"Pavel Ivanov" <[email protected]> schrieb im Newsbeitrag news:[email protected]... > > Would the authors be kind to implement such optimization? > > I'm not author but I believe the answer to this question is No. > Because this fix is good enough for you but may be not good for > others. Your fix gets more memory than is really needed and it > can be a problem for embedded devices.
But that would end up with only twice the allocated Bufferspace than before (for the GroupConcat-String(s)) - and that only in the worst-case. Maybe a factor 1.5 (instead of the 2) would be a better compromise between "wasted buffer-memory" and significantly reduced realloc-calls in these typical "growing-append-space" scenarios. So, as long as sqlite3StrAccumAppend() is *not* used "pretty much everywhere, blowing up *total* mem-usage of a running SQLite-engine-instance by on average 25%", I'd vote for the change... ;-) Olaf Schmidt _______________________________________________ sqlite-users mailing list [email protected] http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
