Elapsed milliseconds are zero

[pnmcosta]

Hi @hryz,

I've now completed the integration of the package in both .net framework and .net core samples and it's working as expected, thank you for all your help.

Only issue i'm coming across is with the Elapsed times, these are zero with the quickest queries, I believe this is because of the Stopwatch precision.

For a more precise measure, please see https://github.com/serilog/serilog-aspnetcore/blob/dev/src/Serilog.AspNetCore/AspNetCore/RequestLoggingMiddleware.cs#L55

Thanks,
P.

[hryz]

Hi @pnmcosta,

I've checked the type definitions and see a problem:
IHookedDbConnectionFactory<T> and ISqlHooks<T> are public interfaces.
In order to have a sub-millisecond value in the logging messages, we have to change the type of elapsedMs from long to double. This would be a breaking change. If someone implemented their own implementation of hooks (for example for telemetry) after upgrade of the package their code will be broken.
Also, a double value will have ~16 digits in the log message. The default format must be specified to limit them.

In my opinion, if a query execution time is less than 1 ms, it doesn't really matter how much less, unless it's a microbenchmark like Benchmark.Net (but they are dealing with micro and nanoseconds). The typical application of this lib is debugging and testing the system, not benchmarking.

If you don't mind, I would not introduce a breaking change for extra precision.
However, in the next major version, it should be changed, I think. The app metrics/ telemetry scenario will benefit from this for sure.

[pnmcosta]

Yup, totally understand.

As our library is new and we require that level of precision, I'll do the change on the fork and pack that instead.

I'll create a PR for you to integrate if you'd like, i'll try to keep it up-to-date as much as possible and match your version

Thanks so much,
P.