Enhance / redesign the time module

[boggle]

• nanosleep (difficult on windows but doable with 100 ns solution using win32 API)
• function for the normalization of timespec values
• substraction, addition, comparison functions are needed
• hash
• support for printing times in fmt
• serialization helpers would be nice, too

[boggle]

http://www.c-plusplus.de/forum/43364

(win32 nanosleep)

[brson]

I don't know if we want to encourage sleep functions that actually put the thread to sleep (though obviously they are useful for some purposes). Sleeping should generally just block the task and let the runtime continue to schedule other tasks. Timers is one of the first things I was planning on implementing with libuv.

[boggle]

Ah, I didn't consider doing timers via libuv. Just thought that we should have a high resolution sleep function that matches to the timespec type. I was thinking about all things in the context of recv with timeout. Keeping this in mind, whatever sleep/timer function will be used maybe should go into the runtime for consistency of implementation.

Regarding a blocking sleep, a frequent use case I had in golang was in tasks that are locked onto a fixed thread due to C interop requirements (libzeromq bindings).

[catamorphism]

Nominating to change milestone to feature-complete.

[graydon]

time module is kinda ad-hoc at this point, needs more principled (re)design

[graydon]

accepted for backwards-compatible milestone

[huonw]

cc https://github.com/mozilla/rust/wiki/Lib-datetime and #2637

[gsingh93]

This should probably be moved to https://github.com/rust-lang/time, unless we want to add any of these features to std::time.

[steveklabnik]

Agreed.