Support millisecond granularity in logs

[optix2000]

Looks like timestamps are converted to longs. At high request rates (100+ QPS), this causes batches of balls to get sent every second instead of as a stream. This not only ruins the effect, but also causes lag as hundreds of balls are rendered in batches.

[bentcoder]

Same here. Instead of throwing the balls continuously as soon as they hit the log file, Logstalgia seems to be storing them as in batches and throws them all in one go every 10ish seconds - (a bit worse than "every second" in like @optix2000 described above).

@optix2000 Have you found a solution yet?

[bentcoder]

@acaudwell Is there a fix for this issue?

[acaudwell]

10 seconds sounds like there is external buffering going on before they reach Logstalgia so I'd look into that. For instance some commands will buffer by default and you need to change a setting to flush immediately. Logstalgia linearly distributes the release of the last seconds requests in the order they were received so you shouldn't see that kind of delay.

Maybe also try the --sync option if you're not using that.