relative time as timestamp

[arvidn]

The message timestamp is expressed as a relative time duration (std::chrono::milliseconds) here.

That seems like the wrong type, it really should be a time_point, right?

[mfontanini]

The timestamp is expressed as a relative time since epoch. So it's arguably the wrong type but not necessarily.

[rriggs]

So, we expect everyone to write:

.timestamp(duration_cast<milliseconds>(system_clock::now() - system_clock::from_time_t(0)))

Whenever they need a timestamp? That seems less than ideal.

[accelerated]

.timestamp(duration_cast<milliseconds>(std::chrono::system_clock::now().time_since_epoch())) should do.

[accelerated]

.timestamp(duration_cast<milliseconds>(std::chrono::system_clock::now().time_since_epoch())) should do.

But I agree that .timestamp(std::chrono::system_clock::now()) would be more user-friendly. I wonder if @mfontanini is ok with adding an overload to this function taking a time_point?

[mfontanini]

Yeah that should work.

[mfontanini]

Also, note that you don't have to specify anything if you want now to be the timestamp in the message. Messages will have now as the timestamp by default so you should only specify one if you want to override it (e.g. you want the timestamp to be something that comes out of the message's contents or something like that).

[accelerated]

Yeah that should work.

@mfontanini, You mean adding the overload or the code snipped I posted?

[mfontanini]

Yep, I was referring to that comment.

[mfontanini]

Sorry, I meant your code snippet :)

[mfontanini]

I mean, adding the overload. Sorry, for the spam!

[accelerated]

See PR#128.