It is generally fastest by far to treat all data as string that one reads, say, 1 mb at a time and then parse the string for discreet records/fields performing type conversions where required.
Individual read and write operations for floats will slow things down measurably.
There is indeed considerable overhead to opening and closing files. It is far faster to leave them open. There can be a drawback. Some filing operations only update the file contents during a closefile operation or when a percentage of the file buffer is reached. I seem to remember that Blitz writes data `real time` but I cannot remember for sure.
Does that help?
Individual read and write operations for floats will slow things down measurably.
There is indeed considerable overhead to opening and closing files. It is far faster to leave them open. There can be a drawback. Some filing operations only update the file contents during a closefile operation or when a percentage of the file buffer is reached. I seem to remember that Blitz writes data `real time` but I cannot remember for sure.
Does that help?