datetime review part 2 [Update 5]

[Dmitry Olshansky]

On 19.11.2010 8:10, Jonathan M Davis wrote:
> Most Recent: http://is.gd/hovyR
>
> Other than some minor documentation changes, all of the changes are to the
> implementation rather than the API, though those changes are fairly minimal as
> well (albeit important).
>
> I fixed it so that the stopwatch code uses a monotonic clock on both Windows and
> Linux (at least, it will once clock_gettime() is fully in druntime rather than
> commented out) and so that getting the current time doesn't use a monotonic
> clock (before Linux was using a normal or real time clock for both whereas
> Windows was using a monotonic clock for both). I also consolidated some of the
> code in static ifs to minimize code duplication.
>
> As a reminder, if you want to use the code on Windows, you need to link with
> advapi32.lib.
>
> - Jonathan M Davis
Great!
Have you found a way to deal with high redundancy in API docs?
For instance in PosInfInterval (and in the others, most likely):

template expand(D) if (__traits(compiles,begin + duration))
Expands the interval backwards in time. Effectively, it does begin -= 
duration.
Parameters:
duration    The duration to expand the interval by.
dir    The direction in time to expand the interval.
Examples:
[snip]
     nothrow void expand(D duration);
     Expands the interval backwards in time. Effectively, it does begin 
-= duration.
     ...[exactly the same as above with examples ]..

This appears to plague all the templated in such a way methods of all 
Interval structs.

Also, as  I said before,  summary table with common methods won't hurt. 
In fact it would show the implicit interface - what TP provides, most 
Intervals provide.

As to my suggested API changes, now  I have a better plan for more 
generic and easy to use API wrt ranges, just need some time to 
experiment and write it down.

-- 
Dmitry Olshansky