[f450f2f] | 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 |
---|
[7d25f44] | 11 | // Last Modified By : Peter A. Buhr |
---|
[06280ad] | 12 | // Last Modified On : Tue Jan 16 22:19:27 2024 |
---|
| 13 | // Update Count : 35 |
---|
[f450f2f] | 14 | // |
---|
| 15 | |
---|
| 16 | #include "string_res.hfa" |
---|
[0f781fb8] | 17 | #include "string_sharectx.hfa" |
---|
[08ed947] | 18 | #include "stdlib.hfa" |
---|
[d32679d5] | 19 | #include <ctype.h> |
---|
[08ed947] | 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) |
---|
[681e12f] | 24 | #define TEMP_ALLOC(T, n) (( T * ) malloc( n * sizeof( T ) )) |
---|
[0f781fb8] | 25 | |
---|
[218096f] | 26 | #include <assert.h> |
---|
[f450f2f] | 27 | |
---|
| 28 | //######################### VbyteHeap "header" ######################### |
---|
| 29 | |
---|
| 30 | #ifdef VbyteDebug |
---|
[6cc87c0] | 31 | HandleNode *HeaderPtr; |
---|
[f450f2f] | 32 | #endif // VbyteDebug |
---|
| 33 | |
---|
| 34 | struct VbyteHeap { |
---|
[681e12f] | 35 | int NoOfCompactions; // number of compactions of the byte area |
---|
| 36 | int NoOfExtensions; // number of extensions in the size of the byte area |
---|
| 37 | int NoOfReductions; // number of reductions in the size of the byte area |
---|
[f450f2f] | 38 | |
---|
[681e12f] | 39 | int InitSize; // initial number of bytes in the byte-string area |
---|
| 40 | int CurrSize; // current number of bytes in the byte-string area |
---|
| 41 | char *StartVbyte; // pointer to the `st byte of the start of the byte-string area |
---|
| 42 | char *EndVbyte; // pointer to the next byte after the end of the currently used portion of byte-string area |
---|
| 43 | void *ExtVbyte; // pointer to the next byte after the end of the byte-string area |
---|
[f450f2f] | 44 | |
---|
[681e12f] | 45 | HandleNode Header; // header node for handle list |
---|
[f450f2f] | 46 | }; // VbyteHeap |
---|
| 47 | |
---|
| 48 | |
---|
[681e12f] | 49 | static void compaction( VbyteHeap & ); // compaction of the byte area |
---|
| 50 | static void garbage( VbyteHeap &, int ); // garbage collect the byte area |
---|
[4e8df745] | 51 | static void extend( VbyteHeap &, int ); // extend the size of the byte area |
---|
| 52 | static void reduce( VbyteHeap &, int ); // reduce the size of the byte area |
---|
[f450f2f] | 53 | |
---|
[4e8df745] | 54 | static void ?{}( VbyteHeap &, size_t = 1000 ); |
---|
| 55 | static void ^?{}( VbyteHeap & ); |
---|
[94647b0b] | 56 | |
---|
[4e8df745] | 57 | static int ByteCmp( char *, int, int, char *, int, int ); // compare 2 blocks of bytes |
---|
| 58 | static char *VbyteAlloc( VbyteHeap &, int ); // allocate a block bytes in the heap |
---|
| 59 | static char *VbyteTryAdjustLast( VbyteHeap &, int ); |
---|
[f450f2f] | 60 | |
---|
[4e8df745] | 61 | static void AddThisAfter( HandleNode &, HandleNode & ); |
---|
| 62 | static void DeleteNode( HandleNode & ); |
---|
| 63 | static void MoveThisAfter( HandleNode &, const HandleNode & ); // move current handle after parameter handle |
---|
[f450f2f] | 64 | |
---|
| 65 | |
---|
| 66 | // Allocate the storage for the variable sized area and intialize the heap variables. |
---|
| 67 | |
---|
[681e12f] | 68 | static void ?{}( VbyteHeap & s, size_t Size ) with(s) { |
---|
[f450f2f] | 69 | #ifdef VbyteDebug |
---|
[681e12f] | 70 | serr | "enter:VbyteHeap::VbyteHeap, s:" | &s | " Size:" | Size; |
---|
[f450f2f] | 71 | #endif // VbyteDebug |
---|
| 72 | NoOfCompactions = NoOfExtensions = NoOfReductions = 0; |
---|
| 73 | InitSize = CurrSize = Size; |
---|
[08ed947] | 74 | StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize); |
---|
[f450f2f] | 75 | ExtVbyte = (void *)( StartVbyte + CurrSize ); |
---|
| 76 | Header.flink = Header.blink = &Header; |
---|
[681e12f] | 77 | Header.ulink = &s; |
---|
[f450f2f] | 78 | #ifdef VbyteDebug |
---|
| 79 | HeaderPtr = &Header; |
---|
[681e12f] | 80 | serr | "exit:VbyteHeap::VbyteHeap, s:" | &s; |
---|
[f450f2f] | 81 | #endif // VbyteDebug |
---|
| 82 | } // VbyteHeap |
---|
| 83 | |
---|
| 84 | |
---|
| 85 | // Release the dynamically allocated storage for the byte area. |
---|
| 86 | |
---|
[681e12f] | 87 | static void ^?{}( VbyteHeap & s ) with(s) { |
---|
[f450f2f] | 88 | free( StartVbyte ); |
---|
| 89 | } // ~VbyteHeap |
---|
| 90 | |
---|
| 91 | |
---|
| 92 | //######################### HandleNode ######################### |
---|
| 93 | |
---|
| 94 | |
---|
| 95 | // Create a handle node. The handle is not linked into the handle list. This is the responsibilitiy of the handle |
---|
| 96 | // creator. |
---|
| 97 | |
---|
[681e12f] | 98 | static void ?{}( HandleNode & s ) with(s) { |
---|
[f450f2f] | 99 | #ifdef VbyteDebug |
---|
[681e12f] | 100 | serr | "enter:HandleNode::HandleNode, s:" | &s; |
---|
[f450f2f] | 101 | #endif // VbyteDebug |
---|
| 102 | s = 0; |
---|
| 103 | lnth = 0; |
---|
| 104 | #ifdef VbyteDebug |
---|
[681e12f] | 105 | serr | "exit:HandleNode::HandleNode, s:" | &s; |
---|
[f450f2f] | 106 | #endif // VbyteDebug |
---|
| 107 | } // HandleNode |
---|
| 108 | |
---|
| 109 | // Create a handle node. The handle is linked into the handle list at the end. This means that this handle will NOT be |
---|
| 110 | // in order by string address, but this is not a problem because a string with length zero does nothing during garbage |
---|
| 111 | // collection. |
---|
| 112 | |
---|
[681e12f] | 113 | static void ?{}( HandleNode & s, VbyteHeap & vh ) with(s) { |
---|
[f450f2f] | 114 | #ifdef VbyteDebug |
---|
[681e12f] | 115 | serr | "enter:HandleNode::HandleNode, s:" | &s; |
---|
[f450f2f] | 116 | #endif // VbyteDebug |
---|
| 117 | s = 0; |
---|
| 118 | lnth = 0; |
---|
[0f781fb8] | 119 | ulink = &vh; |
---|
[681e12f] | 120 | AddThisAfter( s, *vh.Header.blink ); |
---|
[f450f2f] | 121 | #ifdef VbyteDebug |
---|
[681e12f] | 122 | serr | "exit:HandleNode::HandleNode, s:" | &s; |
---|
[f450f2f] | 123 | #endif // VbyteDebug |
---|
| 124 | } // HandleNode |
---|
| 125 | |
---|
| 126 | |
---|
| 127 | // Delete a node from the handle list by unchaining it from the list. If the handle node was allocated dynamically, it |
---|
| 128 | // is the responsibility of the creator to destroy it. |
---|
| 129 | |
---|
[681e12f] | 130 | static void ^?{}( HandleNode & s ) with(s) { |
---|
[f450f2f] | 131 | #ifdef VbyteDebug |
---|
[681e12f] | 132 | serr | "enter:HandleNode::~HandleNode, s:" | & s; |
---|
[f450f2f] | 133 | { |
---|
| 134 | serr | nlOff; |
---|
| 135 | serr | " lnth:" | lnth | " s:" | (void *)s | ",\""; |
---|
[9ca5e56] | 136 | for ( i; lnth ) { |
---|
[f450f2f] | 137 | serr | s[i]; |
---|
| 138 | } // for |
---|
| 139 | serr | "\" flink:" | flink | " blink:" | blink | nl; |
---|
| 140 | serr | nlOn; |
---|
| 141 | } |
---|
| 142 | #endif // VbyteDebug |
---|
[681e12f] | 143 | DeleteNode( s ); |
---|
[f450f2f] | 144 | } // ~HandleNode |
---|
| 145 | |
---|
| 146 | |
---|
[0f781fb8] | 147 | //######################### String Sharing Context ######################### |
---|
[f450f2f] | 148 | |
---|
[0f781fb8] | 149 | static string_sharectx * ambient_string_sharectx; // fickle top of stack |
---|
| 150 | static string_sharectx default_string_sharectx = {NEW_SHARING}; // stable bottom of stack |
---|
[f450f2f] | 151 | |
---|
[681e12f] | 152 | void ?{}( string_sharectx & s, StringSharectx_Mode mode ) with( s ) { |
---|
[0f781fb8] | 153 | (older){ ambient_string_sharectx }; |
---|
| 154 | if ( mode == NEW_SHARING ) { |
---|
[804bf677] | 155 | (activeHeap){ new( (size_t) 1000 ) }; |
---|
[0f781fb8] | 156 | } else { |
---|
| 157 | verify( mode == NO_SHARING ); |
---|
| 158 | (activeHeap){ 0p }; |
---|
| 159 | } |
---|
[681e12f] | 160 | ambient_string_sharectx = & s; |
---|
[0f781fb8] | 161 | } |
---|
| 162 | |
---|
[681e12f] | 163 | void ^?{}( string_sharectx & s ) with( s ) { |
---|
[0f781fb8] | 164 | if ( activeHeap ) delete( activeHeap ); |
---|
| 165 | |
---|
[681e12f] | 166 | // unlink s from older-list starting from ambient_string_sharectx |
---|
| 167 | // usually, s==ambient_string_sharectx and the loop runs zero times |
---|
[0f781fb8] | 168 | string_sharectx *& c = ambient_string_sharectx; |
---|
[681e12f] | 169 | while ( c != &s ) &c = &c->older; // find s |
---|
| 170 | c = s.older; // unlink |
---|
[0f781fb8] | 171 | } |
---|
| 172 | |
---|
| 173 | //######################### String Resource ######################### |
---|
| 174 | |
---|
| 175 | |
---|
| 176 | VbyteHeap * DEBUG_string_heap() { |
---|
| 177 | assert( ambient_string_sharectx->activeHeap && "No sharing context is active" ); |
---|
| 178 | return ambient_string_sharectx->activeHeap; |
---|
| 179 | } |
---|
[6cc87c0] | 180 | |
---|
| 181 | size_t DEBUG_string_bytes_avail_until_gc( VbyteHeap * heap ) { |
---|
[06280ad] | 182 | return ((char *)heap->ExtVbyte) - heap->EndVbyte; |
---|
[6cc87c0] | 183 | } |
---|
| 184 | |
---|
[7b0e8b7] | 185 | size_t DEBUG_string_bytes_in_heap( VbyteHeap * heap ) { |
---|
| 186 | return heap->CurrSize; |
---|
| 187 | } |
---|
| 188 | |
---|
[6cc87c0] | 189 | const char * DEBUG_string_heap_start( VbyteHeap * heap ) { |
---|
| 190 | return heap->StartVbyte; |
---|
| 191 | } |
---|
| 192 | |
---|
[f450f2f] | 193 | // Returns the size of the string in bytes |
---|
[681e12f] | 194 | size_t size(const string_res & s) with(s) { |
---|
[f450f2f] | 195 | return Handle.lnth; |
---|
| 196 | } |
---|
| 197 | |
---|
| 198 | // Output operator |
---|
[681e12f] | 199 | ofstream & ?|?(ofstream & out, const string_res & s) { |
---|
[9ca5e56] | 200 | // CFA string is NOT null terminated, so print exactly lnth characters in a minimum width of 0. |
---|
| 201 | out | wd( 0, s.Handle.lnth, s.Handle.s ) | nonl; |
---|
[f450f2f] | 202 | return out; |
---|
| 203 | } |
---|
| 204 | |
---|
[681e12f] | 205 | void ?|?(ofstream & out, const string_res & s) { |
---|
[9ca5e56] | 206 | (ofstream &)(out | s); ends( out ); |
---|
[f450f2f] | 207 | } |
---|
| 208 | |
---|
[d32679d5] | 209 | // Input operator |
---|
[681e12f] | 210 | ifstream & ?|?(ifstream & in, string_res & s) { |
---|
[d32679d5] | 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. |
---|
[0860d9c] | 218 | for (bool cont = true; cont; ) { |
---|
| 219 | cont = false; |
---|
| 220 | |
---|
[d32679d5] | 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 | |
---|
[0860d9c] | 230 | // rest of heap is available to read into |
---|
[06280ad] | 231 | int lenReadable = (char *)temp.Handle.ulink->ExtVbyte - temp.Handle.ulink->EndVbyte; |
---|
[0860d9c] | 232 | assert (lenReadable >= 2); |
---|
[d32679d5] | 233 | |
---|
| 234 | // get bytes |
---|
[0860d9c] | 235 | try { |
---|
[37ceccb] | 236 | *(temp.Handle.ulink->EndVbyte) = '\0'; // pre-assign empty cstring |
---|
[0860d9c] | 237 | in | wdi( lenReadable, temp.Handle.ulink->EndVbyte ); |
---|
[681e12f] | 238 | } catch (cstring_length *) { |
---|
[0860d9c] | 239 | cont = true; |
---|
| 240 | } |
---|
[d32679d5] | 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 | |
---|
[37ceccb] | 248 | if ( temp.Handle.lnth > 0 ) s = temp; |
---|
[d32679d5] | 249 | return in; |
---|
| 250 | } |
---|
| 251 | |
---|
[681e12f] | 252 | void ?|?( ifstream & in, string_res & s ) { |
---|
| 253 | (ifstream &)(in | s); |
---|
[ff56dd2e] | 254 | } |
---|
| 255 | |
---|
| 256 | ifstream & ?|?( ifstream & is, _Istream_Rstr f ) { |
---|
| 257 | // .---------------, |
---|
| 258 | // | | | | |...|0|0| null terminator and guard if missing |
---|
| 259 | // `---------------' |
---|
| 260 | enum { gwd = 128 + 1, wd = gwd - 1 }; // guard and unguard width |
---|
| 261 | char cstr[gwd]; // read in chunks |
---|
| 262 | bool cont = false; |
---|
| 263 | |
---|
| 264 | _Istream_Cstr cf = { cstr, (_Istream_str_base)f }; |
---|
| 265 | if ( ! cf.flags.rwd ) cf.wd = wd; |
---|
| 266 | |
---|
| 267 | cstr[wd] = '\0'; // guard null terminate string |
---|
| 268 | try { |
---|
[37ceccb] | 269 | cstr[0] = '\0'; // pre-assign as empty cstring |
---|
[ff56dd2e] | 270 | is | cf; |
---|
| 271 | } catch( cstring_length * ) { |
---|
| 272 | cont = true; |
---|
| 273 | } finally { |
---|
[681e12f] | 274 | if ( ! cf.flags.ignore // ok to initialize string |
---|
| 275 | // && cstr[0] != '\0' // something was read |
---|
| 276 | ) { |
---|
[37ceccb] | 277 | *(f.s) = cstr; |
---|
| 278 | } |
---|
[ff56dd2e] | 279 | } // try |
---|
| 280 | for ( ; cont; ) { // overflow read ? |
---|
| 281 | cont = false; |
---|
| 282 | try { |
---|
[37ceccb] | 283 | cstr[0] = '\0'; // pre-assign as empty cstring |
---|
[ff56dd2e] | 284 | is | cf; |
---|
| 285 | } catch( cstring_length * ) { |
---|
| 286 | cont = true; // continue not allowed |
---|
| 287 | } finally { |
---|
[681e12f] | 288 | if ( ! cf.flags.ignore && cstr[0] != '\0' ) { // something was read |
---|
[37ceccb] | 289 | *(f.s) += cstr; // build string chunk at a time |
---|
| 290 | } |
---|
[ff56dd2e] | 291 | } // try |
---|
| 292 | } // for |
---|
| 293 | return is; |
---|
| 294 | } // ?|? |
---|
| 295 | |
---|
| 296 | void ?|?( ifstream & in, _Istream_Rstr f ) { |
---|
[f842032] | 297 | (ifstream &)(in | f); |
---|
[ff56dd2e] | 298 | } |
---|
| 299 | |
---|
[d32679d5] | 300 | |
---|
[f450f2f] | 301 | // Empty constructor |
---|
[681e12f] | 302 | void ?{}(string_res & s) with(s) { |
---|
[804bf677] | 303 | if( ambient_string_sharectx->activeHeap ) { |
---|
| 304 | (Handle){ * ambient_string_sharectx->activeHeap }; |
---|
| 305 | (shareEditSet_owns_ulink){ false }; |
---|
[2b30370] | 306 | verify( Handle.s == 0p && Handle.lnth == 0 ); |
---|
[804bf677] | 307 | } else { |
---|
| 308 | (Handle){ * new( (size_t) 10 ) }; // TODO: can I lazily avoid allocating for empty string |
---|
| 309 | (shareEditSet_owns_ulink){ true }; |
---|
[2b30370] | 310 | Handle.s = Handle.ulink->StartVbyte; |
---|
| 311 | verify( Handle.lnth == 0 ); |
---|
[804bf677] | 312 | } |
---|
[f450f2f] | 313 | s.shareEditSet_prev = &s; |
---|
| 314 | s.shareEditSet_next = &s; |
---|
| 315 | } |
---|
| 316 | |
---|
[06280ad] | 317 | static void eagerCopyCtorHelper(string_res & s, const char * rhs, size_t rhslnth) with(s) { |
---|
[804bf677] | 318 | if( ambient_string_sharectx->activeHeap ) { |
---|
| 319 | (Handle){ * ambient_string_sharectx->activeHeap }; |
---|
| 320 | (shareEditSet_owns_ulink){ false }; |
---|
| 321 | } else { |
---|
| 322 | (Handle){ * new( rhslnth ) }; |
---|
| 323 | (shareEditSet_owns_ulink){ true }; |
---|
| 324 | } |
---|
[0f781fb8] | 325 | Handle.s = VbyteAlloc(*Handle.ulink, rhslnth); |
---|
[f450f2f] | 326 | Handle.lnth = rhslnth; |
---|
[08ed947] | 327 | memmove( Handle.s, rhs, rhslnth ); |
---|
[f450f2f] | 328 | s.shareEditSet_prev = &s; |
---|
| 329 | s.shareEditSet_next = &s; |
---|
| 330 | } |
---|
| 331 | |
---|
[4b3b352] | 332 | // Constructor from a raw buffer and size |
---|
[06280ad] | 333 | void ?{}(string_res & s, const char * rhs, size_t rhslnth) with(s) { |
---|
[4b3b352] | 334 | eagerCopyCtorHelper(s, rhs, rhslnth); |
---|
| 335 | } |
---|
| 336 | |
---|
[fe18b46] | 337 | // private ctor (not in header): use specified heap (ignore ambient) and copy chars in |
---|
[06280ad] | 338 | void ?{}( string_res & s, VbyteHeap & heap, const char * rhs, size_t rhslnth ) with(s) { |
---|
[fe18b46] | 339 | (Handle){ heap }; |
---|
| 340 | Handle.s = VbyteAlloc(*Handle.ulink, rhslnth); |
---|
| 341 | Handle.lnth = rhslnth; |
---|
| 342 | (s.shareEditSet_owns_ulink){ false }; |
---|
[08ed947] | 343 | memmove( Handle.s, rhs, rhslnth ); |
---|
[fe18b46] | 344 | s.shareEditSet_prev = &s; |
---|
| 345 | s.shareEditSet_next = &s; |
---|
| 346 | } |
---|
| 347 | |
---|
[f450f2f] | 348 | // General copy constructor |
---|
[681e12f] | 349 | void ?{}(string_res & s, const string_res & s2, StrResInitMode mode, size_t start, size_t end ) { |
---|
[f450f2f] | 350 | |
---|
[218096f] | 351 | verify( start <= end && end <= s2.Handle.lnth ); |
---|
| 352 | |
---|
[fe18b46] | 353 | if (s2.Handle.ulink != ambient_string_sharectx->activeHeap && mode == COPY_VALUE) { |
---|
| 354 | // crossing heaps (including private): copy eagerly |
---|
| 355 | eagerCopyCtorHelper(s, s2.Handle.s + start, end - start); |
---|
| 356 | verify(s.shareEditSet_prev == &s); |
---|
| 357 | verify(s.shareEditSet_next == &s); |
---|
| 358 | } else { |
---|
[4b3b352] | 359 | (s.Handle){}; |
---|
| 360 | s.Handle.s = s2.Handle.s + start; |
---|
| 361 | s.Handle.lnth = end - start; |
---|
[fe18b46] | 362 | s.Handle.ulink = s2.Handle.ulink; |
---|
| 363 | |
---|
[4b3b352] | 364 | AddThisAfter(s.Handle, s2.Handle ); // insert this handle after rhs handle |
---|
| 365 | // ^ bug? skip others at early point in string |
---|
[fe18b46] | 366 | |
---|
[4b3b352] | 367 | if (mode == COPY_VALUE) { |
---|
[fe18b46] | 368 | verify(s2.Handle.ulink == ambient_string_sharectx->activeHeap); |
---|
| 369 | // requested logical copy in same heap: defer copy until write |
---|
| 370 | |
---|
| 371 | (s.shareEditSet_owns_ulink){ false }; |
---|
| 372 | |
---|
[4b3b352] | 373 | // make s alone in its shareEditSet |
---|
| 374 | s.shareEditSet_prev = &s; |
---|
| 375 | s.shareEditSet_next = &s; |
---|
| 376 | } else { |
---|
| 377 | verify( mode == SHARE_EDITS ); |
---|
[fe18b46] | 378 | // sharing edits with source forces same heap as source (ignore context) |
---|
| 379 | |
---|
| 380 | (s.shareEditSet_owns_ulink){ s2.shareEditSet_owns_ulink }; |
---|
[f450f2f] | 381 | |
---|
[4b3b352] | 382 | // s2 is logically const but not implementation const |
---|
| 383 | string_res & s2mod = (string_res &) s2; |
---|
[f450f2f] | 384 | |
---|
[4b3b352] | 385 | // insert s after s2 on shareEditSet |
---|
| 386 | s.shareEditSet_next = s2mod.shareEditSet_next; |
---|
| 387 | s.shareEditSet_prev = &s2mod; |
---|
| 388 | s.shareEditSet_next->shareEditSet_prev = &s; |
---|
| 389 | s.shareEditSet_prev->shareEditSet_next = &s; |
---|
[6cc87c0] | 390 | } |
---|
[f450f2f] | 391 | } |
---|
[4b3b352] | 392 | } |
---|
[f450f2f] | 393 | |
---|
[681e12f] | 394 | static void assignEditSet(string_res & s, string_res * shareEditSetStartPeer, string_res * shareEditSetEndPeer, |
---|
[4b3b352] | 395 | char * resultSesStart, |
---|
| 396 | size_t resultSesLnth, |
---|
| 397 | HandleNode * resultPadPosition, size_t bsize ) { |
---|
[6cc87c0] | 398 | |
---|
| 399 | char * beforeBegin = shareEditSetStartPeer->Handle.s; |
---|
[681e12f] | 400 | size_t beforeLen = s.Handle.s - beforeBegin; |
---|
[f450f2f] | 401 | |
---|
[681e12f] | 402 | char * afterBegin = s.Handle.s + s.Handle.lnth; |
---|
[6cc87c0] | 403 | size_t afterLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - afterBegin; |
---|
[f450f2f] | 404 | |
---|
[681e12f] | 405 | size_t oldLnth = s.Handle.lnth; |
---|
[f450f2f] | 406 | |
---|
[681e12f] | 407 | s.Handle.s = resultSesStart + beforeLen; |
---|
| 408 | s.Handle.lnth = bsize; |
---|
[6f7aff3] | 409 | if (resultPadPosition) |
---|
[681e12f] | 410 | MoveThisAfter( s.Handle, *resultPadPosition ); |
---|
[f450f2f] | 411 | |
---|
| 412 | // adjust all substring string and handle locations, and check if any substring strings are outside the new base string |
---|
[4b3b352] | 413 | char *limit = resultSesStart + resultSesLnth; |
---|
[681e12f] | 414 | for ( string_res * p = s.shareEditSet_next; p != &s; p = p->shareEditSet_next ) { |
---|
[218096f] | 415 | verify (p->Handle.s >= beforeBegin); |
---|
[6cc87c0] | 416 | if ( p->Handle.s >= afterBegin ) { |
---|
[218096f] | 417 | verify ( p->Handle.s <= afterBegin + afterLen ); |
---|
| 418 | verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen ); |
---|
[6cc87c0] | 419 | // p starts after the edit |
---|
| 420 | // take start and end as end-anchored |
---|
| 421 | size_t startOffsetFromEnd = afterBegin + afterLen - p->Handle.s; |
---|
| 422 | p->Handle.s = limit - startOffsetFromEnd; |
---|
| 423 | // p->Handle.lnth unaffected |
---|
| 424 | } else if ( p->Handle.s <= beforeBegin + beforeLen ) { |
---|
| 425 | // p starts before, or at the start of, the edit |
---|
| 426 | if ( p->Handle.s + p->Handle.lnth <= beforeBegin + beforeLen ) { |
---|
[f450f2f] | 427 | // p ends before the edit |
---|
| 428 | // take end as start-anchored too |
---|
| 429 | // p->Handle.lnth unaffected |
---|
| 430 | } else if ( p->Handle.s + p->Handle.lnth < afterBegin ) { |
---|
[6cc87c0] | 431 | // p ends during the edit; p does not include the last character replaced |
---|
[f450f2f] | 432 | // clip end of p to end at start of edit |
---|
| 433 | p->Handle.lnth = beforeLen - ( p->Handle.s - beforeBegin ); |
---|
| 434 | } else { |
---|
| 435 | // p ends after the edit |
---|
[218096f] | 436 | verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen ); |
---|
[f450f2f] | 437 | // take end as end-anchored |
---|
| 438 | // stretch-shrink p according to the edit |
---|
[681e12f] | 439 | p->Handle.lnth += s.Handle.lnth; |
---|
[f450f2f] | 440 | p->Handle.lnth -= oldLnth; |
---|
| 441 | } |
---|
| 442 | // take start as start-anchored |
---|
| 443 | size_t startOffsetFromStart = p->Handle.s - beforeBegin; |
---|
[4b3b352] | 444 | p->Handle.s = resultSesStart + startOffsetFromStart; |
---|
[6cc87c0] | 445 | } else { |
---|
[218096f] | 446 | verify ( p->Handle.s < afterBegin ); |
---|
[f450f2f] | 447 | // p starts during the edit |
---|
[218096f] | 448 | verify( p->Handle.s + p->Handle.lnth >= beforeBegin + beforeLen ); |
---|
[f450f2f] | 449 | if ( p->Handle.s + p->Handle.lnth < afterBegin ) { |
---|
[6cc87c0] | 450 | // p ends during the edit; p does not include the last character replaced |
---|
[f450f2f] | 451 | // set p to empty string at start of edit |
---|
[681e12f] | 452 | p->Handle.s = s.Handle.s; |
---|
[f450f2f] | 453 | p->Handle.lnth = 0; |
---|
| 454 | } else { |
---|
[6cc87c0] | 455 | // p includes the end of the edit |
---|
[f450f2f] | 456 | // clip start of p to start at end of edit |
---|
[6cc87c0] | 457 | int charsToClip = afterBegin - p->Handle.s; |
---|
[681e12f] | 458 | p->Handle.s = s.Handle.s + s.Handle.lnth; |
---|
[6cc87c0] | 459 | p->Handle.lnth -= charsToClip; |
---|
[f450f2f] | 460 | } |
---|
| 461 | } |
---|
[6f7aff3] | 462 | if (resultPadPosition) |
---|
| 463 | MoveThisAfter( p->Handle, *resultPadPosition ); // move substring handle to maintain sorted order by string position |
---|
[4b3b352] | 464 | } |
---|
| 465 | } |
---|
| 466 | |
---|
[681e12f] | 467 | // traverse the share-edit set (SES) to recover the range of a base string to which `s` belongs |
---|
| 468 | static void locateInShareEditSet( string_res & s, string_res *& shareEditSetStartPeer, string_res *& shareEditSetEndPeer ) { |
---|
| 469 | shareEditSetStartPeer = & s; |
---|
| 470 | shareEditSetEndPeer = & s; |
---|
| 471 | for (string_res * editPeer = s.shareEditSet_next; editPeer != &s; editPeer = editPeer->shareEditSet_next) { |
---|
[4b3b352] | 472 | if ( editPeer->Handle.s < shareEditSetStartPeer->Handle.s ) { |
---|
| 473 | shareEditSetStartPeer = editPeer; |
---|
| 474 | } |
---|
| 475 | if ( shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth < editPeer->Handle.s + editPeer->Handle.lnth) { |
---|
| 476 | shareEditSetEndPeer = editPeer; |
---|
| 477 | } |
---|
[f450f2f] | 478 | } |
---|
[d32679d5] | 479 | } |
---|
| 480 | |
---|
[06280ad] | 481 | static string_res & assign_(string_res & s, const char * buffer, size_t bsize, const string_res & valSrc) { |
---|
[d32679d5] | 482 | |
---|
| 483 | string_res * shareEditSetStartPeer; |
---|
| 484 | string_res * shareEditSetEndPeer; |
---|
[681e12f] | 485 | locateInShareEditSet( s, shareEditSetStartPeer, shareEditSetEndPeer ); |
---|
[4b3b352] | 486 | |
---|
| 487 | verify( shareEditSetEndPeer->Handle.s >= shareEditSetStartPeer->Handle.s ); |
---|
[0ca15b7] | 488 | size_t origEditSetLength = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - shareEditSetStartPeer->Handle.s; |
---|
[681e12f] | 489 | verify( origEditSetLength >= s.Handle.lnth ); |
---|
[4b3b352] | 490 | |
---|
[681e12f] | 491 | if ( s.shareEditSet_owns_ulink ) { // assigning to private context |
---|
[6f7aff3] | 492 | // ok to overwrite old value within LHS |
---|
| 493 | char * prefixStartOrig = shareEditSetStartPeer->Handle.s; |
---|
[681e12f] | 494 | int prefixLen = s.Handle.s - prefixStartOrig; |
---|
| 495 | char * suffixStartOrig = s.Handle.s + s.Handle.lnth; |
---|
[6f7aff3] | 496 | int suffixLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - suffixStartOrig; |
---|
| 497 | |
---|
[681e12f] | 498 | int delta = bsize - s.Handle.lnth; |
---|
| 499 | if ( char * oldBytes = VbyteTryAdjustLast( *s.Handle.ulink, delta ) ) { |
---|
[6f7aff3] | 500 | // growing: copy from old to new |
---|
[681e12f] | 501 | char * dest = VbyteAlloc( *s.Handle.ulink, origEditSetLength + delta ); |
---|
[fe18b46] | 502 | char *destCursor = dest; memcpy(destCursor, prefixStartOrig, prefixLen); |
---|
| 503 | destCursor += prefixLen; memcpy(destCursor, buffer , bsize ); |
---|
| 504 | destCursor += bsize; memcpy(destCursor, suffixStartOrig, suffixLen); |
---|
[681e12f] | 505 | assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
[6f7aff3] | 506 | dest, |
---|
[0ca15b7] | 507 | origEditSetLength + delta, |
---|
[6f7aff3] | 508 | 0p, bsize); |
---|
| 509 | free( oldBytes ); |
---|
| 510 | } else { |
---|
| 511 | // room is already allocated in-place: bubble suffix and overwite middle |
---|
| 512 | memmove( suffixStartOrig + delta, suffixStartOrig, suffixLen ); |
---|
[681e12f] | 513 | memcpy( s.Handle.s, buffer, bsize ); |
---|
[6f7aff3] | 514 | |
---|
[681e12f] | 515 | assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
[6f7aff3] | 516 | shareEditSetStartPeer->Handle.s, |
---|
[0ca15b7] | 517 | origEditSetLength + delta, |
---|
[6f7aff3] | 518 | 0p, bsize); |
---|
| 519 | } |
---|
[4b3b352] | 520 | |
---|
[6f7aff3] | 521 | } else if ( // assigning to shared context |
---|
[681e12f] | 522 | s.Handle.lnth == origEditSetLength && // overwriting entire run of SES |
---|
[6f7aff3] | 523 | & valSrc && // sourcing from a managed string |
---|
[681e12f] | 524 | valSrc.Handle.ulink == s.Handle.ulink ) { // sourcing from same heap |
---|
[4b3b352] | 525 | |
---|
[6f7aff3] | 526 | // SES's result will only use characters from the source string => reuse source |
---|
[681e12f] | 527 | assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
[4b3b352] | 528 | valSrc.Handle.s, |
---|
| 529 | valSrc.Handle.lnth, |
---|
| 530 | &((string_res&)valSrc).Handle, bsize); |
---|
| 531 | |
---|
| 532 | } else { |
---|
[6f7aff3] | 533 | // overwriting a proper substring of some string: mash characters from old and new together (copy on write) |
---|
| 534 | // OR we are importing characters: need to copy eagerly (can't refer to source) |
---|
[4b3b352] | 535 | |
---|
| 536 | // full string is from start of shareEditSetStartPeer thru end of shareEditSetEndPeer |
---|
[681e12f] | 537 | // `s` occurs in the middle of it, to be replaced |
---|
[4b3b352] | 538 | // build up the new text in `pasting` |
---|
| 539 | |
---|
| 540 | string_res pasting = { |
---|
[681e12f] | 541 | * s.Handle.ulink, // maintain same heap, regardless of context |
---|
[4b3b352] | 542 | shareEditSetStartPeer->Handle.s, // start of SES |
---|
[681e12f] | 543 | s.Handle.s - shareEditSetStartPeer->Handle.s }; // length of SES, before s |
---|
[4b3b352] | 544 | append( pasting, |
---|
[681e12f] | 545 | buffer, // start of replacement for s |
---|
| 546 | bsize ); // length of replacement for s |
---|
[4b3b352] | 547 | append( pasting, |
---|
[681e12f] | 548 | s.Handle.s + s.Handle.lnth, // start of SES after s |
---|
[4b3b352] | 549 | shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - |
---|
[681e12f] | 550 | (s.Handle.s + s.Handle.lnth) ); // length of SES, after s |
---|
[4b3b352] | 551 | |
---|
| 552 | // The above string building can trigger compaction. |
---|
| 553 | // The reference points (that are arguments of the string building) may move during that building. |
---|
[681e12f] | 554 | // From s point on, they are stable. |
---|
[4b3b352] | 555 | |
---|
[681e12f] | 556 | assignEditSet(s, shareEditSetStartPeer, shareEditSetEndPeer, |
---|
[4b3b352] | 557 | pasting.Handle.s, |
---|
| 558 | pasting.Handle.lnth, |
---|
| 559 | &pasting.Handle, bsize); |
---|
| 560 | } |
---|
[1733184] | 561 | |
---|
[681e12f] | 562 | return s; |
---|
[4b3b352] | 563 | } |
---|
| 564 | |
---|
[e891349] | 565 | string_res & assign(string_res & s, const string_res & src, size_t maxlen) { |
---|
| 566 | return assign_(s, src.Handle.s, min(src.Handle.lnth, maxlen), *0p); |
---|
| 567 | } |
---|
| 568 | |
---|
[06280ad] | 569 | string_res & assign(string_res & s, const char * buffer, size_t bsize) { |
---|
[681e12f] | 570 | return assign_(s, buffer, bsize, *0p); |
---|
[f450f2f] | 571 | } |
---|
| 572 | |
---|
[681e12f] | 573 | string_res & ?=?(string_res & s, char c) { |
---|
| 574 | return assign(s, &c, 1); |
---|
[d8d512e] | 575 | } |
---|
| 576 | |
---|
| 577 | // Copy assignment operator |
---|
[681e12f] | 578 | string_res & ?=?(string_res & s, const string_res & rhs) with( s ) { |
---|
| 579 | return assign_(s, rhs.Handle.s, rhs.Handle.lnth, rhs); |
---|
[d8d512e] | 580 | } |
---|
| 581 | |
---|
[681e12f] | 582 | string_res & ?=?(string_res & s, string_res & rhs) with( s ) { |
---|
[f450f2f] | 583 | const string_res & rhs2 = rhs; |
---|
[681e12f] | 584 | return s = rhs2; |
---|
[f450f2f] | 585 | } |
---|
| 586 | |
---|
| 587 | |
---|
| 588 | // Destructor |
---|
[681e12f] | 589 | void ^?{}(string_res & s) with(s) { |
---|
[f450f2f] | 590 | // much delegated to implied ^VbyteSM |
---|
| 591 | |
---|
| 592 | // sever s from its share-edit peers, if any (four no-ops when already solo) |
---|
| 593 | s.shareEditSet_prev->shareEditSet_next = s.shareEditSet_next; |
---|
| 594 | s.shareEditSet_next->shareEditSet_prev = s.shareEditSet_prev; |
---|
[218096f] | 595 | // s.shareEditSet_next = &s; |
---|
| 596 | // s.shareEditSet_prev = &s; |
---|
[804bf677] | 597 | |
---|
| 598 | if (shareEditSet_owns_ulink && s.shareEditSet_next == &s) { // last one out |
---|
| 599 | delete( s.Handle.ulink ); |
---|
| 600 | } |
---|
[f450f2f] | 601 | } |
---|
| 602 | |
---|
| 603 | |
---|
| 604 | // Returns the character at the given index |
---|
| 605 | // With unicode support, this may be different from just the byte at the given |
---|
| 606 | // offset from the start of the string. |
---|
[681e12f] | 607 | char ?[?](const string_res & s, size_t index) with(s) { |
---|
[f450f2f] | 608 | //TODO: Check if index is valid (no exceptions yet) |
---|
| 609 | return Handle.s[index]; |
---|
| 610 | } |
---|
| 611 | |
---|
[681e12f] | 612 | void assignAt(const string_res & s, size_t index, char val) { |
---|
[218096f] | 613 | string_res editZone = { s, SHARE_EDITS, index, index+1 }; |
---|
| 614 | assign(editZone, &val, 1); |
---|
| 615 | } |
---|
| 616 | |
---|
[d8d512e] | 617 | |
---|
[f450f2f] | 618 | /////////////////////////////////////////////////////////////////// |
---|
[d8d512e] | 619 | // Concatenation |
---|
[f450f2f] | 620 | |
---|
[681e12f] | 621 | void append(string_res & str1, const char * buffer, size_t bsize) { |
---|
[4e8df745] | 622 | size_t clnth = str1.Handle.lnth + bsize; |
---|
| 623 | if ( str1.Handle.s + str1.Handle.lnth == buffer ) { // already juxtapose ? |
---|
[d8d512e] | 624 | // no-op |
---|
[f450f2f] | 625 | } else { // must copy some text |
---|
[4e8df745] | 626 | if ( str1.Handle.s + str1.Handle.lnth == VbyteAlloc(*str1.Handle.ulink, 0) ) { // str1 at end of string area ? |
---|
[0f781fb8] | 627 | VbyteAlloc( *str1.Handle.ulink, bsize ); // create room for 2nd part at the end of string area |
---|
[f450f2f] | 628 | } else { // copy the two parts |
---|
[fe18b46] | 629 | char * str1newBuf = VbyteAlloc( *str1.Handle.ulink, clnth ); |
---|
| 630 | char * str1oldBuf = str1.Handle.s; // must read after VbyteAlloc call in case it gs's |
---|
| 631 | str1.Handle.s = str1newBuf; |
---|
[4e8df745] | 632 | memcpy( str1.Handle.s, str1oldBuf, str1.Handle.lnth ); |
---|
[f450f2f] | 633 | } // if |
---|
[4e8df745] | 634 | memcpy( str1.Handle.s + str1.Handle.lnth, buffer, bsize ); |
---|
[f450f2f] | 635 | } // if |
---|
| 636 | str1.Handle.lnth = clnth; |
---|
| 637 | } |
---|
| 638 | |
---|
[681e12f] | 639 | void ?+=?(string_res & str1, const string_res & str2) { |
---|
[d8d512e] | 640 | append( str1, str2.Handle.s, str2.Handle.lnth ); |
---|
[f450f2f] | 641 | } |
---|
| 642 | |
---|
[e891349] | 643 | void append(string_res & str1, const string_res & str2, size_t maxlen) { |
---|
| 644 | append( str1, str2.Handle.s, min(str2.Handle.lnth, maxlen) ); |
---|
| 645 | } |
---|
| 646 | |
---|
[681e12f] | 647 | void ?+=?(string_res & s, char c) { |
---|
| 648 | append( s, & c, 1 ); |
---|
[f450f2f] | 649 | } |
---|
[e891349] | 650 | void ?+=?(string_res & s, const char * c) { |
---|
| 651 | append( s, c, strlen(c) ); |
---|
| 652 | } |
---|
[f450f2f] | 653 | |
---|
[38951c31] | 654 | /////////////////////////////////////////////////////////////////// |
---|
| 655 | // Repetition |
---|
| 656 | |
---|
| 657 | void ?*=?(string_res & s, size_t factor) { |
---|
| 658 | string_res s2 = { s, COPY_VALUE, 0, size(s) }; |
---|
| 659 | s = ""; |
---|
| 660 | for (factor) s += s2; |
---|
| 661 | } |
---|
| 662 | |
---|
[f450f2f] | 663 | ////////////////////////////////////////////////////////// |
---|
| 664 | // Comparisons |
---|
| 665 | |
---|
[681e12f] | 666 | int strcmp(const string_res & s1, const string_res & s2) { |
---|
[416b443] | 667 | // return 0; |
---|
| 668 | int ans1 = memcmp(s1.Handle.s, s2.Handle.s, min(s1.Handle.lnth, s2.Handle.lnth)); |
---|
| 669 | if (ans1 != 0) return ans1; |
---|
| 670 | return s1.Handle.lnth - s2.Handle.lnth; |
---|
[f450f2f] | 671 | } |
---|
| 672 | |
---|
[681e12f] | 673 | bool ?==?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) == 0; } |
---|
| 674 | bool ?!=?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) != 0; } |
---|
| 675 | bool ?>? (const string_res & s1, const string_res & s2) { return strcmp(s1, s2) > 0; } |
---|
| 676 | bool ?>=?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) >= 0; } |
---|
| 677 | bool ?<=?(const string_res & s1, const string_res & s2) { return strcmp(s1, s2) <= 0; } |
---|
| 678 | bool ?<? (const string_res & s1, const string_res & s2) { return strcmp(s1, s2) < 0; } |
---|
[416b443] | 679 | |
---|
[06280ad] | 680 | int strcmp (const string_res & s1, const char * s2) { |
---|
[416b443] | 681 | string_res s2x = s2; |
---|
[681e12f] | 682 | return strcmp(s1, s2x); |
---|
[f450f2f] | 683 | } |
---|
[416b443] | 684 | |
---|
[06280ad] | 685 | bool ?==?(const string_res & s1, const char * s2) { return strcmp(s1, s2) == 0; } |
---|
| 686 | bool ?!=?(const string_res & s1, const char * s2) { return strcmp(s1, s2) != 0; } |
---|
| 687 | bool ?>? (const string_res & s1, const char * s2) { return strcmp(s1, s2) > 0; } |
---|
| 688 | bool ?>=?(const string_res & s1, const char * s2) { return strcmp(s1, s2) >= 0; } |
---|
| 689 | bool ?<=?(const string_res & s1, const char * s2) { return strcmp(s1, s2) <= 0; } |
---|
| 690 | bool ?<? (const string_res & s1, const char * s2) { return strcmp(s1, s2) < 0; } |
---|
[416b443] | 691 | |
---|
[06280ad] | 692 | int strcmp (const char * s1, const string_res & s2) { |
---|
[416b443] | 693 | string_res s1x = s1; |
---|
[681e12f] | 694 | return strcmp(s1x, s2); |
---|
[f450f2f] | 695 | } |
---|
| 696 | |
---|
[06280ad] | 697 | bool ?==?(const char * s1, const string_res & s2) { return strcmp(s1, s2) == 0; } |
---|
| 698 | bool ?!=?(const char * s1, const string_res & s2) { return strcmp(s1, s2) != 0; } |
---|
| 699 | bool ?>? (const char * s1, const string_res & s2) { return strcmp(s1, s2) > 0; } |
---|
| 700 | bool ?>=?(const char * s1, const string_res & s2) { return strcmp(s1, s2) >= 0; } |
---|
| 701 | bool ?<=?(const char * s1, const string_res & s2) { return strcmp(s1, s2) <= 0; } |
---|
| 702 | bool ?<? (const char * s1, const string_res & s2) { return strcmp(s1, s2) < 0; } |
---|
[416b443] | 703 | |
---|
| 704 | |
---|
[f450f2f] | 705 | |
---|
| 706 | ////////////////////////////////////////////////////////// |
---|
| 707 | // Search |
---|
| 708 | |
---|
[681e12f] | 709 | bool contains(const string_res & s, char ch) { |
---|
[9ca5e56] | 710 | for ( i; size(s) ) { |
---|
[f450f2f] | 711 | if (s[i] == ch) return true; |
---|
| 712 | } |
---|
| 713 | return false; |
---|
| 714 | } |
---|
| 715 | |
---|
[681e12f] | 716 | int find(const string_res & s, char search) { |
---|
[08ed947] | 717 | return findFrom(s, 0, search); |
---|
[f450f2f] | 718 | } |
---|
| 719 | |
---|
[681e12f] | 720 | int findFrom(const string_res & s, size_t fromPos, char search) { |
---|
[08ed947] | 721 | // FIXME: This paricular overload (find of single char) is optimized to use memchr. |
---|
| 722 | // The general overload (find of string, memchr applying to its first character) and `contains` should be adjusted to match. |
---|
| 723 | char * searchFrom = s.Handle.s + fromPos; |
---|
| 724 | size_t searchLnth = s.Handle.lnth - fromPos; |
---|
| 725 | int searchVal = search; |
---|
| 726 | char * foundAt = (char *) memchr(searchFrom, searchVal, searchLnth); |
---|
| 727 | if (foundAt == 0p) return s.Handle.lnth; |
---|
| 728 | else return foundAt - s.Handle.s; |
---|
| 729 | } |
---|
[f450f2f] | 730 | |
---|
[681e12f] | 731 | int find(const string_res & s, const string_res & search) { |
---|
[08ed947] | 732 | return findFrom(s, 0, search); |
---|
| 733 | } |
---|
| 734 | |
---|
[681e12f] | 735 | int findFrom(const string_res & s, size_t fromPos, const string_res & search) { |
---|
[08ed947] | 736 | return findFrom(s, fromPos, search.Handle.s, search.Handle.lnth); |
---|
[f450f2f] | 737 | } |
---|
| 738 | |
---|
[06280ad] | 739 | int find(const string_res & s, const char * search) { |
---|
[08ed947] | 740 | return findFrom(s, 0, search); |
---|
| 741 | } |
---|
[06280ad] | 742 | int findFrom(const string_res & s, size_t fromPos, const char * search) { |
---|
[08ed947] | 743 | return findFrom(s, fromPos, search, strlen(search)); |
---|
[f450f2f] | 744 | } |
---|
| 745 | |
---|
[06280ad] | 746 | int find(const string_res & s, const char * search, size_t searchsize) { |
---|
[08ed947] | 747 | return findFrom(s, 0, search, searchsize); |
---|
| 748 | } |
---|
| 749 | |
---|
[06280ad] | 750 | int findFrom(const string_res & s, size_t fromPos, const char * search, size_t searchsize) { |
---|
[08ed947] | 751 | |
---|
| 752 | /* Remaining implementations essentially ported from Sunjay's work */ |
---|
| 753 | |
---|
| 754 | |
---|
[f450f2f] | 755 | // FIXME: This is a naive algorithm. We probably want to switch to someting |
---|
| 756 | // like Boyer-Moore in the future. |
---|
| 757 | // https://en.wikipedia.org/wiki/String_searching_algorithm |
---|
| 758 | |
---|
| 759 | // Always find the empty string |
---|
| 760 | if (searchsize == 0) { |
---|
| 761 | return 0; |
---|
| 762 | } |
---|
| 763 | |
---|
[9ca5e56] | 764 | for ( i; fromPos ~ s.Handle.lnth ) { |
---|
[f450f2f] | 765 | size_t remaining = s.Handle.lnth - i; |
---|
| 766 | // Never going to find the search string if the remaining string is |
---|
| 767 | // smaller than search |
---|
| 768 | if (remaining < searchsize) { |
---|
| 769 | break; |
---|
| 770 | } |
---|
| 771 | |
---|
| 772 | bool matched = true; |
---|
[9ca5e56] | 773 | for ( j; searchsize ) { |
---|
[f450f2f] | 774 | if (search[j] != s.Handle.s[i + j]) { |
---|
| 775 | matched = false; |
---|
| 776 | break; |
---|
| 777 | } |
---|
| 778 | } |
---|
| 779 | if (matched) { |
---|
| 780 | return i; |
---|
| 781 | } |
---|
| 782 | } |
---|
| 783 | |
---|
| 784 | return s.Handle.lnth; |
---|
| 785 | } |
---|
| 786 | |
---|
[681e12f] | 787 | bool includes(const string_res & s, const string_res & search) { |
---|
[f450f2f] | 788 | return includes(s, search.Handle.s, search.Handle.lnth); |
---|
| 789 | } |
---|
| 790 | |
---|
[06280ad] | 791 | bool includes(const string_res & s, const char * search) { |
---|
[f450f2f] | 792 | return includes(s, search, strlen(search)); |
---|
| 793 | } |
---|
| 794 | |
---|
[06280ad] | 795 | bool includes(const string_res & s, const char * search, size_t searchsize) { |
---|
[f450f2f] | 796 | return find(s, search, searchsize) < s.Handle.lnth; |
---|
| 797 | } |
---|
| 798 | |
---|
[681e12f] | 799 | bool startsWith(const string_res & s, const string_res & prefix) { |
---|
[f450f2f] | 800 | return startsWith(s, prefix.Handle.s, prefix.Handle.lnth); |
---|
| 801 | } |
---|
| 802 | |
---|
[06280ad] | 803 | bool startsWith(const string_res & s, const char * prefix) { |
---|
[f450f2f] | 804 | return startsWith(s, prefix, strlen(prefix)); |
---|
| 805 | } |
---|
| 806 | |
---|
[06280ad] | 807 | bool startsWith(const string_res & s, const char * prefix, size_t prefixsize) { |
---|
[f450f2f] | 808 | if (s.Handle.lnth < prefixsize) { |
---|
| 809 | return false; |
---|
| 810 | } |
---|
| 811 | return memcmp(s.Handle.s, prefix, prefixsize) == 0; |
---|
| 812 | } |
---|
| 813 | |
---|
[681e12f] | 814 | bool endsWith(const string_res & s, const string_res & suffix) { |
---|
[f450f2f] | 815 | return endsWith(s, suffix.Handle.s, suffix.Handle.lnth); |
---|
| 816 | } |
---|
| 817 | |
---|
[06280ad] | 818 | bool endsWith(const string_res & s, const char * suffix) { |
---|
[f450f2f] | 819 | return endsWith(s, suffix, strlen(suffix)); |
---|
| 820 | } |
---|
| 821 | |
---|
[06280ad] | 822 | bool endsWith(const string_res & s, const char * suffix, size_t suffixsize) { |
---|
[f450f2f] | 823 | if (s.Handle.lnth < suffixsize) { |
---|
| 824 | return false; |
---|
| 825 | } |
---|
| 826 | // Amount to offset the bytes pointer so that we are comparing the end of s |
---|
| 827 | // to suffix. s.bytes + offset should be the first byte to compare against suffix |
---|
| 828 | size_t offset = s.Handle.lnth - suffixsize; |
---|
| 829 | return memcmp(s.Handle.s + offset, suffix, suffixsize) == 0; |
---|
| 830 | } |
---|
| 831 | |
---|
| 832 | /* Back to Mike's work */ |
---|
| 833 | |
---|
| 834 | |
---|
| 835 | /////////////////////////////////////////////////////////////////////////// |
---|
| 836 | // charclass, include, exclude |
---|
| 837 | |
---|
[681e12f] | 838 | void ?{}( charclass_res & s, const string_res & chars) { |
---|
| 839 | (s){ chars.Handle.s, chars.Handle.lnth }; |
---|
[f450f2f] | 840 | } |
---|
| 841 | |
---|
[681e12f] | 842 | void ?{}( charclass_res & s, const char * chars ) { |
---|
| 843 | (s){ chars, strlen(chars) }; |
---|
[f450f2f] | 844 | } |
---|
| 845 | |
---|
[681e12f] | 846 | void ?{}( charclass_res & s, const char * chars, size_t charssize ) { |
---|
| 847 | (s.chars){ chars, charssize }; |
---|
[f450f2f] | 848 | // now sort it ? |
---|
| 849 | } |
---|
| 850 | |
---|
[681e12f] | 851 | void ^?{}( charclass_res & s ) { |
---|
| 852 | ^(s.chars){}; |
---|
[f450f2f] | 853 | } |
---|
| 854 | |
---|
| 855 | static bool test( const charclass_res & mask, char c ) { |
---|
| 856 | // instead, use sorted char list? |
---|
| 857 | return contains( mask.chars, c ); |
---|
| 858 | } |
---|
| 859 | |
---|
[681e12f] | 860 | int exclude(const string_res & s, const charclass_res & mask) { |
---|
[9ca5e56] | 861 | for ( i; size(s) ) { |
---|
[f450f2f] | 862 | if ( test(mask, s[i]) ) return i; |
---|
| 863 | } |
---|
| 864 | return size(s); |
---|
| 865 | } |
---|
| 866 | |
---|
[681e12f] | 867 | int include(const string_res & s, const charclass_res & mask) { |
---|
[9ca5e56] | 868 | for ( i; size(s) ) { |
---|
[f450f2f] | 869 | if ( ! test(mask, s[i]) ) return i; |
---|
| 870 | } |
---|
| 871 | return size(s); |
---|
| 872 | } |
---|
| 873 | |
---|
| 874 | //######################### VbyteHeap "implementation" ######################### |
---|
| 875 | |
---|
| 876 | |
---|
| 877 | // Add a new HandleNode node n after the current HandleNode node. |
---|
| 878 | |
---|
[681e12f] | 879 | static void AddThisAfter( HandleNode & s, HandleNode & n ) with(s) { |
---|
[f450f2f] | 880 | #ifdef VbyteDebug |
---|
[681e12f] | 881 | serr | "enter:AddThisAfter, s:" | &s | " n:" | &n; |
---|
[f450f2f] | 882 | #endif // VbyteDebug |
---|
[08ed947] | 883 | // 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). |
---|
[0f781fb8] | 884 | verify( n.ulink != 0p ); |
---|
[681e12f] | 885 | verify( s.ulink == n.ulink ); |
---|
[f450f2f] | 886 | flink = n.flink; |
---|
| 887 | blink = &n; |
---|
[681e12f] | 888 | n.flink->blink = &s; |
---|
| 889 | n.flink = &s; |
---|
[f450f2f] | 890 | #ifdef VbyteDebug |
---|
| 891 | { |
---|
| 892 | serr | "HandleList:"; |
---|
| 893 | serr | nlOff; |
---|
| 894 | for ( HandleNode *ni = HeaderPtr->flink; ni != HeaderPtr; ni = ni->flink ) { |
---|
| 895 | serr | "\tnode:" | ni | " lnth:" | ni->lnth | " s:" | (void *)ni->s | ",\""; |
---|
[9ca5e56] | 896 | for ( i; ni->lnth ) { |
---|
[f450f2f] | 897 | serr | ni->s[i]; |
---|
| 898 | } // for |
---|
| 899 | serr | "\" flink:" | ni->flink | " blink:" | ni->blink | nl; |
---|
| 900 | } // for |
---|
| 901 | serr | nlOn; |
---|
| 902 | } |
---|
| 903 | serr | "exit:AddThisAfter"; |
---|
| 904 | #endif // VbyteDebug |
---|
| 905 | } // AddThisAfter |
---|
| 906 | |
---|
| 907 | |
---|
| 908 | // Delete the current HandleNode node. |
---|
| 909 | |
---|
[681e12f] | 910 | static void DeleteNode( HandleNode & s ) with(s) { |
---|
[f450f2f] | 911 | #ifdef VbyteDebug |
---|
[681e12f] | 912 | serr | "enter:DeleteNode, s:" | &s; |
---|
[f450f2f] | 913 | #endif // VbyteDebug |
---|
| 914 | flink->blink = blink; |
---|
| 915 | blink->flink = flink; |
---|
| 916 | #ifdef VbyteDebug |
---|
| 917 | serr | "exit:DeleteNode"; |
---|
| 918 | #endif // VbyteDebug |
---|
| 919 | } // DeleteNode |
---|
| 920 | |
---|
| 921 | |
---|
| 922 | // Allocates specified storage for a string from byte-string area. If not enough space remains to perform the |
---|
[6f7aff3] | 923 | // allocation, the garbage collection routine is called. |
---|
[f450f2f] | 924 | |
---|
[681e12f] | 925 | static char * VbyteAlloc( VbyteHeap & s, int size ) with(s) { |
---|
[f450f2f] | 926 | #ifdef VbyteDebug |
---|
| 927 | serr | "enter:VbyteAlloc, size:" | size; |
---|
| 928 | #endif // VbyteDebug |
---|
| 929 | uintptr_t NoBytes; |
---|
| 930 | char *r; |
---|
| 931 | |
---|
| 932 | NoBytes = ( uintptr_t )EndVbyte + size; |
---|
[681e12f] | 933 | if ( NoBytes > ( uintptr_t )ExtVbyte ) { // enough room for new byte-string ? |
---|
| 934 | garbage( s, size ); // firer up the garbage collector |
---|
[4e8df745] | 935 | verify( (( uintptr_t )EndVbyte + size) <= ( uintptr_t )ExtVbyte && "garbage run did not free up required space" ); |
---|
[f450f2f] | 936 | } // if |
---|
| 937 | r = EndVbyte; |
---|
| 938 | EndVbyte += size; |
---|
| 939 | #ifdef VbyteDebug |
---|
| 940 | serr | "exit:VbyteAlloc, r:" | (void *)r | " EndVbyte:" | (void *)EndVbyte | " ExtVbyte:" | ExtVbyte; |
---|
| 941 | #endif // VbyteDebug |
---|
| 942 | return r; |
---|
| 943 | } // VbyteAlloc |
---|
| 944 | |
---|
| 945 | |
---|
[6f7aff3] | 946 | // Adjusts the last allocation in this heap by delta bytes, or resets this heap to be able to offer |
---|
| 947 | // new allocations of its original size + delta bytes. Positive delta means bigger; |
---|
| 948 | // negative means smaller. A null return indicates that the original heap location has room for |
---|
| 949 | // the requested growth. A non-null return indicates that copying to a new location is required |
---|
| 950 | // but has not been done; the returned value is the old heap storage location; `this` heap is |
---|
| 951 | // modified to reference the new location. In the copy-requred case, the caller should use |
---|
| 952 | // VbyteAlloc to claim the new space, while doing optimal copying from old to new, then free old. |
---|
| 953 | |
---|
[681e12f] | 954 | static char * VbyteTryAdjustLast( VbyteHeap & s, int delta ) with(s) { |
---|
[6f7aff3] | 955 | if ( ( uintptr_t )EndVbyte + delta <= ( uintptr_t )ExtVbyte ) { |
---|
| 956 | // room available |
---|
| 957 | EndVbyte += delta; |
---|
| 958 | return 0p; |
---|
| 959 | } |
---|
| 960 | |
---|
| 961 | char *oldBytes = StartVbyte; |
---|
| 962 | |
---|
| 963 | NoOfExtensions += 1; |
---|
| 964 | CurrSize *= 2; |
---|
[08ed947] | 965 | StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize); |
---|
[6f7aff3] | 966 | ExtVbyte = StartVbyte + CurrSize; |
---|
| 967 | |
---|
| 968 | return oldBytes; |
---|
| 969 | } |
---|
| 970 | |
---|
| 971 | |
---|
[f450f2f] | 972 | // Move an existing HandleNode node h somewhere after the current HandleNode node so that it is in ascending order by |
---|
| 973 | // the address in the byte string area. |
---|
| 974 | |
---|
[681e12f] | 975 | static void MoveThisAfter( HandleNode & s, const HandleNode & h ) with(s) { |
---|
[f450f2f] | 976 | #ifdef VbyteDebug |
---|
[681e12f] | 977 | serr | "enter:MoveThisAfter, s:" | & s | " h:" | & h; |
---|
[f450f2f] | 978 | #endif // VbyteDebug |
---|
[0f781fb8] | 979 | verify( h.ulink != 0p ); |
---|
[681e12f] | 980 | verify( s.ulink == h.ulink ); |
---|
[f450f2f] | 981 | if ( s < h.s ) { // check argument values |
---|
| 982 | // serr | "VbyteSM: Error - Cannot move byte string starting at:" | s | " after byte string starting at:" |
---|
| 983 | // | ( h->s ) | " and keep handles in ascending order"; |
---|
| 984 | // exit(-1 ); |
---|
[218096f] | 985 | verify( 0 && "VbyteSM: Error - Cannot move byte strings as requested and keep handles in ascending order"); |
---|
[f450f2f] | 986 | } // if |
---|
| 987 | |
---|
| 988 | HandleNode *i; |
---|
| 989 | for ( i = h.flink; i->s != 0 && s > ( i->s ); i = i->flink ); // find the position for this node after h |
---|
[681e12f] | 990 | if ( & s != i->blink ) { |
---|
| 991 | DeleteNode( s ); |
---|
| 992 | AddThisAfter( s, *i->blink ); |
---|
[f450f2f] | 993 | } // if |
---|
| 994 | #ifdef VbyteDebug |
---|
| 995 | { |
---|
| 996 | serr | "HandleList:"; |
---|
| 997 | serr | nlOff; |
---|
| 998 | for ( HandleNode *n = HeaderPtr->flink; n != HeaderPtr; n = n->flink ) { |
---|
| 999 | serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; |
---|
[9ca5e56] | 1000 | for ( i; n->lnth ) { |
---|
| 1001 | serr | n->s[i]; |
---|
[f450f2f] | 1002 | } // for |
---|
[6cc87c0] | 1003 | serr | "\" flink:" | n->flink | " blink:" | n->blink | nl; |
---|
[f450f2f] | 1004 | } // for |
---|
| 1005 | serr | nlOn; |
---|
| 1006 | } |
---|
[6cc87c0] | 1007 | serr | "exit:MoveThisAfter"; |
---|
[f450f2f] | 1008 | #endif // VbyteDebug |
---|
| 1009 | } // MoveThisAfter |
---|
| 1010 | |
---|
| 1011 | |
---|
| 1012 | |
---|
| 1013 | |
---|
| 1014 | |
---|
| 1015 | //######################### VbyteHeap ######################### |
---|
| 1016 | |
---|
| 1017 | // Compare two byte strings in the byte-string area. The routine returns the following values: |
---|
| 1018 | // |
---|
| 1019 | // 1 => Src1-byte-string > Src2-byte-string |
---|
| 1020 | // 0 => Src1-byte-string = Src2-byte-string |
---|
| 1021 | // -1 => Src1-byte-string < Src2-byte-string |
---|
| 1022 | |
---|
[0f781fb8] | 1023 | int ByteCmp( char *Src1, int Src1Start, int Src1Lnth, char *Src2, int Src2Start, int Src2Lnth ) { |
---|
[f450f2f] | 1024 | #ifdef VbyteDebug |
---|
| 1025 | serr | "enter:ByteCmp, Src1Start:" | Src1Start | " Src1Lnth:" | Src1Lnth | " Src2Start:" | Src2Start | " Src2Lnth:" | Src2Lnth; |
---|
| 1026 | #endif // VbyteDebug |
---|
| 1027 | int cmp; |
---|
| 1028 | |
---|
| 1029 | CharZip: for ( int i = 0; ; i += 1 ) { |
---|
| 1030 | if ( i == Src2Lnth - 1 ) { |
---|
| 1031 | for ( ; ; i += 1 ) { |
---|
| 1032 | if ( i == Src1Lnth - 1 ) { |
---|
| 1033 | cmp = 0; |
---|
| 1034 | break CharZip; |
---|
| 1035 | } // exit |
---|
| 1036 | if ( Src1[Src1Start + i] != ' ') { |
---|
| 1037 | // SUSPECTED BUG: this could be be why Peter got the bug report about == " " (why is this case here at all?) |
---|
| 1038 | cmp = 1; |
---|
| 1039 | break CharZip; |
---|
| 1040 | } // exit |
---|
| 1041 | } // for |
---|
| 1042 | } // exit |
---|
| 1043 | if ( i == Src1Lnth - 1 ) { |
---|
| 1044 | for ( ; ; i += 1 ) { |
---|
| 1045 | if ( i == Src2Lnth - 1 ) { |
---|
| 1046 | cmp = 0; |
---|
| 1047 | break CharZip; |
---|
| 1048 | } // exit |
---|
| 1049 | if ( Src2[Src2Start + i] != ' ') { |
---|
| 1050 | cmp = -1; |
---|
| 1051 | break CharZip; |
---|
| 1052 | } // exit |
---|
| 1053 | } // for |
---|
| 1054 | } // exit |
---|
| 1055 | if ( Src2[Src2Start + i] != Src1[Src1Start+ i]) { |
---|
| 1056 | cmp = Src1[Src1Start + i] > Src2[Src2Start + i] ? 1 : -1; |
---|
| 1057 | break CharZip; |
---|
| 1058 | } // exit |
---|
| 1059 | } // for |
---|
| 1060 | #ifdef VbyteDebug |
---|
| 1061 | serr | "exit:ByteCmp, cmp:" | cmp; |
---|
| 1062 | #endif // VbyteDebug |
---|
| 1063 | return cmp; |
---|
| 1064 | } // ByteCmp |
---|
| 1065 | |
---|
| 1066 | |
---|
| 1067 | // The compaction moves all of the byte strings currently in use to the beginning of the byte-string area and modifies |
---|
| 1068 | // the handles to reflect the new positions of the byte strings. Compaction assumes that the handle list is in ascending |
---|
| 1069 | // order by pointers into the byte-string area. The strings associated with substrings do not have to be moved because |
---|
| 1070 | // the containing string has been moved. Hence, they only require that their string pointers be adjusted. |
---|
| 1071 | |
---|
[681e12f] | 1072 | void compaction(VbyteHeap & s) with(s) { |
---|
[f450f2f] | 1073 | HandleNode *h; |
---|
| 1074 | char *obase, *nbase, *limit; |
---|
| 1075 | |
---|
| 1076 | NoOfCompactions += 1; |
---|
| 1077 | EndVbyte = StartVbyte; |
---|
| 1078 | h = Header.flink; // ignore header node |
---|
[9ca5e56] | 1079 | for () { |
---|
[94647b0b] | 1080 | memmove( EndVbyte, h->s, h->lnth ); |
---|
[f450f2f] | 1081 | obase = h->s; |
---|
| 1082 | h->s = EndVbyte; |
---|
| 1083 | nbase = h->s; |
---|
| 1084 | EndVbyte += h->lnth; |
---|
| 1085 | limit = obase + h->lnth; |
---|
| 1086 | h = h->flink; |
---|
| 1087 | |
---|
| 1088 | // check if any substrings are allocated within a string |
---|
| 1089 | |
---|
[9ca5e56] | 1090 | for () { |
---|
[f450f2f] | 1091 | if ( h == &Header ) break; // end of header list ? |
---|
| 1092 | if ( h->s >= limit ) break; // outside of current string ? |
---|
| 1093 | h->s = nbase + (( uintptr_t )h->s - ( uintptr_t )obase ); |
---|
| 1094 | h = h->flink; |
---|
| 1095 | } // for |
---|
| 1096 | if ( h == &Header ) break; // end of header list ? |
---|
| 1097 | } // for |
---|
| 1098 | } // compaction |
---|
| 1099 | |
---|
| 1100 | |
---|
[08ed947] | 1101 | static double heap_expansion_freespace_threshold = 0.1; // default inherited from prior work: expand heap when less than 10% "free" (i.e. garbage) |
---|
| 1102 | // probably an unreasonable default, but need to assess early-round tests on changing it |
---|
| 1103 | |
---|
| 1104 | void TUNING_set_string_heap_liveness_threshold( double val ) { |
---|
| 1105 | heap_expansion_freespace_threshold = 1.0 - val; |
---|
| 1106 | } |
---|
| 1107 | |
---|
| 1108 | |
---|
[f450f2f] | 1109 | // Garbage determines the amount of free space left in the heap and then reduces, leave the same, or extends the size of |
---|
| 1110 | // the heap. The heap is then compacted in the existing heap or into the newly allocated heap. |
---|
| 1111 | |
---|
[681e12f] | 1112 | void garbage(VbyteHeap & s, int minreq ) with(s) { |
---|
[f450f2f] | 1113 | #ifdef VbyteDebug |
---|
| 1114 | serr | "enter:garbage"; |
---|
| 1115 | { |
---|
| 1116 | serr | "HandleList:"; |
---|
| 1117 | for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) { |
---|
| 1118 | serr | nlOff; |
---|
| 1119 | serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; |
---|
[9ca5e56] | 1120 | for ( i; n->lnth ) { |
---|
[f450f2f] | 1121 | serr | n->s[i]; |
---|
| 1122 | } // for |
---|
| 1123 | serr | nlOn; |
---|
| 1124 | serr | "\" flink:" | n->flink | " blink:" | n->blink; |
---|
| 1125 | } // for |
---|
| 1126 | } |
---|
| 1127 | #endif // VbyteDebug |
---|
| 1128 | int AmountUsed, AmountFree; |
---|
| 1129 | |
---|
| 1130 | AmountUsed = 0; |
---|
| 1131 | for ( HandleNode *i = Header.flink; i != &Header; i = i->flink ) { // calculate amount of byte area used |
---|
| 1132 | AmountUsed += i->lnth; |
---|
| 1133 | } // for |
---|
| 1134 | AmountFree = ( uintptr_t )ExtVbyte - ( uintptr_t )StartVbyte - AmountUsed; |
---|
| 1135 | |
---|
[08ed947] | 1136 | if ( ( double ) AmountFree < ( CurrSize * heap_expansion_freespace_threshold ) || AmountFree < minreq ) { // free space less than threshold or not enough to serve cur request |
---|
[f450f2f] | 1137 | |
---|
[681e12f] | 1138 | extend( s, max( CurrSize, minreq ) ); // extend the heap |
---|
[f450f2f] | 1139 | |
---|
| 1140 | // Peter says, "This needs work before it should be used." |
---|
| 1141 | // } else if ( AmountFree > CurrSize / 2 ) { // free space greater than 3 times the initial allocation ? |
---|
| 1142 | // reduce(( AmountFree / CurrSize - 3 ) * CurrSize ); // reduce the memory |
---|
| 1143 | |
---|
[97c215f] | 1144 | // `extend` implies a `compaction` during the copy |
---|
| 1145 | |
---|
| 1146 | } else { |
---|
[681e12f] | 1147 | compaction(s); // in-place |
---|
[97c215f] | 1148 | }// if |
---|
[f450f2f] | 1149 | #ifdef VbyteDebug |
---|
| 1150 | { |
---|
| 1151 | serr | "HandleList:"; |
---|
| 1152 | for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) { |
---|
| 1153 | serr | nlOff; |
---|
| 1154 | serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; |
---|
[9ca5e56] | 1155 | for ( i; n->lnth ) { |
---|
[f450f2f] | 1156 | serr | n->s[i]; |
---|
| 1157 | } // for |
---|
| 1158 | serr | nlOn; |
---|
| 1159 | serr | "\" flink:" | n->flink | " blink:" | n->blink; |
---|
| 1160 | } // for |
---|
| 1161 | } |
---|
| 1162 | serr | "exit:garbage"; |
---|
| 1163 | #endif // VbyteDebug |
---|
| 1164 | } // garbage |
---|
| 1165 | |
---|
| 1166 | #undef VbyteDebug |
---|
| 1167 | |
---|
| 1168 | |
---|
| 1169 | |
---|
| 1170 | // Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string |
---|
| 1171 | // area is deleted. |
---|
| 1172 | |
---|
[681e12f] | 1173 | void extend( VbyteHeap & s, int size ) with (s) { |
---|
[f450f2f] | 1174 | #ifdef VbyteDebug |
---|
| 1175 | serr | "enter:extend, size:" | size; |
---|
| 1176 | #endif // VbyteDebug |
---|
| 1177 | char *OldStartVbyte; |
---|
| 1178 | |
---|
| 1179 | NoOfExtensions += 1; |
---|
| 1180 | OldStartVbyte = StartVbyte; // save previous byte area |
---|
| 1181 | |
---|
| 1182 | CurrSize += size > InitSize ? size : InitSize; // minimum extension, initial size |
---|
[08ed947] | 1183 | StartVbyte = EndVbyte = TEMP_ALLOC(char, CurrSize); |
---|
[f450f2f] | 1184 | ExtVbyte = (void *)( StartVbyte + CurrSize ); |
---|
[681e12f] | 1185 | compaction(s); // copy from old heap to new & adjust pointers to new heap |
---|
[7b0e8b7] | 1186 | free( OldStartVbyte ); // release old heap |
---|
[f450f2f] | 1187 | #ifdef VbyteDebug |
---|
| 1188 | serr | "exit:extend, CurrSize:" | CurrSize; |
---|
| 1189 | #endif // VbyteDebug |
---|
| 1190 | } // extend |
---|
| 1191 | |
---|
[7b0e8b7] | 1192 | //WIP |
---|
| 1193 | #if 0 |
---|
[f450f2f] | 1194 | |
---|
| 1195 | // Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string |
---|
| 1196 | // area is deleted. |
---|
| 1197 | |
---|
| 1198 | void VbyteHeap::reduce( int size ) { |
---|
| 1199 | #ifdef VbyteDebug |
---|
| 1200 | serr | "enter:reduce, size:" | size; |
---|
| 1201 | #endif // VbyteDebug |
---|
| 1202 | char *OldStartVbyte; |
---|
| 1203 | |
---|
| 1204 | NoOfReductions += 1; |
---|
| 1205 | OldStartVbyte = StartVbyte; // save previous byte area |
---|
| 1206 | |
---|
| 1207 | CurrSize -= size; |
---|
| 1208 | StartVbyte = EndVbyte = new char[CurrSize]; |
---|
| 1209 | ExtVbyte = (void *)( StartVbyte + CurrSize ); |
---|
| 1210 | compaction(); // copy from old heap to new & adjust pointers to new heap |
---|
| 1211 | delete OldStartVbyte; // release old heap |
---|
| 1212 | #ifdef VbyteDebug |
---|
| 1213 | serr | "exit:reduce, CurrSize:" | CurrSize; |
---|
| 1214 | #endif // VbyteDebug |
---|
| 1215 | } // reduce |
---|
| 1216 | |
---|
| 1217 | |
---|
| 1218 | #endif |
---|