Abdelrazak Younes wrote:
Juergen,There is something fishy in this method. set<char_type> Encoding::getSymbolsList() const { // assure the used encoding is properly initialized init(); // first all encodable characters CharSet symbols = encodable_; // add those below start_encodable_ for (char_type c = 0; c < start_encodable_; ++c) symbols.insert(c); // now the ones from the unicodesymbols file CharInfoMap::const_iterator const end = unicodesymbols.end(); CharInfoMap::const_iterator it = unicodesymbols.begin(); for (; it != end; ++it) symbols.insert(it->first); return symbols; }The lengthy initialization for utf8* encoding is not due to iconv and this is normal as we shouldn't have to do a lookup for utf8 encodable characters, all of them are. No, the problem lies is when we insert the symbols from the unicodesymbols file. For utf8, we shouldn't do that because _all_ symbols are already in there. On each insertion, std::set() has to search if the given symbols is not already present; as you have 1114112 symbols...
Well the initial insertion from 0 to ucs4_max is also quite lengthy of course. I guess the solution is just to use a vector instead of a set.
Abdel.
