--- parser3/src/main/pa_string.C 2001/11/21 08:26:55 1.124 +++ parser3/src/main/pa_string.C 2002/02/20 12:35:29 1.142 @@ -1,16 +1,13 @@ /** @file Parser: string class. @see untasize_t.C. - Copyright (c) 2001 ArtLebedev Group (http://www.artlebedev.com) - Author: Alexander Petrosyan (http://paf.design.ru) + Copyright (c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com) + Author: Alexandr Petrosian (http://paf.design.ru) - $Id: pa_string.C,v 1.124 2001/11/21 08:26:55 paf Exp $ + $Id: pa_string.C,v 1.142 2002/02/20 12:35:29 paf Exp $ */ -#include "pa_config_includes.h" - #include "pcre.h" -#include "internal.h" #include "pa_pool.h" #include "pa_string.h" @@ -21,14 +18,21 @@ #include "pa_globals.h" #include "pa_table.h" #include "pa_dictionary.h" +#include "pa_charset.h" + +#define DEBUG_STRING_APPENDS_VS_EXPANDS + + +#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS +ulong string_piece_appends=0; +#endif String::String(Pool& apool, const char *src, size_t src_size, bool tainted) : Pooled(apool) { last_chunk=&head; head.count=CR_PREALLOCATED_COUNT; append_here=head.rows; - head_link=0; - link_row=&head.rows[head.count]; + initial_head_link=0; if(src) if(tainted) @@ -39,60 +43,12 @@ String::String(Pool& apool, const char * String::String(const String& src) : Pooled(src.pool()) { + last_chunk=&head; head.count=CR_PREALLOCATED_COUNT; - - uint src_used_rows=src.used_rows(); - if(src_used_rows<=head.count) { - // all new rows fit size_to preallocated area - last_chunk=&head; - uint curr_chunk_rows=head.count; - memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*src_used_rows); - append_here=&head.rows[src_used_rows]; - link_row=&head.rows[curr_chunk_rows]; - } else { - // warning: - // heavily relies on the fact - // "preallocated area is the same for all strings" - // - // info: - // allocating only enough mem to fit src string rows - // next append would allocate a new chunk - // - // new rows don't fit size_to preallocated area: splitting size_to two chunks - // preallocated chunk src to constructing head - memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*head.count); - // remaining rows size_to new_chunk - uint curr_chunk_rows=src_used_rows-head.count; - last_chunk=static_cast( - malloc(sizeof(uint)+sizeof(Chunk::Row)*curr_chunk_rows+sizeof(Chunk *), 9)); - last_chunk->count=curr_chunk_rows; - head_link=last_chunk; - append_here=link_row=&last_chunk->rows[last_chunk->count]; - - Chunk *old_chunk=src.head_link; - Chunk::Row *new_rows=last_chunk->rows; - uint rows_left_to_copy=last_chunk->count; - while(true) { - uint old_count=old_chunk->count; - Chunk *next_chunk=old_chunk->rows[old_count].link; - if(next_chunk) { - // not last source chunk - // taking it all - memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*old_count); - new_rows+=old_count; - rows_left_to_copy-=old_count; + append_here=head.rows; + initial_head_link=0; - old_chunk=next_chunk; - } else { - // the last source chunk - // taking only those rows of chunk that _left_to_copy - memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*rows_left_to_copy); - break; - } - } - } - link_row->link=0; - src_used_rows; + append(src, UL_UNSPECIFIED); } size_t String::size() const { @@ -114,20 +70,25 @@ break2: return result; } void String::expand() { - uint new_chunk_count=last_chunk->count+CR_GROW_COUNT; - if(new_chunk_count>MAX_USHORT) - new_chunk_count=MAX_USHORT; - - last_chunk=static_cast( - malloc(sizeof(uint)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *), 10)); - last_chunk->count=new_chunk_count; - link_row->link=last_chunk; + Chunk::count_type new_chunk_count=last_chunk->count+CR_GROW_COUNT; + if(new_chunk_count>max_integral(Chunk::count_type)) + new_chunk_count=max_integral(Chunk::count_type); + + Chunk *new_chunk=static_cast( + malloc(sizeof(Chunk::count_type)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *), 10)); + new_chunk->rows[new_chunk->count=new_chunk_count].link=0; + last_chunk->rows[last_chunk->count].link=new_chunk; + + last_chunk=new_chunk; append_here=last_chunk->rows; - link_row=&last_chunk->rows[last_chunk->count]; - link_row->link=0; } String& String::real_append(STRING_APPEND_PARAMS) { + if(!last_chunk) // growth stopped [we're appended as string to somebody] + throw Exception(0, 0, + this, + "string growth stopped (append cstr)"); + if(!src) return *this; if(!size) @@ -135,12 +96,17 @@ String& String::real_append(STRING_APPEN if(!size) return *this; - while(size>MAX_USHORT) { +#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS + string_piece_appends++; +#endif + + // manually unrolled to avoid extra check + while(size>max_integral(Chunk::Row::item_size_type)) { if(chunk_is_full()) expand(); append_here->item.ptr=src; - append_here->item.size=MAX_USHORT; + append_here->item.size=max_integral(Chunk::Row::item_size_type); append_here->item.lang=lang; #ifndef NO_STRING_ORIGIN append_here->item.origin.file=file; @@ -148,8 +114,8 @@ String& String::real_append(STRING_APPEN #endif append_here++; - src+=MAX_USHORT; - size-=MAX_USHORT; + src+=max_integral(Chunk::Row::item_size_type); + size-=max_integral(Chunk::Row::item_size_type); } if(chunk_is_full()) @@ -168,7 +134,7 @@ String& String::real_append(STRING_APPEN } char String::first_char() const { - if(!used_rows()) + if(is_empty()) throw Exception(0, 0, this, "getting first char of empty string"); @@ -189,7 +155,8 @@ break2: int String::cmp(int& partial, const String& src, size_t this_offset, Untaint_lang lang) const { partial=-1; - this_offset=min(this_offset, size()-1); + size_t a_size=size(); + this_offset=min(this_offset, a_size-1); const Chunk *a_chunk=&head; const Chunk *b_chunk=&src.head; @@ -204,11 +171,11 @@ int String::cmp(int& partial, const Stri int result; size_t pos=0; - bool a_break=size()==0; - bool b_break=src.size()==0; + bool a_break=a_size==0; + bool b_break=src.is_empty(); if(!(a_break || b_break)) while(true) { if(pos+a_row->item.size > this_offset) { - if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang) + if(lang!=UL_UNSPECIFIED && a_row->item.lang>lang) return -1; // wrong lang -- bail out int size_diff= @@ -277,8 +244,9 @@ int String::cmp(int& partial, const Stri int String::cmp(int& partial, const char* b_ptr, size_t src_size, size_t this_offset, Untaint_lang lang) const { partial=-1; + size_t a_size=size(); size_t b_size=src_size?src_size:b_ptr?strlen(b_ptr):0; - this_offset=min(this_offset, size()-1); + this_offset=min(this_offset, a_size-1); const Chunk *a_chunk=&head; const Chunk::Row *a_row=a_chunk->rows; @@ -288,11 +256,11 @@ int String::cmp(int& partial, const char uint a_countdown=a_chunk->count; size_t pos=0; - bool a_break=size()==0; + bool a_break=a_size==0; bool b_break=b_size==0; if(!(a_break || b_break)) while(true) { if(pos+a_row->item.size > this_offset) { - if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang) + if(lang!=UL_UNSPECIFIED && a_row->item.lang>lang) return -1; // wrong lang -- bail out int size_diff= @@ -346,7 +314,7 @@ int String::cmp(int& partial, const char #ifndef NO_STRING_ORIGIN const Origin& String::origin() const { - if(!used_rows()) { + if(is_empty()) { static const Origin empty_origin={"empty string"}; return empty_origin; } @@ -365,7 +333,7 @@ const Origin& String::origin() const { String& String::mid(size_t start, size_t finish) const { String& result=*NEW String(pool()); - start=max(0, start); + start=max(size_t(0), start); finish=min(size(), finish); if(start==finish) return result; @@ -374,7 +342,7 @@ String& String::mid(size_t start, size_t STRING_FOREACH_ROW( size_t item_finish=pos+row->item.size; if(item_finish > start) { // started now or already? - bool started=result.size()==0; // started now? + bool started=result.is_empty(); // started now? bool finished=finish <= item_finish; // finished now? size_t offset=started?start-pos:0; size_t size=finished?finish-pos:row->item.size; @@ -395,7 +363,8 @@ break2: int String::pos(const String& substr, int result, Untaint_lang lang) const { - for(; resultrows; - for(uint i=0; icount; i++, row++) { + for(Chunk::count_type i=0; icount; i++, row++) { if(row==append_here) goto break2; @@ -674,7 +645,7 @@ void String::join_chain(Pool& pool, if(row->item.lang==joined_lang) joined_size+=row->item.size; else - break; + goto break21; } if(chunk=row->link) { start_i=0; @@ -689,8 +660,12 @@ break21:; if(joined_size==arow->item.size) { joined_ptr=arow->item.ptr; ai++; arow++; - if(ai==achunk->count) - achunk=arow->link; + if(ai==achunk->count) { + if(achunk=arow->link) { + ai=0; + arow=achunk->rows; + } + } } else { // join adjacent rows char *ptr=(char *)pool.malloc(joined_size,13); @@ -710,7 +685,7 @@ break21:; memcpy(ptr, row->item.ptr, row->item.size); ptr+=row->item.size; } else - break; + goto break22; } if(chunk=row->link) { start_i=0; @@ -731,26 +706,27 @@ String& String::reconstruct(Pool& pool) //_asm int 3; String& result=*new(pool) String(pool); const Chunk *chunk=&head; - do { - const Chunk::Row *row=chunk->rows; - for(uint i=0; icount; ) { - if(row==append_here) - goto break2; + const Chunk::Row *row=chunk->rows; + for(uint i=0; icount; ) { + if(row==append_here) + break; - uchar joined_lang; - const char *joined_ptr; - size_t joined_size; - join_chain(pool, i, chunk, row, - joined_lang, joined_ptr, joined_size); + uchar joined_lang; + const char *joined_ptr; + size_t joined_size; +#ifndef NO_STRING_ORIGIN + const char *joined_origin_file=row->item.origin.file; + const size_t joined_origin_line=row->item.origin.line; +#endif + join_chain(pool, i, chunk, row, + joined_lang, joined_ptr, joined_size); - result.APPEND(joined_ptr, joined_size, - joined_lang, - row->item.origin.file, row->item.origin.line); - if(!chunk) - goto break2; - } - } while(true); -break2: + result.APPEND(joined_ptr, joined_size, joined_lang, + joined_origin_file, joined_origin_line); + + if(!chunk) + break; + } return result; }; @@ -848,6 +824,13 @@ int String::as_int() const { return result; } +inline void ushort2uchars(ushort word, uchar& byte1, uchar& byte2) { + byte1=word&0xFF; + byte2=word>>8; +} +inline ushort uchars2ushort(uchar byte1, uchar byte2) { + return (byte2<<8) | byte1; +} /* @todo maybe network order worth spending some effort? don't bothering myself with network byte order, am not planning to be able to move resulting file across platforms @@ -856,18 +839,20 @@ int String::as_int() const { void String::serialize(size_t prolog_size, void *& buf, size_t& buf_size) const { buf_size= prolog_size - +used_rows()*(sizeof(Untaint_lang)+sizeof(size_t)) + +used_rows()*(sizeof(uchar)+sizeof(ushort)) +size(); buf=malloc(buf_size,15); char *cur=(char *)buf+prolog_size; STRING_FOREACH_ROW( // lang - memcpy(cur, &row->item.lang, sizeof(Untaint_lang)); - cur+=sizeof(Untaint_lang); + memcpy(cur, &row->item.lang, sizeof(uchar)); + cur+=sizeof(uchar); // size - memcpy(cur, &row->item.size, sizeof(size_t)); - cur+=sizeof(size_t); + uchar byte1; uchar byte2; + ushort2uchars(row->item.size, byte1, byte2); + memcpy(cur, &byte1, sizeof(uchar)); cur+=sizeof(uchar); + memcpy(cur, &byte2, sizeof(uchar)); cur+=sizeof(uchar); // bytes memcpy(cur, row->item.ptr, row->item.size); cur+=row->item.size; @@ -875,31 +860,26 @@ void String::serialize(size_t prolog_siz break2: ; } - -/* @todo maybe network order worth spending some effort? - don't bothering myself with network byte order, - am not planning to be able to move resulting file across platforms - for now -*/ -#ifndef DOXYGEN -struct Serialized_piece { - String::Untaint_lang lang; - size_t size; - char ptr[1]; -}; -#endif - void String::deserialize(size_t prolog_size, void *buf, size_t buf_size, const char *file) { - char *cur=((char *)buf)+prolog_size; + if(buf_size<=prolog_size) + return; + + char *cur=(char *)buf+prolog_size; buf_size-=prolog_size; while(buf_size) { - Serialized_piece& p=*(Serialized_piece *)cur; - APPEND(p.ptr, p.size, p.lang, file, 0); + uchar lang=*(uchar *)(cur); + + ushort size=uchars2ushort( + *(uchar*)(cur+sizeof(uchar)*1), + *(uchar*)(cur+sizeof(uchar)*2) + ); + + const char *ptr=(const char*)(cur+sizeof(uchar)*3); + APPEND(ptr, size, lang, file, 0); - size_t piece_size=sizeof(p.lang)+sizeof(p.size)+p.size; + size_t piece_size=sizeof(uchar)+sizeof(ushort)+size; cur+=piece_size; buf_size-=piece_size; } } -