For tables that can't fit into memory we currently process 1 row at a time.
It may be faster to read in say 1M rows and process those, since there are fewer system calls.
On the other hand we'd have to allocate strings for each cell before parsing them to the output, rather than parse them directly to output and avoiding the string allocation. To be investigated.
For tables that can't fit into memory we currently process 1 row at a time. It may be faster to read in say 1M rows and process those, since there are fewer system calls. On the other hand we'd have to allocate strings for each cell before parsing them to the output, rather than parse them directly to output and avoiding the string allocation. To be investigated.