string_res & s, const char * search, size_t searchsize) {
879	return find(s, search, searchsize) < s.Handle.lnth;
880	}
881
882	bool startsWith(const string_res & s, const string_res & prefix) {
883	return startsWith(s, prefix.Handle.s, prefix.Handle.lnth);
884	}
885
886	bool startsWith(const string_res & s, const char * prefix) {
887	return startsWith(s, prefix, strlen(prefix));
888	}
889
890	bool startsWith(const string_res & s, const char * prefix, size_t prefixsize) {
891	if (s.Handle.lnth < prefixsize) {
892	return false;
893	}
894	return memcmp(s.Handle.s, prefix, prefixsize) == 0;
895	}
896
897	bool endsWith(const string_res & s, const string_res & suffix) {
898	return endsWith(s, suffix.Handle.s, suffix.Handle.lnth);
899	}
900
901	bool endsWith(const string_res & s, const char * suffix) {
902	return endsWith(s, suffix, strlen(suffix));
903	}
904
905	bool endsWith(const string_res & s, const char * suffix, size_t suffixsize) {
906	if (s.Handle.lnth < suffixsize) {
907	return false;
908	}
909	// Amount to offset the bytes pointer so that we are comparing the end of s
910	// to suffix. s.bytes + offset should be the first byte to compare against suffix
911	size_t offset = s.Handle.lnth - suffixsize;
912	return memcmp(s.Handle.s + offset, suffix, suffixsize) == 0;
913	}
914
915	/* Back to Mike's work */
916
917	///////////////////////////////////////////////////////////////////////////
918	// charclass, include, exclude
919
920	void ?{}( charclass_res & s, const string_res & chars) {
921	(s){ chars.Handle.s, chars.Handle.lnth };
922	}
923
924	void ?{}( charclass_res & s, const char * chars ) {
925	(s){ chars, strlen(chars) };
926	}
927
928	void ?{}( charclass_res & s, const char * chars, size_t charssize ) {
929	(s.chars){ chars, charssize };
930	// now sort it ?
931	}
932
933	void ^?{}( charclass_res & s ) {
934	^(s.chars){};
935	}
936
937	static bool test( const charclass_res & mask, char c ) {
938	// instead, use sorted char list?
939	return contains( mask.chars, c );
940	}
941
942	int exclude(const string_res & s, const charclass_res & mask) {
943	for ( i; size(s) ) {
944	if ( test(mask, s[i]) ) return i;
945	}
946	return size(s);
947	}
948
949	int include(const string_res & s, const charclass_res & mask) {
950	for ( i; size(s) ) {
951	if ( ! test(mask, s[i]) ) return i;
952	}
953	return size(s);
954	}
955
956	//######################### VbyteHeap "implementation" #########################
957
958
959	// Add a new HandleNode node n after the current HandleNode node.
960
961	static void AddThisAfter( HandleNode & s, HandleNode & n ) with(s) {
962	#ifdef VbyteDebug
963	serr \| "enter:AddThisAfter, s:" \| &s \| " n:" \| &n;
964	#endif // VbyteDebug
965	// 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).
966	verify( n.ulink != 0p );
967	verify( s.ulink == n.ulink );
968	flink = n.flink;
969	blink = &n;
970	n.flink->blink = &s;
971	n.flink = &s;
972	#ifdef VbyteDebug
973	{
974	serr \| "HandleList:";
975	serr \| nlOff;
976	for ( HandleNode *ni = HeaderPtr->flink; ni != HeaderPtr; ni = ni->flink ) {
977	serr \| "\tnode:" \| ni \| " lnth:" \| ni->lnth \| " s:" \| (void *)ni->s \| ",\"";
978	for ( i; ni->lnth ) {
979	serr \| ni->s[i];
980	} // for
981	serr \| "\" flink:" \| ni->flink \| " blink:" \| ni->blink \| nl;
982	} // for
983	serr \| nlOn;
984	}
985	serr \| "exit:AddThisAfter";
986	#endif // VbyteDebug
987	} // AddThisAfter
988
989
990	// Delete the current HandleNode node.
991
992	static void DeleteNode( HandleNode & s ) with(s) {
993	#ifdef VbyteDebug
994	serr \| "enter:DeleteNode, s:" \| &s;
995	#endif // VbyteDebug
996	flink->blink = blink;
997	blink->flink = flink;
998	#ifdef VbyteDebug
999	serr \| "exit:DeleteNode";
1000	#endif // VbyteDebug
1001	} // DeleteNode
1002
1003
1004	// Allocates specified storage for a string from byte-string area. If not enough space remains to perform the
1005	// allocation, the garbage collection routine is called.
1006
1007	static char * VbyteAlloc( VbyteHeap & s, int size ) with(s) {
1008	#ifdef VbyteDebug
1009	serr \| "enter:VbyteAlloc, size:" \| size;
1010	#endif // VbyteDebug
1011	uintptr_t NoBytes;
1012	char *r;
1013
1014	NoBytes = ( uintptr_t )EndVbyte + size;
1015	if ( NoBytes > ( uintptr_t )ExtVbyte ) { // enough room for new byte-string ?
1016	garbage( s, size ); // firer up the garbage collector
1017	verify( (( uintptr_t )EndVbyte + size) <= ( uintptr_t )ExtVbyte && "garbage run did not free up required space" );
1018	} // if
1019	r = EndVbyte;
1020	EndVbyte += size;
1021	#ifdef VbyteDebug
1022	serr \| "exit:VbyteAlloc, r:" \| (void )r \| " EndVbyte:" \| (void )EndVbyte \| " ExtVbyte:" \| ExtVbyte;
1023	#endif // VbyteDebug
1024	return r;
1025	} // VbyteAlloc
1026
1027
1028	// Adjusts the last allocation in this heap by delta bytes, or resets this heap to be able to offer
1029	// new allocations of its original size + delta bytes. Positive delta means bigger;
1030	// negative means smaller. A null return indicates that the original heap location has room for
1031	// the requested growth. A non-null return indicates that copying to a new location is required
1032	// but has not been done; the returned value is the old heap storage location; `this` heap is
1033	// modified to reference the new location. In the copy-requred case, the caller should use
1034	// VbyteAlloc to claim the new space, while doing optimal copying from old to new, then free old.
1035
1036	static char * VbyteTryAdjustLast( VbyteHeap & s, int delta ) with(s) {
1037	if ( ( uintptr_t )EndVbyte + delta <= ( uintptr_t )ExtVbyte ) {
1038	// room available
1039	EndVbyte += delta;
1040	return 0p;
1041	}
1042
1043	char *oldBytes = StartVbyte;
1044
1045	NoOfExtensions += 1;
1046	CurrSize *= 2;
1047	StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize);
1048	ExtVbyte = StartVbyte + CurrSize;
1049
1050	return oldBytes;
1051	}
1052
1053
1054	// Move an existing HandleNode node h somewhere after the current HandleNode node so that it is in ascending order by
1055	// the address in the byte string area.
1056
1057	static void MoveThisAfter( HandleNode & s, const HandleNode & h ) with(s) {
1058	#ifdef VbyteDebug
1059	serr \| "enter:MoveThisAfter, s:" \| & s \| " h:" \| & h;
1060	#endif // VbyteDebug
1061	verify( h.ulink != 0p );
1062	verify( s.ulink == h.ulink );
1063	if ( s < h.s ) { // check argument values
1064	// serr \| "VbyteSM: Error - Cannot move byte string starting at:" \| s \| " after byte string starting at:"
1065	// \| ( h->s ) \| " and keep handles in ascending order";
1066	// exit(-1 );
1067	verify( 0 && "VbyteSM: Error - Cannot move byte strings as requested and keep handles in ascending order");
1068	} // if
1069
1070	HandleNode *i;
1071	for ( i = h.flink; i->s != 0 && s > ( i->s ); i = i->flink ); // find the position for this node after h
1072	if ( & s != i->blink ) {
1073	DeleteNode( s );
1074	AddThisAfter( s, *i->blink );
1075	} // if
1076	#ifdef VbyteDebug
1077	{
1078	serr \| "HandleList:";
1079	serr \| nlOff;
1080	for ( HandleNode *n = HeaderPtr->flink; n != HeaderPtr; n = n->flink ) {
1081	serr \| "\tnode:" \| n \| " lnth:" \| n->lnth \| " s:" \| (void *)n->s \| ",\"";
1082	for ( i; n->lnth ) {
1083	serr \| n->s[i];
1084	} // for
1085	serr \| "\" flink:" \| n->flink \| " blink:" \| n->blink \| nl;
1086	} // for
1087	serr \| nlOn;
1088	}
1089	serr \| "exit:MoveThisAfter";
1090	#endif // VbyteDebug
1091	} // MoveThisAfter
1092
1093
1094	//######################### VbyteHeap #########################
1095
1096	// Compare two byte strings in the byte-string area. The routine returns the following values:
1097	//
1098	// 1 => Src1-byte-string > Src2-byte-string
1099	// 0 => Src1-byte-string = Src2-byte-string
1100	// -1 => Src1-byte-string < Src2-byte-string
1101
1102	int ByteCmp( char Src1, int Src1Start, int Src1Lnth, char Src2, int Src2Start, int Src2Lnth ) {
1103	#ifdef VbyteDebug
1104	serr \| "enter:ByteCmp, Src1Start:" \| Src1Start \| " Src1Lnth:" \| Src1Lnth \| " Src2Start:" \| Src2Start \| " Src2Lnth:" \| Src2Lnth;
1105	#endif // VbyteDebug
1106	int cmp;
1107
1108	CharZip: for ( int i = 0; ; i += 1 ) {
1109	if ( i == Src2Lnth - 1 ) {
1110	for ( ; ; i += 1 ) {
1111	if ( i == Src1Lnth - 1 ) {
1112	cmp = 0;
1113	break CharZip;
1114	} // exit
1115	if ( Src1[Src1Start + i] != ' ') {
1116	// SUSPECTED BUG: this could be be why Peter got the bug report about == " " (why is this case here at all?)
1117	cmp = 1;
1118	break CharZip;
1119	} // exit
1120	} // for
1121	} // exit
1122	if ( i == Src1Lnth - 1 ) {
1123	for ( ; ; i += 1 ) {
1124	if ( i == Src2Lnth - 1 ) {
1125	cmp = 0;
1126	break CharZip;
1127	} // exit
1128	if ( Src2[Src2Start + i] != ' ') {
1129	cmp = -1;
1130	break CharZip;
1131	} // exit
1132	} // for
1133	} // exit
1134	if ( Src2[Src2Start + i] != Src1[Src1Start+ i]) {
1135	cmp = Src1[Src1Start + i] > Src2[Src2Start + i] ? 1 : -1;
1136	break CharZip;
1137	} // exit
1138	} // for
1139	#ifdef VbyteDebug
1140	serr \| "exit:ByteCmp, cmp:" \| cmp;
1141	#endif // VbyteDebug
1142	return cmp;
1143	} // ByteCmp
1144
1145
1146	// The compaction moves all of the byte strings currently in use to the beginning of the byte-string area and modifies
1147	// the handles to reflect the new positions of the byte strings. Compaction assumes that the handle list is in ascending
1148	// order by pointers into the byte-string area. The strings associated with substrings do not have to be moved because
1149	// the containing string has been moved. Hence, they only require that their string pointers be adjusted.
1150
1151	void compaction(VbyteHeap & s) with(s) {
1152	HandleNode *h;
1153	char obase, nbase, *limit;
1154
1155	NoOfCompactions += 1;
1156	EndVbyte = StartVbyte;
1157	h = Header.flink; // ignore header node
1158	for () {
1159	memmove( EndVbyte, h->s, h->lnth );
1160	obase = h->s;
1161	h->s = EndVbyte;
1162	nbase = h->s;
1163	EndVbyte += h->lnth;
1164	limit = obase + h->lnth;
1165	h = h->flink;
1166
1167	// check if any substrings are allocated within a string
1168
1169	for () {
1170	if ( h == &Header ) break; // end of header list ?
1171	if ( h->s >= limit ) break; // outside of current string ?
1172	h->s = nbase + (( uintptr_t )h->s - ( uintptr_t )obase );
1173	h = h->flink;
1174	} // for
1175	if ( h == &Header ) break; // end of header list ?
1176	} // for
1177	} // compaction
1178
1179
1180	static double heap_expansion_freespace_threshold = 0.1; // default inherited from prior work: expand heap when less than 10% "free" (i.e. garbage)
1181	// probably an unreasonable default, but need to assess early-round tests on changing it
1182
1183	void TUNING_set_string_heap_liveness_threshold( double val ) {
1184	heap_expansion_freespace_threshold = 1.0 - val;
1185	}
1186
1187
1188	// Garbage determines the amount of free space left in the heap and then reduces, leave the same, or extends the size of
1189	// the heap. The heap is then compacted in the existing heap or into the newly allocated heap.
1190
1191	void garbage(VbyteHeap & s, int minreq ) with(s) {
1192	#ifdef VbyteDebug
1193	serr \| "enter:garbage";
1194	{
1195	serr \| "HandleList:";
1196	for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) {
1197	serr \| nlOff;
1198	serr \| "\tnode:" \| n \| " lnth:" \| n->lnth \| " s:" \| (void *)n->s \| ",\"";
1199	for ( i; n->lnth ) {
1200	serr \| n->s[i];
1201	} // for
1202	serr \| nlOn;
1203	serr \| "\" flink:" \| n->flink \| " blink:" \| n->blink;
1204	} // for
1205	}
1206	#endif // VbyteDebug
1207	int AmountUsed, AmountFree;
1208
1209	AmountUsed = 0;
1210	for ( HandleNode *i = Header.flink; i != &Header; i = i->flink ) { // calculate amount of byte area used
1211	AmountUsed += i->lnth;
1212	} // for
1213	AmountFree = ( uintptr_t )ExtVbyte - ( uintptr_t )StartVbyte - AmountUsed;
1214
1215	if ( ( double ) AmountFree < ( CurrSize * heap_expansion_freespace_threshold ) \|\| AmountFree < minreq ) { // free space less than threshold or not enough to serve cur request
1216
1217	extend( s, max( CurrSize, minreq ) ); // extend the heap
1218
1219	// Peter says, "This needs work before it should be used."
1220	// } else if ( AmountFree > CurrSize / 2 ) { // free space greater than 3 times the initial allocation ?
1221	// reduce(( AmountFree / CurrSize - 3 ) * CurrSize ); // reduce the memory
1222
1223	// `extend` implies a `compaction` during the copy
1224
1225	} else {
1226	compaction(s); // in-place
1227	}// if
1228	#ifdef VbyteDebug
1229	{
1230	serr \| "HandleList:";
1231	for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) {
1232	serr \| nlOff;
1233	serr \| "\tnode:" \| n \| " lnth:" \| n->lnth \| " s:" \| (void *)n->s \| ",\"";
1234	for ( i; n->lnth ) {
1235	serr \| n->s[i];
1236	} // for
1237	serr \| nlOn;
1238	serr \| "\" flink:" \| n->flink \| " blink:" \| n->blink;
1239	} // for
1240	}
1241	serr \| "exit:garbage";
1242	#endif // VbyteDebug
1243	} // garbage
1244
1245	#undef VbyteDebug
1246
1247
1248
1249	// Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string
1250	// area is deleted.
1251
1252	void extend( VbyteHeap & s, int size ) with (s) {
1253	#ifdef VbyteDebug
1254	serr \| "enter:extend, size:" \| size;
1255	#endif // VbyteDebug
1256	char *OldStartVbyte;
1257
1258	NoOfExtensions += 1;
1259	OldStartVbyte = StartVbyte; // save previous byte area
1260
1261	CurrSize += size > InitSize ? size : InitSize; // minimum extension, initial size
1262	StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize);
1263	ExtVbyte = (void *)( StartVbyte + CurrSize );
1264	compaction(s); // copy from old heap to new & adjust pointers to new heap
1265	free( OldStartVbyte ); // release old heap
1266	#ifdef VbyteDebug
1267	serr \| "exit:extend, CurrSize:" \| CurrSize;
1268	#endif // VbyteDebug
1269	} // extend
1270
1271	//WIP
1272	#if 0
1273
1274	// Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string
1275	// area is deleted.
1276
1277	void VbyteHeap::reduce( int size ) {
1278	#ifdef VbyteDebug
1279	serr \| "enter:reduce, size:" \| size;
1280	#endif // VbyteDebug
1281	char *OldStartVbyte;
1282
1283	NoOfReductions += 1;
1284	OldStartVbyte = StartVbyte; // save previous byte area
1285
1286	CurrSize -= size;
1287	StartVbyte = EndVbyte = new char[CurrSize];
1288	ExtVbyte = (void *)( StartVbyte + CurrSize );
1289	compaction(); // copy from old heap to new & adjust pointers to new heap
1290	delete OldStartVbyte; // release old heap
1291	#ifdef VbyteDebug
1292	!serr \| "exit:reduce, CurrSize:" \| CurrSize;
1293	#endif // VbyteDebug
1294	} // reduce
1295
1296
1297	#endif

Note: See TracBrowser for help on using the repository browser.

Download in other formats:

Original Format