// // Cforall Version 1.0.0 Copyright (C) 2016 University of Waterloo // // The contents of this file are covered under the licence agreement in the // file "LICENCE" distributed with Cforall. // // string_res -- variable-length, mutable run of text, with resource semantics // // Author : Michael L. Brooks // Created On : Fri Sep 03 11:00:00 2021 // Last Modified By : Michael L. Brooks // Last Modified On : Fri Sep 03 11:00:00 2021 // Update Count : 1 // #include "string_res.hfa" #include "string_sharectx.hfa" #include // e.g. malloc #include // e.g. strlen #include //######################### VbyteHeap "header" ######################### #ifdef VbyteDebug HandleNode *HeaderPtr; #endif // VbyteDebug struct VbyteHeap { int NoOfCompactions; // number of compactions of the byte area int NoOfExtensions; // number of extensions in the size of the byte area int NoOfReductions; // number of reductions in the size of the byte area int InitSize; // initial number of bytes in the byte-string area int CurrSize; // current number of bytes in the byte-string area char *StartVbyte; // pointer to the `st byte of the start of the byte-string area char *EndVbyte; // pointer to the next byte after the end of the currently used portion of byte-string area void *ExtVbyte; // pointer to the next byte after the end of the byte-string area HandleNode Header; // header node for handle list }; // VbyteHeap static inline void compaction( VbyteHeap & ); // compaction of the byte area static inline void garbage( VbyteHeap & ); // garbage collect the byte area static inline void extend( VbyteHeap &, int ); // extend the size of the byte area static inline void reduce( VbyteHeap &, int ); // reduce the size of the byte area static inline void ?{}( VbyteHeap &, int = 1000 ); static inline void ^?{}( VbyteHeap & ); static inline void ByteCopy( char *, int, int, char *, int, int ); // copy a block of bytes from one location in the heap to another static inline int ByteCmp( char *, int, int, char *, int, int ); // compare 2 blocks of bytes static inline char *VbyteAlloc( VbyteHeap &, int ); // allocate a block bytes in the heap static inline void AddThisAfter( HandleNode &, HandleNode & ); static inline void DeleteNode( HandleNode & ); static inline void MoveThisAfter( HandleNode &, const HandleNode & ); // move current handle after parameter handle // Allocate the storage for the variable sized area and intialize the heap variables. static inline void ?{}( VbyteHeap & this, int Size ) with(this) { #ifdef VbyteDebug serr | "enter:VbyteHeap::VbyteHeap, this:" | &this | " Size:" | Size; #endif // VbyteDebug NoOfCompactions = NoOfExtensions = NoOfReductions = 0; InitSize = CurrSize = Size; StartVbyte = EndVbyte = alloc(CurrSize); ExtVbyte = (void *)( StartVbyte + CurrSize ); Header.flink = Header.blink = &Header; Header.ulink = & this; #ifdef VbyteDebug HeaderPtr = &Header; serr | "exit:VbyteHeap::VbyteHeap, this:" | &this; #endif // VbyteDebug } // VbyteHeap // Release the dynamically allocated storage for the byte area. static inline void ^?{}( VbyteHeap & this ) with(this) { free( StartVbyte ); } // ~VbyteHeap //######################### HandleNode ######################### // Create a handle node. The handle is not linked into the handle list. This is the responsibilitiy of the handle // creator. static inline void ?{}( HandleNode & this ) with(this) { #ifdef VbyteDebug serr | "enter:HandleNode::HandleNode, this:" | &this; #endif // VbyteDebug s = 0; lnth = 0; #ifdef VbyteDebug serr | "exit:HandleNode::HandleNode, this:" | &this; #endif // VbyteDebug } // HandleNode // Create a handle node. The handle is linked into the handle list at the end. This means that this handle will NOT be // in order by string address, but this is not a problem because a string with length zero does nothing during garbage // collection. static inline void ?{}( HandleNode & this, VbyteHeap & vh ) with(this) { #ifdef VbyteDebug serr | "enter:HandleNode::HandleNode, this:" | &this; #endif // VbyteDebug s = 0; lnth = 0; ulink = &vh; AddThisAfter( this, *vh.Header.blink ); #ifdef VbyteDebug serr | "exit:HandleNode::HandleNode, this:" | &this; #endif // VbyteDebug } // HandleNode // Delete a node from the handle list by unchaining it from the list. If the handle node was allocated dynamically, it // is the responsibility of the creator to destroy it. static inline void ^?{}( HandleNode & this ) with(this) { #ifdef VbyteDebug serr | "enter:HandleNode::~HandleNode, this:" | & this; { serr | nlOff; serr | " lnth:" | lnth | " s:" | (void *)s | ",\""; for ( int i = 0; i < lnth; i += 1 ) { serr | s[i]; } // for serr | "\" flink:" | flink | " blink:" | blink | nl; serr | nlOn; } #endif // VbyteDebug DeleteNode( this ); } // ~HandleNode //######################### String Sharing Context ######################### static string_sharectx * ambient_string_sharectx; // fickle top of stack static string_sharectx default_string_sharectx = {NEW_SHARING}; // stable bottom of stack void ?{}( string_sharectx & this, StringSharectx_Mode mode ) with( this ) { (older){ ambient_string_sharectx }; if ( mode == NEW_SHARING ) { (activeHeap){ new( 1000 ) }; } else { verify( mode == NO_SHARING ); (activeHeap){ 0p }; } ambient_string_sharectx = & this; } void ^?{}( string_sharectx & this ) with( this ) { if ( activeHeap ) delete( activeHeap ); // unlink this from older-list starting from ambient_string_sharectx // usually, this==ambient_string_sharectx and the loop runs zero times string_sharectx *& c = ambient_string_sharectx; while ( c != &this ) &c = &c->older; // find this c = this.older; // unlink } //######################### String Resource ######################### VbyteHeap * DEBUG_string_heap() { assert( ambient_string_sharectx->activeHeap && "No sharing context is active" ); return ambient_string_sharectx->activeHeap; } size_t DEBUG_string_bytes_avail_until_gc( VbyteHeap * heap ) { return ((char*)heap->ExtVbyte) - heap->EndVbyte; } const char * DEBUG_string_heap_start( VbyteHeap * heap ) { return heap->StartVbyte; } // Returns the size of the string in bytes size_t size(const string_res &s) with(s) { return Handle.lnth; } // Output operator ofstream & ?|?(ofstream &out, const string_res &s) { // Store auto-newline state so it can be restored bool anl = getANL$(out); nlOff(out); for (size_t i = 0; i < s.Handle.lnth; i++) { out | s[i]; } out | sep; // Re-apply newlines after done, for chaining version if (anl) nlOn(out); return out; } void ?|?(ofstream &out, const string_res &s) { // Store auto-newline state so it can be restored bool anl = getANL$(out); nlOff(out); for (size_t i = 0; i < s.Handle.lnth; i++) { // Need to re-apply on the last output operator, for whole-statement version if (anl && i == s.Handle.lnth-1) nlOn(out); out | s[i]; } return out; } // Empty constructor void ?{}(string_res &s) with(s) { assert( ambient_string_sharectx->activeHeap && "Need to implement private contexts" ); (Handle){ * ambient_string_sharectx->activeHeap }; s.shareEditSet_prev = &s; s.shareEditSet_next = &s; } // Constructor from a raw buffer and size void ?{}(string_res &s, const char* rhs, size_t rhslnth) with(s) { assert( ambient_string_sharectx->activeHeap && "Need to implement private contexts" ); (Handle){ * ambient_string_sharectx->activeHeap }; Handle.s = VbyteAlloc(*Handle.ulink, rhslnth); Handle.lnth = rhslnth; for ( int i = 0; i < rhslnth; i += 1 ) { // copy characters Handle.s[i] = rhs[i]; } // for s.shareEditSet_prev = &s; s.shareEditSet_next = &s; } // String literal constructor void ?{}(string_res &s, const char* rhs) { (s){ rhs, strlen(rhs) }; } // General copy constructor void ?{}(string_res &s, const string_res & s2, StrResInitMode mode, size_t start, size_t end ) { assert( ambient_string_sharectx->activeHeap && "Need to implement private contexts" ); assert( s2.Handle.ulink == ambient_string_sharectx->activeHeap && "need to implement context crossing"); verify( start <= end && end <= s2.Handle.lnth ); (s.Handle){}; s.Handle.s = s2.Handle.s + start; s.Handle.lnth = end - start; s.Handle.ulink = ambient_string_sharectx->activeHeap; AddThisAfter(s.Handle, s2.Handle ); // insert this handle after rhs handle // ^ bug? skip others at early point in string if (mode == COPY_VALUE) { // make s alone in its shareEditSet s.shareEditSet_prev = &s; s.shareEditSet_next = &s; } else { verify( mode == SHARE_EDITS ); // s2 is logically const but not implementation const string_res & s2mod = (string_res &) s2; // insert s after s2 on shareEditSet s.shareEditSet_next = s2mod.shareEditSet_next; s.shareEditSet_prev = &s2mod; s.shareEditSet_next->shareEditSet_prev = &s; s.shareEditSet_prev->shareEditSet_next = &s; } } void assign(string_res &this, const char* buffer, size_t bsize) { // traverse the incumbent share-edit set (SES) to recover the range of a base string to which `this` belongs string_res * shareEditSetStartPeer = & this; string_res * shareEditSetEndPeer = & this; for (string_res * editPeer = this.shareEditSet_next; editPeer != &this; editPeer = editPeer->shareEditSet_next) { if ( editPeer->Handle.s < shareEditSetStartPeer->Handle.s ) { shareEditSetStartPeer = editPeer; } if ( shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth < editPeer->Handle.s + editPeer->Handle.lnth) { shareEditSetEndPeer = editPeer; } } // full string is from start of shareEditSetStartPeer thru end of shareEditSetEndPeer // `this` occurs in the middle of it, to be replaced // build up the new text in `pasting` string_res pasting = { shareEditSetStartPeer->Handle.s, // start of SES this.Handle.s - shareEditSetStartPeer->Handle.s }; // length of SES, before this append( pasting, buffer, // start of replacement for this bsize ); // length of replacement for this append( pasting, this.Handle.s + this.Handle.lnth, // start of SES after this shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - (this.Handle.s + this.Handle.lnth) ); // length of SES, after this // The above string building can trigger compaction. // The reference points (that are arguments of the string building) may move during that building. // From this point on, they are stable. // So now, capture their values for use in the overlap cases, below. // Do not factor these definitions with the arguments used above. char * beforeBegin = shareEditSetStartPeer->Handle.s; size_t beforeLen = this.Handle.s - beforeBegin; char * afterBegin = this.Handle.s + this.Handle.lnth; size_t afterLen = shareEditSetEndPeer->Handle.s + shareEditSetEndPeer->Handle.lnth - afterBegin; size_t oldLnth = this.Handle.lnth; this.Handle.s = pasting.Handle.s + beforeLen; this.Handle.lnth = bsize; MoveThisAfter( this.Handle, pasting.Handle ); // adjust all substring string and handle locations, and check if any substring strings are outside the new base string char *limit = pasting.Handle.s + pasting.Handle.lnth; for (string_res * p = this.shareEditSet_next; p != &this; p = p->shareEditSet_next) { verify (p->Handle.s >= beforeBegin); if ( p->Handle.s >= afterBegin ) { verify ( p->Handle.s <= afterBegin + afterLen ); verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen ); // p starts after the edit // take start and end as end-anchored size_t startOffsetFromEnd = afterBegin + afterLen - p->Handle.s; p->Handle.s = limit - startOffsetFromEnd; // p->Handle.lnth unaffected } else if ( p->Handle.s <= beforeBegin + beforeLen ) { // p starts before, or at the start of, the edit if ( p->Handle.s + p->Handle.lnth <= beforeBegin + beforeLen ) { // p ends before the edit // take end as start-anchored too // p->Handle.lnth unaffected } else if ( p->Handle.s + p->Handle.lnth < afterBegin ) { // p ends during the edit; p does not include the last character replaced // clip end of p to end at start of edit p->Handle.lnth = beforeLen - ( p->Handle.s - beforeBegin ); } else { // p ends after the edit verify ( p->Handle.s + p->Handle.lnth <= afterBegin + afterLen ); // take end as end-anchored // stretch-shrink p according to the edit p->Handle.lnth += this.Handle.lnth; p->Handle.lnth -= oldLnth; } // take start as start-anchored size_t startOffsetFromStart = p->Handle.s - beforeBegin; p->Handle.s = pasting.Handle.s + startOffsetFromStart; } else { verify ( p->Handle.s < afterBegin ); // p starts during the edit verify( p->Handle.s + p->Handle.lnth >= beforeBegin + beforeLen ); if ( p->Handle.s + p->Handle.lnth < afterBegin ) { // p ends during the edit; p does not include the last character replaced // set p to empty string at start of edit p->Handle.s = this.Handle.s; p->Handle.lnth = 0; } else { // p includes the end of the edit // clip start of p to start at end of edit int charsToClip = afterBegin - p->Handle.s; p->Handle.s = this.Handle.s + this.Handle.lnth; p->Handle.lnth -= charsToClip; } } MoveThisAfter( p->Handle, pasting.Handle ); // move substring handle to maintain sorted order by string position } } void ?=?(string_res &s, const char* other) { assign(s, other, strlen(other)); } void ?=?(string_res &s, char other) { assign(s, &other, 1); } // Copy assignment operator void ?=?(string_res & this, const string_res & rhs) with( this ) { assign(this, rhs.Handle.s, rhs.Handle.lnth); } void ?=?(string_res & this, string_res & rhs) with( this ) { const string_res & rhs2 = rhs; this = rhs2; } // Destructor void ^?{}(string_res &s) with(s) { // much delegated to implied ^VbyteSM // sever s from its share-edit peers, if any (four no-ops when already solo) s.shareEditSet_prev->shareEditSet_next = s.shareEditSet_next; s.shareEditSet_next->shareEditSet_prev = s.shareEditSet_prev; // s.shareEditSet_next = &s; // s.shareEditSet_prev = &s; } // Returns the character at the given index // With unicode support, this may be different from just the byte at the given // offset from the start of the string. char ?[?](const string_res &s, size_t index) with(s) { //TODO: Check if index is valid (no exceptions yet) return Handle.s[index]; } void assignAt(const string_res &s, size_t index, char val) { string_res editZone = { s, SHARE_EDITS, index, index+1 }; assign(editZone, &val, 1); } /////////////////////////////////////////////////////////////////// // Concatenation void append(string_res &str1, const char * buffer, size_t bsize) { size_t clnth = size(str1) + bsize; if ( str1.Handle.s + size(str1) == buffer ) { // already juxtapose ? // no-op } else { // must copy some text if ( str1.Handle.s + size(str1) == VbyteAlloc(*str1.Handle.ulink, 0) ) { // str1 at end of string area ? VbyteAlloc( *str1.Handle.ulink, bsize ); // create room for 2nd part at the end of string area } else { // copy the two parts char * str1oldBuf = str1.Handle.s; str1.Handle.s = VbyteAlloc( *str1.Handle.ulink, clnth ); ByteCopy( str1.Handle.s, 0, str1.Handle.lnth, str1oldBuf, 0, str1.Handle.lnth); } // if ByteCopy( str1.Handle.s, str1.Handle.lnth, bsize, (char*)buffer, 0, (int)bsize); // VbyteHeap & this, char *Dst, int DstStart, int DstLnth, char *Src, int SrcStart, int SrcLnth } // if str1.Handle.lnth = clnth; } void ?+=?(string_res &str1, const string_res &str2) { append( str1, str2.Handle.s, str2.Handle.lnth ); } void ?+=?(string_res &s, char other) { append( s, &other, 1 ); } void ?+=?(string_res &s, const char* other) { append( s, other, strlen(other) ); } ////////////////////////////////////////////////////////// // Comparisons bool ?==?(const string_res &s1, const string_res &s2) { return ByteCmp( s1.Handle.s, 0, s1.Handle.lnth, s2.Handle.s, 0, s2.Handle.lnth) == 0; } bool ?!=?(const string_res &s1, const string_res &s2) { return !(s1 == s2); } bool ?==?(const string_res &s, const char* other) { string_res sother = other; return s == sother; } bool ?!=?(const string_res &s, const char* other) { return !(s == other); } ////////////////////////////////////////////////////////// // Search bool contains(const string_res &s, char ch) { for (i; size(s)) { if (s[i] == ch) return true; } return false; } int find(const string_res &s, char search) { for (i; size(s)) { if (s[i] == search) return i; } return size(s); } /* Remaining implementations essentially ported from Sunjay's work */ int find(const string_res &s, const string_res &search) { return find(s, search.Handle.s, search.Handle.lnth); } int find(const string_res &s, const char* search) { return find(s, search, strlen(search)); } int find(const string_res &s, const char* search, size_t searchsize) { // FIXME: This is a naive algorithm. We probably want to switch to someting // like Boyer-Moore in the future. // https://en.wikipedia.org/wiki/String_searching_algorithm // Always find the empty string if (searchsize == 0) { return 0; } for (size_t i = 0; i < s.Handle.lnth; i++) { size_t remaining = s.Handle.lnth - i; // Never going to find the search string if the remaining string is // smaller than search if (remaining < searchsize) { break; } bool matched = true; for (size_t j = 0; j < searchsize; j++) { if (search[j] != s.Handle.s[i + j]) { matched = false; break; } } if (matched) { return i; } } return s.Handle.lnth; } bool includes(const string_res &s, const string_res &search) { return includes(s, search.Handle.s, search.Handle.lnth); } bool includes(const string_res &s, const char* search) { return includes(s, search, strlen(search)); } bool includes(const string_res &s, const char* search, size_t searchsize) { return find(s, search, searchsize) < s.Handle.lnth; } bool startsWith(const string_res &s, const string_res &prefix) { return startsWith(s, prefix.Handle.s, prefix.Handle.lnth); } bool startsWith(const string_res &s, const char* prefix) { return startsWith(s, prefix, strlen(prefix)); } bool startsWith(const string_res &s, const char* prefix, size_t prefixsize) { if (s.Handle.lnth < prefixsize) { return false; } return memcmp(s.Handle.s, prefix, prefixsize) == 0; } bool endsWith(const string_res &s, const string_res &suffix) { return endsWith(s, suffix.Handle.s, suffix.Handle.lnth); } bool endsWith(const string_res &s, const char* suffix) { return endsWith(s, suffix, strlen(suffix)); } bool endsWith(const string_res &s, const char* suffix, size_t suffixsize) { if (s.Handle.lnth < suffixsize) { return false; } // Amount to offset the bytes pointer so that we are comparing the end of s // to suffix. s.bytes + offset should be the first byte to compare against suffix size_t offset = s.Handle.lnth - suffixsize; return memcmp(s.Handle.s + offset, suffix, suffixsize) == 0; } /* Back to Mike's work */ /////////////////////////////////////////////////////////////////////////// // charclass, include, exclude void ?{}( charclass_res & this, const string_res & chars) { (this){ chars.Handle.s, chars.Handle.lnth }; } void ?{}( charclass_res & this, const char * chars ) { (this){ chars, strlen(chars) }; } void ?{}( charclass_res & this, const char * chars, size_t charssize ) { (this.chars){ chars, charssize }; // now sort it ? } void ^?{}( charclass_res & this ) { ^(this.chars){}; } static bool test( const charclass_res & mask, char c ) { // instead, use sorted char list? return contains( mask.chars, c ); } int exclude(const string_res &s, const charclass_res &mask) { for (int i = 0; i < size(s); i++) { if ( test(mask, s[i]) ) return i; } return size(s); } int include(const string_res &s, const charclass_res &mask) { for (int i = 0; i < size(s); i++) { if ( ! test(mask, s[i]) ) return i; } return size(s); } //######################### VbyteHeap "implementation" ######################### // Add a new HandleNode node n after the current HandleNode node. static inline void AddThisAfter( HandleNode & this, HandleNode & n ) with(this) { #ifdef VbyteDebug serr | "enter:AddThisAfter, this:" | &this | " n:" | &n; #endif // VbyteDebug verify( n.ulink != 0p ); verify( this.ulink == n.ulink ); flink = n.flink; blink = &n; n.flink->blink = &this; n.flink = &this; #ifdef VbyteDebug { serr | "HandleList:"; serr | nlOff; for ( HandleNode *ni = HeaderPtr->flink; ni != HeaderPtr; ni = ni->flink ) { serr | "\tnode:" | ni | " lnth:" | ni->lnth | " s:" | (void *)ni->s | ",\""; for ( int i = 0; i < ni->lnth; i += 1 ) { serr | ni->s[i]; } // for serr | "\" flink:" | ni->flink | " blink:" | ni->blink | nl; } // for serr | nlOn; } serr | "exit:AddThisAfter"; #endif // VbyteDebug } // AddThisAfter // Delete the current HandleNode node. static inline void DeleteNode( HandleNode & this ) with(this) { #ifdef VbyteDebug serr | "enter:DeleteNode, this:" | &this; #endif // VbyteDebug flink->blink = blink; blink->flink = flink; #ifdef VbyteDebug serr | "exit:DeleteNode"; #endif // VbyteDebug } // DeleteNode // Allocates specified storage for a string from byte-string area. If not enough space remains to perform the // allocation, the garbage collection routine is called and a second attempt is made to allocate the space. If the // second attempt fails, a further attempt is made to create a new, larger byte-string area. static inline char * VbyteAlloc( VbyteHeap & this, int size ) with(this) { #ifdef VbyteDebug serr | "enter:VbyteAlloc, size:" | size; #endif // VbyteDebug uintptr_t NoBytes; char *r; NoBytes = ( uintptr_t )EndVbyte + size; if ( NoBytes > ( uintptr_t )ExtVbyte ) { // enough room for new byte-string ? garbage( this ); // firer up the garbage collector NoBytes = ( uintptr_t )EndVbyte + size; // try again if ( NoBytes > ( uintptr_t )ExtVbyte ) { // enough room for new byte-string ? // unimplemented feature - assert, not verify assert( 0 && "need to implement actual growth" ); // extend( size ); // extend the byte-string area } // if } // if r = EndVbyte; EndVbyte += size; #ifdef VbyteDebug serr | "exit:VbyteAlloc, r:" | (void *)r | " EndVbyte:" | (void *)EndVbyte | " ExtVbyte:" | ExtVbyte; #endif // VbyteDebug return r; } // VbyteAlloc // Move an existing HandleNode node h somewhere after the current HandleNode node so that it is in ascending order by // the address in the byte string area. static inline void MoveThisAfter( HandleNode & this, const HandleNode & h ) with(this) { #ifdef VbyteDebug serr | "enter:MoveThisAfter, this:" | & this | " h:" | & h; #endif // VbyteDebug verify( h.ulink != 0p ); verify( this.ulink == h.ulink ); if ( s < h.s ) { // check argument values // serr | "VbyteSM: Error - Cannot move byte string starting at:" | s | " after byte string starting at:" // | ( h->s ) | " and keep handles in ascending order"; // exit(-1 ); verify( 0 && "VbyteSM: Error - Cannot move byte strings as requested and keep handles in ascending order"); } // if HandleNode *i; for ( i = h.flink; i->s != 0 && s > ( i->s ); i = i->flink ); // find the position for this node after h if ( & this != i->blink ) { DeleteNode( this ); AddThisAfter( this, *i->blink ); } // if #ifdef VbyteDebug { serr | "HandleList:"; serr | nlOff; for ( HandleNode *n = HeaderPtr->flink; n != HeaderPtr; n = n->flink ) { serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; for ( int i = 0; i < n->lnth; i += 1 ) { serr | n->s[i]; } // for serr | "\" flink:" | n->flink | " blink:" | n->blink | nl; } // for serr | nlOn; } serr | "exit:MoveThisAfter"; #endif // VbyteDebug } // MoveThisAfter //######################### VbyteHeap ######################### // Move characters from one location in the byte-string area to another. The routine handles the following situations: // // if the |Src| > |Dst| => truncate // if the |Dst| > |Src| => pad Dst with blanks void ByteCopy( char *Dst, int DstStart, int DstLnth, char *Src, int SrcStart, int SrcLnth ) { for ( int i = 0; i < DstLnth; i += 1 ) { if ( i == SrcLnth ) { // |Dst| > |Src| for ( ; i < DstLnth; i += 1 ) { // pad Dst with blanks Dst[DstStart + i] = ' '; } // for break; } // exit Dst[DstStart + i] = Src[SrcStart + i]; } // for } // ByteCopy // Compare two byte strings in the byte-string area. The routine returns the following values: // // 1 => Src1-byte-string > Src2-byte-string // 0 => Src1-byte-string = Src2-byte-string // -1 => Src1-byte-string < Src2-byte-string int ByteCmp( char *Src1, int Src1Start, int Src1Lnth, char *Src2, int Src2Start, int Src2Lnth ) { #ifdef VbyteDebug serr | "enter:ByteCmp, Src1Start:" | Src1Start | " Src1Lnth:" | Src1Lnth | " Src2Start:" | Src2Start | " Src2Lnth:" | Src2Lnth; #endif // VbyteDebug int cmp; CharZip: for ( int i = 0; ; i += 1 ) { if ( i == Src2Lnth - 1 ) { for ( ; ; i += 1 ) { if ( i == Src1Lnth - 1 ) { cmp = 0; break CharZip; } // exit if ( Src1[Src1Start + i] != ' ') { // SUSPECTED BUG: this could be be why Peter got the bug report about == " " (why is this case here at all?) cmp = 1; break CharZip; } // exit } // for } // exit if ( i == Src1Lnth - 1 ) { for ( ; ; i += 1 ) { if ( i == Src2Lnth - 1 ) { cmp = 0; break CharZip; } // exit if ( Src2[Src2Start + i] != ' ') { cmp = -1; break CharZip; } // exit } // for } // exit if ( Src2[Src2Start + i] != Src1[Src1Start+ i]) { cmp = Src1[Src1Start + i] > Src2[Src2Start + i] ? 1 : -1; break CharZip; } // exit } // for #ifdef VbyteDebug serr | "exit:ByteCmp, cmp:" | cmp; #endif // VbyteDebug return cmp; } // ByteCmp // The compaction moves all of the byte strings currently in use to the beginning of the byte-string area and modifies // the handles to reflect the new positions of the byte strings. Compaction assumes that the handle list is in ascending // order by pointers into the byte-string area. The strings associated with substrings do not have to be moved because // the containing string has been moved. Hence, they only require that their string pointers be adjusted. void compaction(VbyteHeap & this) with(this) { HandleNode *h; char *obase, *nbase, *limit; NoOfCompactions += 1; EndVbyte = StartVbyte; h = Header.flink; // ignore header node for (;;) { ByteCopy( EndVbyte, 0, h->lnth, h->s, 0, h->lnth ); obase = h->s; h->s = EndVbyte; nbase = h->s; EndVbyte += h->lnth; limit = obase + h->lnth; h = h->flink; // check if any substrings are allocated within a string for (;;) { if ( h == &Header ) break; // end of header list ? if ( h->s >= limit ) break; // outside of current string ? h->s = nbase + (( uintptr_t )h->s - ( uintptr_t )obase ); h = h->flink; } // for if ( h == &Header ) break; // end of header list ? } // for } // compaction // Garbage determines the amount of free space left in the heap and then reduces, leave the same, or extends the size of // the heap. The heap is then compacted in the existing heap or into the newly allocated heap. void garbage(VbyteHeap & this ) with(this) { #ifdef VbyteDebug serr | "enter:garbage"; { serr | "HandleList:"; for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) { serr | nlOff; serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; for ( int i = 0; i < n->lnth; i += 1 ) { serr | n->s[i]; } // for serr | nlOn; serr | "\" flink:" | n->flink | " blink:" | n->blink; } // for } #endif // VbyteDebug int AmountUsed, AmountFree; AmountUsed = 0; for ( HandleNode *i = Header.flink; i != &Header; i = i->flink ) { // calculate amount of byte area used AmountUsed += i->lnth; } // for AmountFree = ( uintptr_t )ExtVbyte - ( uintptr_t )StartVbyte - AmountUsed; if ( AmountFree < ( int )( CurrSize * 0.1 )) { // free space less than 10% ? // unimplemented feature - assert, not verify assert( 0 && "need to implement actual growth" ); // extend( CurrSize ); // extend the heap // Peter says, "This needs work before it should be used." // } else if ( AmountFree > CurrSize / 2 ) { // free space greater than 3 times the initial allocation ? // reduce(( AmountFree / CurrSize - 3 ) * CurrSize ); // reduce the memory } // if compaction(this); // compact the byte area, in the same or new heap area #ifdef VbyteDebug { serr | "HandleList:"; for ( HandleNode *n = Header.flink; n != &Header; n = n->flink ) { serr | nlOff; serr | "\tnode:" | n | " lnth:" | n->lnth | " s:" | (void *)n->s | ",\""; for ( int i = 0; i < n->lnth; i += 1 ) { serr | n->s[i]; } // for serr | nlOn; serr | "\" flink:" | n->flink | " blink:" | n->blink; } // for } serr | "exit:garbage"; #endif // VbyteDebug } // garbage #undef VbyteDebug //WIP #if 0 // Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string // area is deleted. void VbyteHeap::extend( int size ) { #ifdef VbyteDebug serr | "enter:extend, size:" | size; #endif // VbyteDebug char *OldStartVbyte; NoOfExtensions += 1; OldStartVbyte = StartVbyte; // save previous byte area CurrSize += size > InitSize ? size : InitSize; // minimum extension, initial size StartVbyte = EndVbyte = new char[CurrSize]; ExtVbyte = (void *)( StartVbyte + CurrSize ); compaction(); // copy from old heap to new & adjust pointers to new heap delete OldStartVbyte; // release old heap #ifdef VbyteDebug serr | "exit:extend, CurrSize:" | CurrSize; #endif // VbyteDebug } // extend // Extend the size of the byte-string area by creating a new area and copying the old area into it. The old byte-string // area is deleted. void VbyteHeap::reduce( int size ) { #ifdef VbyteDebug serr | "enter:reduce, size:" | size; #endif // VbyteDebug char *OldStartVbyte; NoOfReductions += 1; OldStartVbyte = StartVbyte; // save previous byte area CurrSize -= size; StartVbyte = EndVbyte = new char[CurrSize]; ExtVbyte = (void *)( StartVbyte + CurrSize ); compaction(); // copy from old heap to new & adjust pointers to new heap delete OldStartVbyte; // release old heap #ifdef VbyteDebug serr | "exit:reduce, CurrSize:" | CurrSize; #endif // VbyteDebug } // reduce #endif