Context Navigation

string_res.cfa @ c0d8e62

Last change on this file since c0d8e62 was 714e206, checked in by Peter A. Buhr <pabuhr@…>, 5 months ago
more cleanup, changes related to detection of missing values during input
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 : Sat Feb 10 17:47:22 2024
13	// Update Count : 83
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	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	// Reading into a temp before assigning to s is near zero overhead in typical cases because of sharing.
213	// If s is a substring of something larger, simple assignment takes care of that case correctly.
214	// But directly reading a variable amount of text into the middle of a larger context is not practical.
215	string_res temp;
216
217	// Read in chunks. Often, one chunk is enough. Keep the string that accumulates chunks last in the heap,
218	// so available room is rest of heap. When a chunk fills the heap, force growth then take the next chunk.
219	for (bool cont = true; cont; ) {
220	cont = false;
221
222	// Append dummy content to temp, forcing expansion when applicable (occurs always on subsequent loops)
223	// length 2 ensures room for at least one real char, plus scanf/pipe-cstr's null terminator
224	temp += "--";
225	assert( temp.Handle.ulink->EndVbyte == temp.Handle.s + temp.Handle.lnth ); // last in heap
226
227	// reset, to overwrite the appended "--"
228	temp.Handle.lnth -= 2;
229	temp.Handle.ulink->EndVbyte -= 2;
230
231	// rest of heap is available to read into
232	int lenReadable = (char *)temp.Handle.ulink->ExtVbyte - temp.Handle.ulink->EndVbyte;
233	assert (lenReadable >= 2);
234
235	// get bytes
236	try {
237	*(temp.Handle.ulink->EndVbyte) = '\0'; // pre-assign empty cstring
238	in \| wdi( lenReadable, temp.Handle.ulink->EndVbyte );
239	} catch (cstring_length *) {
240	cont = true;
241	}
242	int lenWasRead = strlen(temp.Handle.ulink->EndVbyte);
243
244	// update metadata
245	temp.Handle.lnth += lenWasRead;
246	temp.Handle.ulink->EndVbyte += lenWasRead;
247	}
248
249	if ( temp.Handle.lnth > 0 ) s = temp;
250	return in;
251	}
252
253	ifstream & ?\|?( ifstream & is, _Istream_Rquoted f ) with( f.rstr ) {
254	if ( eof( is ) ) throwResume ExceptionInst( missing_data );
255	int args;
256	fini: {
257	char rfmt[5] = { ' ', delimiters[0], '%', 'n', '\0' };
258	int len = -1; // may not be set in fmt
259	args = fmt( is, rfmt, &len ); // remove leading whitespace and quote
260	if ( eof( is ) \|\| len == -1 ) break fini;
261
262	// Change the remainder of the read into a getline by reseting the closing delimiter.
263	if ( delimiters[1] != '\0' ) {
264	delimiters[0] = delimiters[1];
265	delimiters[1] = '\0';
266	} // if
267	flags.delimiter = true;
268	return is \| (_Istream_Rstr )&f;
269	} // fini
270	// read failed => no pattern match => set string to null
271	if ( ! flags.ignore && s != 0p && args == 0 ) s[0] = '\0';
272	if ( args == 1 && eof( is ) ) { // data but scan ended at EOF
273	clear( is ); // => reset EOF => detect again on next read
274	} // if
275	return is;
276	}
277
278	ifstream & ?\|?( ifstream & is, _Istream_Rstr f ) {
279	// .---------------,
280	// \| \| \| \| \|...\|0\|0\| null terminator and guard if missing
281	// `---------------'
282	enum { gwd = 128 + 1, wd = gwd - 1 }; // guard and unguard width
283	char cstr[gwd]; // read in chunks
284	bool cont = false;
285
286	_Istream_Cwidth cf = { cstr, (_Istream_str_base)f };
287	if ( ! cf.flags.rwd ) cf.wd = wd;
288
289	cstr[wd] = '\0'; // guard null terminate string
290	try {
291	cstr[0] = '\0'; // pre-assign as empty cstring
292	is \| cf;
293	} catch( cstring_length * ) {
294	cont = true;
295	} finally {
296	if ( ! cf.flags.ignore // ok to initialize string
297	// && cstr[0] != '\0' // something was read
298	) {
299	*(f.s) = cstr;
300	}
301	} // try
302	for ( ; cont; ) { // overflow read ?
303	cont = false;
304	try {
305	cstr[0] = '\0'; // pre-assign as empty cstring
306	is \| cf;
307	} catch( cstring_length * ) {
308	cont = true; // continue not allowed
309	} finally {
310	if ( ! cf.flags.ignore && cstr[0] != '\0' ) { // something was read
311	*(f.s) += cstr; // build string chunk at a time
312	}
313	} // try
314	} // for
315	return is;
316	} // ?\|?
317
318	// Empty constructor
319	void ?{}(string_res & s) with(s) {
320	if( ambient_string_sharectx->activeHeap ) {
321	(Handle){ * ambient_string_sharectx->activeHeap };
322	(shareEditSet_owns_ulink){ false };
323	verify( Handle.s == 0p && Handle.lnth == 0 );
324	} else {
325	(Handle){ * new( (size_t) 10 ) }; // TODO: can I lazily avoid allocating for empty string
326	(shareEditSet_owns_ulink){ true };
327	Handle.s = Handle.ulink->StartVbyte;
328	verify( Handle.lnth == 0 );
329	}
330	s.shareEditSet_prev = &s;
331	s.shareEditSet_next = &s;
332	}
333
334	static void eagerCopyCtorHelper(string_res & s, const char * rhs, size_t rhslnth) with(s) {
335	if( ambient_string_sharectx->activeHeap ) {
336	(Handle){ * ambient_string_sharectx->activeHeap };
337	(shareEditSet_owns_ulink){ false };
338	} else {
339	(Handle){ * new( rhslnth ) };
340	(shareEditSet_owns_ulink){ true };
341	}
342	Handle.s = VbyteAlloc(*Handle.ulink, rhslnth);
343	Handle.lnth = rhslnth;
344	memmove( Handle.s, rhs, rhslnth );
345	s.shareEditSet_prev = &s;
346	s.shareEditSet_next = &s;
347	}
348
349	// Constructor from a raw buffer and size
350	void ?{}(string_res & s, const char * rhs, size_t rhslnth) with(s) {
351	eagerCopyCtorHelper(s, rhs, rhslnth);
352	}
353
354	void ?{}( string_res & s, ssize_t rhs ) {
355	char buf[64];
356	int len;
357	snprintf( buf, sizeof(buf)-1, "%zd%n", rhs, &len );
358	( s ){ buf, len };
359	}
360	void ?{}( string_res & s, size_t rhs ) {
361	char buf[64];
362	int len;
363	snprintf( buf, sizeof(buf)-1, "%zu%n", rhs, &len );
364	( s ){ buf, len };
365	}
366	void ?{}( string_res & s, double rhs ) {
367	char buf[64];
368	int len;
369	snprintf( buf, sizeof(buf)-1, "%g%n", rhs, &len );
370	( s ){ buf, len };
371	}
372	void ?{}( string_res & s, long double rhs ) {
373	char buf[64];
374	int len;
375	snprintf( buf, sizeof(buf)-1, "%Lg%n", rhs, &len );
376	( s ){ buf, len };
377	}
378	void ?{}( string_res & s, double _Complex rhs ) {
379	char buf[64];
380	int len;
381	snprintf( buf, sizeof(buf)-1, "%g+%gi%n", creal( rhs ), cimag( rhs ), &len );
382	( s ){ buf, len };
383	}
384	void ?{}( string_res & s, long double _Complex rhs ) {
385	char buf[64];
386	int len;
387	snprintf( buf, sizeof(buf)-1, "%Lg+%Lgi%n", creall( rhs ), cimagl( rhs ), &len );
388	( s ){ buf, len };
389	}
390
391	// private ctor (not in header): use specified heap (ignore ambient) and copy chars in
392	void ?{}( string_res & s, VbyteHeap & heap, const char * rhs, size_t rhslnth ) with(s) {
393	(Handle){ heap };
394	Handle.s = VbyteAlloc(*Handle.ulink, rhslnth);
395	Handle.lnth = rhslnth;
396	(s.shareEditSet_owns_ulink){ false };
397	memmove( Handle.s, rhs, rhslnth );
398	s.shareEditSet_prev = &s;
399	s.shareEditSet_next = &s;
400	}
401
402
403	// General copy constructor
404	void ?{}(string_res & s, const string_res & s2, StrResInitMode mode, size_t start, size_t len ) {
405	size_t end = start + len;
406	verify( start <= end && end <= s2.Handle.lnth );
407
408	if (s2.Handle.ulink != ambient_string_sharectx->activeHeap && mode == COPY_VALUE) {
409	// crossing heaps (including private): copy eagerly
410	eagerCopyCtorHelper(s, s2.Handle.s + start, end - start);
411	verify(s.shareEditSet_prev == &s);
412	verify(s.shareEditSet_next == &s);
413	} else {
414	(s.Handle){};
415	s.Handle.s = s2.Handle.s + start;
416	s.Handle.lnth = end - start;
417	s.Handle.ulink = s2.Handle.ulink;
418
419	AddThisAfter(s.Handle, s2.Handle ); // insert this handle after rhs handle
420	// ^ bug? skip others at early point in string
421
422	if (mode == COPY_VALUE) {
423	verify(s2.Handle.ulink == ambient_string_sharectx->activeHeap);
424	// requested logical copy in same heap: defer copy until write
425
426	(s.shareEditSet_owns_ulink){ false };
427
428	// make s alone in its shareEditSet
429	s.shareEditSet_prev = &s;
430	s.shareEditSet_next = &s;
431	} else {
432	verify( mode == SHARE_EDITS );
433	// sharing edits with source forces same heap as source (ignore context)
434
435	(s.shareEditSet_owns_ulink){ s2.shareEditSet_owns_ulink };
436
437	// s2 is logically const but not implementation const
438	string_res & s2mod = (string_res &) s2;
439
440	// insert s after s2 on shareEditSet
441	s.shareEditSet_next = s2mod.shareEditSet_next;
442	s.shareEditSet_prev = &s2mod;
443	s.shareEditSet_next->shareEditSet_prev = &s;
444	s.shareEditSet_prev->shareEditSet_next = &s;
445	}
446	}
447	}
448
449	static void assignEditSet(string_res & s, string_res * shareEditSetStartPeer, string_res * shareEditSetEndPeer,
450	char * resultSesStart,
451	size_t resultSesLnth,
452	HandleNode * resultPadPosition, size_t bsize ) {
453
454	char * beforeBegin = shareEditSetStartPeer->Handle.s;
455	size_t beforeLen = s.Handle.s - beforeBegin;
456
457	char * afterBegin = s.Handle.s + s.Handle.lnth;
458	size_t afterLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - afterBegin;
459
460	size_t oldLnth = s.Handle.lnth;
461
462	s.Handle.s = resultSesStart + beforeLen;
463	s.Handle.lnth = bsize;
464	if (resultPadPosition)
465	MoveThisAfter( s.Handle, *resultPadPosition );
466
467	// adjust all substring string and handle locations, and check if any substring strings are outside the new base string
468	char *limit = resultSesStart + resultSesLnth;
469	for ( string_res * p = s.shareEditSet_next; p != &s; p = p->shareEditSet_next ) {
470	verify (p->Handle.s >= beforeBegin);
471	if ( p->Handle.s >= afterBegin ) {
472	verify ( p->Handle.s <= afterBegin + afterLen );
473	verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen );
474	// p starts after the edit
475	// take start and end as end-anchored
476	size_t startOffsetFromEnd = afterBegin + afterLen - p->Handle.s;
477	p->Handle.s = limit - startOffsetFromEnd;
478	// p->Handle.lnth unaffected
479	} else if ( p->Handle.s <= beforeBegin + beforeLen ) {
480	// p starts before, or at the start of, the edit
481	if ( p->Handle.s + p->Handle.lnth <= beforeBegin + beforeLen ) {
482	// p ends before the edit
483	// take end as start-anchored too
484	// p->Handle.lnth unaffected
485	} else if ( p->Handle.s + p->Handle.lnth < afterBegin ) {
486	// p ends during the edit; p does not include the last character replaced
487	// clip end of p to end at start of edit
488	p->Handle.lnth = beforeLen - ( p->Handle.s - beforeBegin );
489	} else {
490	// p ends after the edit
491	verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen );
492	// take end as end-anchored
493	// stretch-shrink p according to the edit
494	p->Handle.lnth += s.Handle.lnth;
495	p->Handle.lnth -= oldLnth;
496	}
497	// take start as start-anchored
498	size_t startOffsetFromStart = p->Handle.s - beforeBegin;
499	p->Handle.s = resultSesStart + startOffsetFromStart;
500	} else {
501	verify ( p->Handle.s < afterBegin );
502	// p starts during the edit
503	verify( p->Handle.s + p->Handle.lnth >= beforeBegin + beforeLen );
504	if ( p->Handle.s + p->Handle.lnth < afterBegin ) {
505	// p ends during the edit; p does not include the last character replaced
506	// set p to empty string at start of edit
507	p->Handle.s = s.Handle.s;
508	p->Handle.lnth = 0;
509	} else {
510	// p includes the end of the edit
511	// clip start of p to start at end of edit
512	int charsToClip = afterBegin - p->Handle.s;
513	p->Handle.s = s.Handle.s + s.Handle.lnth;
514	p->Handle.lnth -= charsToClip;
515	}
516	}
517	if (resultPadPosition)
518	MoveThisAfter( p->Handle, *resultPadPosition ); // move substring handle to maintain sorted order by string position
519	}
520	}
521
522	// traverse the share-edit set (SES) to recover the range of a base string to which `s` belongs
523	static void locateInShareEditSet( string_res & s, string_res & shareEditSetStartPeer, string_res & shareEditSetEndPeer ) {
524	shareEditSetStartPeer = & s;
525	shareEditSetEndPeer = & s;
526	for (string_res * editPeer = s.shareEditSet_next; editPeer != &s; editPeer = editPeer->shareEditSet_next) {
527	if ( editPeer->Handle.s < shareEditSetStartPeer->Handle.s ) {
528	shareEditSetStartPeer = editPeer;
529	}
530	if ( shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth < editPeer->Handle.s + editPeer->Handle.lnth) {
531	shareEditSetEndPeer = editPeer;
532	}
533	}
534	}
535
536	static string_res & assign_(string_res & s, const char * buffer, size_t bsize, const string_res & valSrc) {
537	string_res * shareEditSetStartPeer;
538	string_res * shareEditSetEndPeer;
539	locateInShareEditSet( s, shareEditSetStartPeer, shareEditSetEndPeer );
540
541	verify( shareEditSetEndPeer->Handle.s >= shareEditSetStartPeer->Handle.s );
542	size_t origEditSetLength = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - shareEditSetStartPeer->Handle.s;
543	verify( origEditSetLength >= s.Handle.lnth );
544
545	if ( s.shareEditSet_owns_ulink ) { // assigning to private context
546	// ok to overwrite old value within LHS
547	char * prefixStartOrig = shareEditSetStartPeer->Handle.s;
548	int prefixLen = s.Handle.s - prefixStartOrig;
549	char * suffixStartOrig = s.Handle.s + s.Handle.lnth;
550	int suffixLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - suffixStartOrig;
551
552	int delta = bsize - s.Handle.lnth;
553	if ( char * oldBytes = VbyteTryAdjustLast( *s.Handle.ulink, delta ) ) {
554	// growing: copy from old to new
555	char * dest = VbyteAlloc( *s.Handle.ulink, origEditSetLength + delta );
556	char *destCursor = dest; memcpy(destCursor, prefixStartOrig, prefixLen);
557	destCursor += prefixLen; memcpy(destCursor, buffer , bsize );
558	destCursor += bsize; memcpy(destCursor, suffixStartOrig, suffixLen);
559	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
560	dest,
561	origEditSetLength + delta,
562	0p, bsize);
563	free( oldBytes );
564	} else {
565	// room is already allocated in-place: bubble suffix and overwite middle
566	memmove( suffixStartOrig + delta, suffixStartOrig, suffixLen );
567	memcpy( s.Handle.s, buffer, bsize );
568
569	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
570	shareEditSetStartPeer->Handle.s,
571	origEditSetLength + delta,
572	0p, bsize);
573	}
574
575	} else if ( // assigning to shared context
576	s.Handle.lnth == origEditSetLength && // overwriting entire run of SES
577	& valSrc && // sourcing from a managed string
578	valSrc.Handle.ulink == s.Handle.ulink ) { // sourcing from same heap
579
580	// SES's result will only use characters from the source string => reuse source
581	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
582	valSrc.Handle.s,
583	valSrc.Handle.lnth,
584	&((string_res&)valSrc).Handle, bsize);
585
586	} else {
587	// overwriting a proper substring of some string: mash characters from old and new together (copy on write)
588	// OR we are importing characters: need to copy eagerly (can't refer to source)
589
590	// full string is from start of shareEditSetStartPeer thru end of shareEditSetEndPeer
591	// `s` occurs in the middle of it, to be replaced
592	// build up the new text in `pasting`
593
594	string_res pasting = {
595	* s.Handle.ulink, // maintain same heap, regardless of context
596	shareEditSetStartPeer->Handle.s, // start of SES
597	s.Handle.s - shareEditSetStartPeer->Handle.s }; // length of SES, before s
598	append( pasting,
599	buffer, // start of replacement for s
600	bsize ); // length of replacement for s
601	append( pasting,
602	s.Handle.s + s.Handle.lnth, // start of SES after s
603	shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth -
604	(s.Handle.s + s.Handle.lnth) ); // length of SES, after s
605
606	// The above string building can trigger compaction.
607	// The reference points (that are arguments of the string building) may move during that building.
608	// From s point on, they are stable.
609
610	assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer,
611	pasting.Handle.s,
612	pasting.Handle.lnth,
613	&pasting.Handle, bsize);
614	}
615
616	return s;
617	}
618
619	string_res & assign(string_res & s, const string_res & src, size_t maxlen) {
620	return assign_(s, src.Handle.s, min(src.Handle.lnth, maxlen), *0p);
621	}
622
623	string_res & assign(string_res & s, const char * buffer, size_t bsize) {
624	return assign_(s, buffer, bsize, *0p);
625	}
626
627	string_res & ?=?(string_res & s, char c) {
628	return assign(s, &c, 1);
629	}
630
631	string_res & ?=?( string_res & s, ssize_t rhs ) {
632	string_res rhs2 = rhs;
633	s = rhs2;
634	return s;
635	}
636	string_res & ?=?( string_res & s, size_t rhs ) {
637	string_res rhs2 = rhs;
638	s = rhs2;
639	return s;
640	}
641	string_res & ?=?( string_res & s, double rhs ) {
642	string_res rhs2 = rhs;
643	s = rhs2;
644	return s;
645	}
646	string_res & ?=?( string_res & s, long double rhs ) {
647	string_res rhs2 = rhs;
648	s = rhs2;
649	return s;
650	}
651	string_res & ?=?( string_res & s, double _Complex rhs ) {
652	string_res rhs2 = rhs;
653	s = rhs2;
654	return s;
655	}
656	string_res & ?=?( string_res & s, long double _Complex rhs ) {
657	string_res rhs2 = rhs;
658	s = rhs2;
659	return s;
660	}
661
662	// Copy assignment operator
663	string_res & ?=?(string_res & s, const string_res & rhs) with( s ) {
664	return assign_(s, rhs.Handle.s, rhs.Handle.lnth, rhs);
665	}
666
667	string_res & ?=?(string_res & s, string_res & rhs) with( s ) {
668	const string_res & rhs2 = rhs;
669	return s = rhs2;
670	}
671
672
673	// Destructor
674	void ^?{}(string_res & s) with(s) {
675	// much delegated to implied ^VbyteSM
676
677	// sever s from its share-edit peers, if any (four no-ops when already solo)
678	s.shareEditSet_prev->shareEditSet_next = s.shareEditSet_next;
679	s.shareEditSet_next->shareEditSet_prev = s.shareEditSet_prev;
680	// s.shareEditSet_next = &s;
681	// s.shareEditSet_prev = &s;
682
683	if (shareEditSet_owns_ulink && s.shareEditSet_next == &s) { // last one out
684	delete( s.Handle.ulink );
685	}
686	}
687
688
689	// Returns the character at the given index
690	// With unicode support, this may be different from just the byte at the given
691	// offset from the start of the string.
692	char ?[?](const string_res & s, size_t index) with(s) {
693	//TODO: Check if index is valid (no exceptions yet)
694	return Handle.s[index];
695	}
696
697	void assignAt(const string_res & s, size_t index, char val) {
698	// caution: not tested (not reachable by string-api-coverage interface)
699	// equivalent form at string level is `s[index] = val`,
700	// which uses the overload that returns a length-1 string
701	string_res editZone = { s, SHARE_EDITS, index, 1 };
702	assign(editZone, &val, 1);
703	}
704
705
706	///////////////////////////////////////////////////////////////////
707	// Concatenation
708
709	void append(string_res & str1, const char * buffer, size_t bsize) {
710	size_t clnth = str1.Handle.lnth + bsize;
711	if ( str1.Handle.s + str1.Handle.lnth == buffer ) { // already juxtapose ?
712	// no-op
713	} else { // must copy some text
714	if ( str1.Handle.s + str1.Handle.lnth == VbyteAlloc(*str1.Handle.ulink, 0) ) { // str1 at end of string area ?
715	VbyteAlloc( *str1.Handle.ulink, bsize ); // create room for 2nd part at the end of string area
716	} else { // copy the two parts
717	char * str1newBuf = VbyteAlloc( *str1.Handle.ulink, clnth );
718	char * str1oldBuf = str1.Handle.s; // must read after VbyteAlloc call in case it gs's
719	str1.Handle.s = str1newBuf;
720	memcpy( str1.Handle.s, str1oldBuf, str1.Handle.lnth );
721	} // if
722	memcpy( str1.Handle.s + str1.Handle.lnth, buffer, bsize );
723	} // if
724	str1.Handle.lnth = clnth;
725	}
726
727	void ?+=?(string_res & str1, const string_res & str2) {
728	append( str1, str2.Handle.s, str2.Handle.lnth );
729	}
730
731	void append(string_res & str1, const string_res & str2, size_t maxlen) {
732	append( str1, str2.Handle.s, min(str2.Handle.lnth, maxlen) );
733	}
734
735	void ?+=?(string_res & s, char c) {
736	append( s, & c, 1 );
737	}
738	void ?+=?(string_res & s, const char * c) {
739	append( s, c, strlen(c) );
740	}
741
742	///////////////////////////////////////////////////////////////////
743	// Repetition
744
745	void ?*=?(string_res & s, size_t factor) {
746	string_res s2 = { s, COPY_VALUE };
747	s = "";
748	for (factor) s += s2;
749	}
750
751	//////////////////////////////////////////////////////////
752	// Comparisons
753
754	int strcmp(const string_res & s1, const string_res & s2) {
755	// return 0;
756	int ans1 = memcmp(s1.Handle.s, s2.Handle.s, min(s1.Handle.lnth, s2.Handle.lnth));
757	if (ans1 != 0) return ans1;
758	return s1.Handle.lnth - s2.Handle.lnth;
759	}
760
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	bool ?<? (const string_res & s1, const string_res & s2) { return strcmp(s1, s2) < 0; }
767
768	int strcmp (const string_res & s1, const char * s2) {
769	string_res s2x = s2;
770	return strcmp(s1, s2x);
771	}
772
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	bool ?<? (const string_res & s1, const char * s2) { return strcmp(s1, s2) < 0; }
779
780	int strcmp (const char * s1, const string_res & s2) {
781	string_res s1x = s1;
782	return strcmp(s1x, s2);
783	}
784
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	bool ?<? (const char * s1, const string_res & s2) { return strcmp(s1, s2) < 0; }
791
792
793	//////////////////////////////////////////////////////////
794	// Search
795
796	bool contains(const string_res & s, char ch) {
797	for ( i; size(s) ) {
798	if (s[i] == ch) return true;
799	}
800	return false;
801	}
802
803	int find(const string_res & s, char search) {
804	return findFrom(s, 0, search);
805	}
806
807	int findFrom(const string_res & s, size_t fromPos, char search) {
808	// FIXME: This paricular overload (find of single char) is optimized to use memchr.
809	// The general overload (find of string, memchr applying to its first character) and `contains` should be adjusted to match.
810	char * searchFrom = s.Handle.s + fromPos;
811	size_t searchLnth = s.Handle.lnth - fromPos;
812	int searchVal = search;
813	char * foundAt = (char *) memchr(searchFrom, searchVal, searchLnth);
814	if (foundAt == 0p) return s.Handle.lnth;
815	else return foundAt - s.Handle.s;
816	}
817
818	int find(const string_res & s, const string_res & search) {
819	return findFrom(s, 0, search);
820	}
821
822	int findFrom(const string_res & s, size_t fromPos, const string_res & search) {
823	return findFrom(s, fromPos, search.Handle.s, search.Handle.lnth);
824	}
825
826	int find(const string_res & s, const char * search) {
827	return findFrom(s, 0, search);
828	}
829	int findFrom(const string_res & s, size_t fromPos, const char * search) {
830	return findFrom(s, fromPos, search, strlen(search));
831	}
832
833	int find(const string_res & s, const char * search, size_t searchsize) {
834	return findFrom(s, 0, search, searchsize);
835	}
836
837	int findFrom(const string_res & s, size_t fromPos, const char * search, size_t searchsize) {
838	/* Remaining implementations essentially ported from Sunjay's work */
839
840	// FIXME: This is a naive algorithm. We probably want to switch to someting
841	// like Boyer-Moore in the future.
842	// https://en.wikipedia.org/wiki/String_searching_algorithm
843
844	// Always find the empty string
845	if (searchsize == 0) {
846	return 0;
847	}
848
849	for ( i; fromPos ~ s.Handle.lnth ) {
850	size_t remaining = s.Handle.lnth - i;
851	// Never going to find the search string if the remaining string is
852	// smaller than search
853	if (remaining < searchsize) {
854	break;
855	}
856
857	bool matched = true;
858	for ( j; searchsize ) {
859	if (search[j] != s.Handle.s[i + j]) {
860	matched = false;
861	break;
862	}
863	}
864	if (matched) {
865	return i;
866	}
867	}
868	return s.Handle.lnth;
869	}
870
871	bool includes(const string_res & s, const string_res & search) {
872	return includes(s, search.Handle.s, search.Handle.lnth);
873	}
874
875	bool includes(const string_res & s, const char * search) {
876	return includes(s, search, strlen(search));
877	}
878
879	bool includes(const string_res & s, const char * search, size_t searchsize) {
880	return find(s, search, searchsize) < s.Handle.lnth;
881	}
882
883	bool startsWith(const string_res & s, const string_res & prefix) {
884	return startsWith(s, prefix.Handle.s, prefix.Handle.lnth);
885	}
886
887	bool startsWith(const string_res & s, const char * prefix) {
888	return startsWith(s, prefix, strlen(prefix));
889	}
890
891	bool startsWith(const string_res & s, const char * prefix, size_t prefixsize) {
892	if (s.Handle.lnth < prefixsize) {
893	return false;
894	}
895	return memcmp(s.Handle.s, prefix, prefixsize) == 0;
896	}
897
898	bool endsWith(const string_res & s, const string_res & suffix) {
899	return endsWith(s, suffix.Handle.s, suffix.Handle.lnth);
900	}
901
902	bool endsWith(const string_res & s, const char * suffix) {
903	return endsWith(s, suffix, strlen(suffix));
904	}
905
906	bool endsWith(const string_res & s, const char * suffix, size_t suffixsize) {
907	if (s.Handle.lnth < suffixsize) {
908	return false;
909	}
910	// Amount to offset the bytes pointer so that we are comparing the end of s
911	// to suffix. s.bytes + offset should be the first byte to compare against suffix
912	size_t offset = s.Handle.lnth - suffixsize;
913	return memcmp(s.Handle.s + offset, suffix, suffixsize) == 0;
914	}
915
916	/* Back to Mike's work */
917
918	///////////////////////////////////////////////////////////////////////////
919	// charclass, include, exclude
920
921	void ?{}( charclass_res & s, const string_res & chars) {
922	(s){ chars.Handle.s, chars.Handle.lnth };
923	}
924
925	void ?{}( charclass_res & s, const char * chars ) {
926	(s){ chars, strlen(chars) };
927	}
928
929	void ?{}( charclass_res & s, const char * chars, size_t charssize ) {
930	(s.chars){ chars, charssize };
931	// now sort it ?
932	}
933
934	void ^?{}( charclass_res & s ) {
935	^(s.chars){};
936	}
937
938	static bool test( const charclass_res & mask, char c ) {
939	// instead, use sorted char list?
940	return contains( mask.chars, c );
941	}
942
943	int exclude(const string_res & s, const charclass_res & mask) {
944	for ( i; size(s) ) {
945	if ( test(mask, s[i]) ) return i;
946	}
947	return size(s);
948	}
949
950	int include(const string_res & s, const charclass_res & mask) {
951	for ( i; size(s) ) {
952	if ( ! test(mask, s[i]) ) return i;
953	}
954	return size(s);
955	}
956
957	//######################### VbyteHeap "implementation" #########################
958
959
960	// Add a new HandleNode node n after the current HandleNode node.
961
962	static void AddThisAfter( HandleNode & s, HandleNode & n ) with(s) {
963	#ifdef VbyteDebug
964	serr \| "enter:AddThisAfter, s:" \| &s \| " n:" \| &n;
965	#endif // VbyteDebug
966	// 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).
967	verify( n.ulink != 0p );
968	verify( s.ulink == n.ulink );
969	flink = n.flink;
970	blink = &n;
971	n.flink->blink = &s;
972	n.flink = &s;
973	#ifdef VbyteDebug
974	{
975	serr \| "HandleList:";
976	serr \| nlOff;
977	for ( HandleNode *ni = HeaderPtr->flink; ni != HeaderPtr; ni = ni->flink ) {
978	serr \| "\tnode:" \| ni \| " lnth:" \| ni->lnth \| " s:" \| (void *)ni->s \| ",\"";
979	for ( i; ni->lnth ) {
980	serr \| ni->s[i];
981	} // for
982	serr \| "\" flink:" \| ni->flink \| " blink:" \| ni->blink \| nl;
983	} // for
984	serr \| nlOn;
985	}
986	serr \| "exit:AddThisAfter";
987	#endif // VbyteDebug
988	} // AddThisAfter
989
990
991	// Delete the current HandleNode node.
992
993	static void DeleteNode( HandleNode & s ) with(s) {
994	#ifdef VbyteDebug
995	serr \| "enter:DeleteNode, s:" \| &s;
996	#endif // VbyteDebug
997	flink->blink = blink;
998	blink->flink = flink;
999	#ifdef VbyteDebug
1000	serr \| "exit:DeleteNode";
1001	#endif // VbyteDebug
1002	} // DeleteNode
1003
1004
1005	// Allocates specified storage for a string from byte-string area. If not enough space remains to perform the
1006	// allocation, the garbage collection routine is called.
1007
1008	static char * VbyteAlloc( VbyteHeap & s, int size ) with(s) {
1009	#ifdef VbyteDebug
1010	serr \| "enter:VbyteAlloc, size:" \| size;
1011	#endif // VbyteDebug
1012	uintptr_t NoBytes;
1013	char *r;
1014
1015	NoBytes = ( uintptr_t )EndVbyte + size;
1016	if ( NoBytes > ( uintptr_t )ExtVbyte ) { // enough room for new byte-string ?
1017	garbage( s, size ); // firer up the garbage collector
1018	verify( (( uintptr_t )EndVbyte + size) <= ( uintptr_t )ExtVbyte && "garbage run did not free up required space" );
1019	} // if
1020	r = EndVbyte;
1021	EndVbyte += size;
1022	#ifdef VbyteDebug
1023	serr \| "exit:VbyteAlloc, r:" \| (void )r \| " EndVbyte:" \| (void )EndVbyte \| " ExtVbyte:" \| ExtVbyte;
1024	#endif // VbyteDebug
1025	return r;
1026	} // VbyteAlloc
1027
1028
1029	// Adjusts the last allocation in this heap by delta bytes, or resets this heap to be able to offer
1030	// new allocations of its original size + delta bytes. Positive delta means bigger;
1031	// negative means smaller. A null return indicates that the original heap location has room for
1032	// the requested growth. A non-null return indicates that copying to a new location is required
1033	// but has not been done; the returned value is the old heap storage location; `this` heap is
1034	// modified to reference the new location. In the copy-requred case, the caller should use
1035	// VbyteAlloc to claim the new space, while doing optimal copying from old to new, then free old.
1036
1037	static char * VbyteTryAdjustLast( VbyteHeap & s, int delta ) with(s) {
1038	if ( ( uintptr_t )EndVbyte + delta <= ( uintptr_t )ExtVbyte ) {
1039	// room available
1040	EndVbyte += delta;
1041	return 0p;
1042	}
1043
1044	char *oldBytes = StartVbyte;
1045
1046	NoOfExtensions += 1;
1047	CurrSize *= 2;
1048	StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize);
1049	ExtVbyte = StartVbyte + CurrSize;
1050
1051	return oldBytes;
1052	}
1053
1054
1055	// Move an existing HandleNode node h somewhere after the current HandleNode node so that it is in ascending order by
1056	// the address in the byte string area.
1057
1058	static void MoveThisAfter( HandleNode & s, const HandleNode & h ) with(s) {
1059	#ifdef VbyteDebug
1060	serr \| "enter:MoveThisAfter, s:" \| & s \| " h:" \| & h;
1061	#endif // VbyteDebug
1062	verify( h.ulink != 0p );
1063	verify( s.ulink == h.ulink );
1064	if ( s < h.s ) { // check argument values
1065	// serr \| "VbyteSM: Error - Cannot move byte string starting at:" \| s \| " after byte string starting at:"
1066	// \| ( h->s ) \| " and keep handles in ascending order";
1067	// exit(-1 );
1068	verify( 0 && "VbyteSM: Error - Cannot move byte strings as requested and keep handles in ascending order");
1069	} // if
1070
1071	HandleNode *i;
1072	for ( i = h.flink; i->s != 0 && s > ( i->s ); i = i->flink ); // find the position for this node after h
1073	if ( & s != i->blink ) {
1074	DeleteNode( s );
1075	AddThisAfter( s, *i->blink );
1076	} // if
1077	#ifdef VbyteDebug
1078	{
1079	serr \| "HandleList:";
1080	serr \| nlOff;
1081	for ( HandleNode *n = HeaderPtr->flink; n != HeaderPtr; n = n->flink ) {
1082	serr \| "\tnode:" \| n \| " lnth:" \| n->lnth \| " s:" \| (void *)n->s \| ",\"";
1083	for ( i; n->lnth ) {
1084	serr \| n->s[i];
1085	} // for
1086	serr \| "\" flink:" \| n->flink \| " blink:" \| n->blink \| nl;
1087	} // for
1088	serr \| nlOn;
1089	}
1090	serr \| "exit:MoveThisAfter";
1091	#endif // VbyteDebug
1092	} // MoveThisAfter
1093
1094
1095	//######################### VbyteHeap #########################
1096
1097	// Compare two byte strings in the byte-string area. The routine returns the following values:
1098	//
1099	// 1 => Src1-byte-string > Src2-byte-string
1100	// 0 => Src1-byte-string = Src2-byte-string
1101	// -1 => Src1-byte-string < Src2-byte-string
1102
1103	int ByteCmp( char Src1, int Src1Start, int Src1Lnth, char Src2, int Src2Start, int Src2Lnth ) {
1104	#ifdef VbyteDebug
1105	serr \| "enter:ByteCmp, Src1Start:" \| Src1Start \| " Src1Lnth:" \| Src1Lnth \| " Src2Start:" \| Src2Start \| " Src2Lnth:" \| Src2Lnth;
1106	#endif // VbyteDebug
1107	int cmp;
1108
1109	CharZip: for ( int i = 0; ; i += 1 ) {
1110	if ( i == Src2Lnth - 1 ) {
1111	for ( ; ; i += 1 ) {
1112	if ( i == Src1Lnth - 1 ) {
1113	cmp = 0;
1114	break CharZip;
1115	} // exit
1116	if ( Src1[Src1Start + i] != ' ') {
1117	// SUSPECTED BUG: this could be be why Peter got the bug report about == " " (why is this case here at all?)
1118	cmp = 1;
1119	break CharZip;
1120	} // exit
1121	} // for
1122	} // exit
1123	if ( i == Src1Lnth - 1 ) {
1124	for ( ; ; i += 1 ) {
1125	if ( i == Src2Lnth - 1 ) {
1126	cmp = 0;
1127	break CharZip;
1128	} // exit
1129	if ( Src2[Src2Start + i] != ' ') {
1130	cmp = -1;
1131	break CharZip;
1132	} // exit
1133	} // for
1134	} // exit
1135	if ( Src2[Src2Start + i] != Src1[Src1Start+ i]) {
1136	cmp = Src1[Src1Start + i] > Src2[Src2Start + i] ? 1 : -1;
1137	break CharZip;
1138	} // exit
1139	} // for
1140	#ifdef VbyteDebug
1141	serr \| "exit:ByteCmp, cmp:" \| cmp;
1142	#endif // VbyteDebug
1143	return cmp;
1144	} // ByteCmp
1145
1146
1147	// The compaction moves all of the byte strings currently in use to the beginning of the byte-string area and modifies
1148	// the handles to reflect the new positions of the byte strings. Compaction assumes that the handle list is in ascending
1149	// order by pointers into the byte-string area. The strings associated with substrings do not have to be moved because
1150	// the containing string has been moved. Hence, they only require that their string pointers be adjusted.
1151
1152	void compaction(VbyteHeap & s) with(s) {
1153	HandleNode *h;
1154	char obase, nbase, *limit;
1155
1156	NoOfCompactions += 1;
1157	EndVbyte = StartVbyte;
1158	h = Header.flink; // ignore header node
1159	for () {
1160	memmove( EndVbyte, h->s, h->lnth );
1161	obase = h->s;
1162	h->s = EndVbyte;
1163	nbase = h->s;
1164	EndVbyte += h->lnth;
1165	limit = obase + h->lnth;
1166	h = h->flink;
1167
1168	// check if any substrings are allocated within a string
1169
1170	for () {
1171	if ( h == &Header ) break; // end of header list ?
1172	if ( h->s >= limit ) break; // outside of current string ?
1173	h->s = nbase + (( uintptr_t )h->s - ( uintptr_t )obase );
1174	h = h->flink;
1175	} // for
1176	if ( h == &Header ) break; // end of header list ?
1177	} // for
1178	} // compaction
1179
1180
1181	static double heap_expansion_freespace_threshold = 0.1; // default inherited from prior work: expand heap when less than 10% "free" (i.e. garbage)
1182	// probably an unreasonable default, but need to assess early-round tests on changing it
1183
1184	void TUNING_set_string_heap_liveness_threshold( double val ) {
1185	heap_expansion_freespace_threshold = 1.0 - val;
1186	}
1187
1188
1189	// Garbage determines the amount of free space left in the heap and then reduces, leave the same, or extends the size of
1190	// the heap. The heap is then compacted in the existing heap or into the newly allocated heap.
1191
1192	void garbage(VbyteHeap & s, int minreq ) with(s) {
1193	#ifdef VbyteDebug
1194	serr \| "enter:garbage";
1195	{
1196	serr \| "HandleList:";
1197	for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) {
1198	serr \| nlOff;
1199	serr \| "\tnode:" \| n \| " lnth:" \| n->lnth \| " s:" \| (void *)n->s \| ",\"";
1200	for ( i; n->lnth ) {
1201	serr \| n->s[i];
1202	} // for
1203	serr \| nlOn;
1204	serr \| "\" flink:" \| n->flink \| " blink:" \| n->blink;
1205	} // for
1206	}
1207	#endif // VbyteDebug
1208	int AmountUsed, AmountFree;
1209
1210	AmountUsed = 0;
1211	for ( HandleNode *i = Header.flink; i != &Header; i = i->flink ) { // calculate amount of byte area used
1212	AmountUsed += i->lnth;
1213	} // for
1214	AmountFree = ( uintptr_t )ExtVbyte - ( uintptr_t )StartVbyte - AmountUsed;
1215
1216	if ( ( double ) AmountFree < ( CurrSize * heap_expansion_freespace_threshold ) \|\| AmountFree < minreq ) { // free space less than threshold or not enough to serve cur request
1217
1218	extend( s, max( CurrSize, minreq ) ); // extend the heap
1219
1220	// Peter says, "This needs work before it should be used."
1221	// } else if ( AmountFree > CurrSize / 2 ) { // free space greater than 3 times the initial allocation ?
1222	// reduce(( AmountFree / CurrSize - 3 ) * CurrSize ); // reduce the memory
1223
1224	// `extend` implies a `compaction` during the copy
1225
1226	} else {
1227	compaction(s); // in-place
1228	}// if
1229	#ifdef VbyteDebug
1230	{
1231	serr \| "HandleList:";
1232	for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) {
1233	serr \| nlOff;
1234	serr \| "\tnode:" \| n \| " lnth:" \| n->lnth \| " s:" \| (void *)n->s \| ",\"";
1235	for ( i; n->lnth ) {
1236	serr \| n->s[i];
1237	} // for
1238	serr \| nlOn;
1239	serr \| "\" flink:" \| n->flink \| " blink:" \| n->blink;
1240	} // for
1241	}
1242	serr \| "exit:garbage";
1243	#endif // VbyteDebug
1244	} // garbage
1245
1246	#undef VbyteDebug
1247
1248
1249
1250	// Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string
1251	// area is deleted.
1252
1253	void extend( VbyteHeap & s, int size ) with (s) {
1254	#ifdef VbyteDebug
1255	serr \| "enter:extend, size:" \| size;
1256	#endif // VbyteDebug
1257	char *OldStartVbyte;
1258
1259	NoOfExtensions += 1;
1260	OldStartVbyte = StartVbyte; // save previous byte area
1261
1262	CurrSize += size > InitSize ? size : InitSize; // minimum extension, initial size
1263	StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize);
1264	ExtVbyte = (void *)( StartVbyte + CurrSize );
1265	compaction(s); // copy from old heap to new & adjust pointers to new heap
1266	free( OldStartVbyte ); // release old heap
1267	#ifdef VbyteDebug
1268	serr \| "exit:extend, CurrSize:" \| CurrSize;
1269	#endif // VbyteDebug
1270	} // extend
1271
1272	//WIP
1273	#if 0
1274
1275	// Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string
1276	// area is deleted.
1277
1278	void VbyteHeap::reduce( int size ) {
1279	#ifdef VbyteDebug
1280	serr \| "enter:reduce, size:" \| size;
1281	#endif // VbyteDebug
1282	char *OldStartVbyte;
1283
1284	NoOfReductions += 1;
1285	OldStartVbyte = StartVbyte; // save previous byte area
1286
1287	CurrSize -= size;
1288	StartVbyte = EndVbyte = new char[CurrSize];
1289	ExtVbyte = (void *)( StartVbyte + CurrSize );
1290	compaction(); // copy from old heap to new & adjust pointers to new heap
1291	delete OldStartVbyte; // release old heap
1292	#ifdef VbyteDebug
1293	!serr \| "exit:reduce, CurrSize:" \| CurrSize;
1294	#endif // VbyteDebug
1295	} // reduce
1296
1297
1298	#endif

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

Download in other formats:

Original Format