www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - SysTime at compile time

reply "Paolo Invernizzi" <paolo.invernizzi gmail.com> writes:
There's a way to use a SysTime at compile-time? It seems 
impossible right now for the presence of the TimeZone.

I want to insert some time-related fields in some structs, but I 
need to have some compile-time check with 'static assert' here 
and there...

Suggestions are welcome, thanks!
Paolo Invernizzi
Aug 23 2013
parent reply "Jonathan M Davis" <jmdavisProg gmx.com> writes:
On Friday, August 23, 2013 10:39:56 Paolo Invernizzi wrote:

 There's a way to use a SysTime at compile-time? It seems
 impossible right now for the presence of the TimeZone.
 
 I want to insert some time-related fields in some structs, but I
 need to have some compile-time check with 'static assert' here
 and there...


Getting the system time would require calling C functions (which never works 
in CTFE), so you can't do what you're trying to do with any CTFE solution, let 
alone SysTime. Even if there were no TimeZone, it couldn't be done.

You might be able to do something with __DATE__, __TIME__, or __TIMESTAMP__ 
though:

http://dlang.org/lex.html#specialtokens

However, they're generated when they're compiled, not when running a function, 
so they coludn't be generated by a CTFE function as it ran, and their values 
might stay the same across the entire build anyway, which may or may not be a 
problem depending on what you're doing. They also go to seconds at the highest 
precision, which may or may not be a problem depending on what you're doing. 
But I think that they're your only chance at getting the time at compile time.

- Jonathan M Davis
Aug 23 2013
parent reply "Paolo Invernizzi" <paolo.invernizzi gmail.com> writes:
On Friday, 23 August 2013 at 19:07:06 UTC, Jonathan M Davis wrote:

 On Friday, August 23, 2013 10:39:56 Paolo Invernizzi wrote:

 There's a way to use a SysTime at compile-time? It seems
 impossible right now for the presence of the TimeZone.
 
 I want to insert some time-related fields in some structs, but 
 I
 need to have some compile-time check with 'static assert' here
 and there...


 Getting the system time would require calling C functions 
 (which never works
 in CTFE), so you can't do what you're trying to do with any 
 CTFE solution, let
 alone SysTime. Even if there were no TimeZone, it couldn't be 
 done.


Thank you Jonathan, in reality I'm not interested in obtaining 
the current time ('Clock.currTIme' et similia), but only in 
checking some basic behaviour involving already forged times.

Something works, in a limited and constrained way, but it seems 
that there's a problem with 'Rebindable' in CTFE:

---
enum t = SysTime(0, UTC());
pragma(msg, t); // SysTime(0L, Rebindable(, (UTC("UTC", "UTC", 
"UTC"))))
// enum t2 = t + msecs(1000); // typecons.d(956): Error: cannot 
read uninitialized variable this.original in ctfe
enum d = SysTime(1000, UTC()) - SysTime(10, UTC());
pragma(msg, d); // Duration(990L)
---

Paolo Invernizzi
Aug 23 2013
parent "Jonathan M Davis" <jmdavisProg gmx.com> writes:
On Friday, August 23, 2013 22:46:34 Paolo Invernizzi wrote:

 Thank you Jonathan, in reality I'm not interested in obtaining
 the current time ('Clock.currTIme' et similia), but only in
 checking some basic behaviour involving already forged times.
 
 Something works, in a limited and constrained way, but it seems
 that there's a problem with 'Rebindable' in CTFE:
 
 ---
 enum t = SysTime(0, UTC());
 pragma(msg, t); // SysTime(0L, Rebindable(, (UTC("UTC", "UTC",
 "UTC"))))
 // enum t2 = t + msecs(1000); // typecons.d(956): Error: cannot
 read uninitialized variable this.original in ctfe
 enum d = SysTime(1000, UTC()) - SysTime(10, UTC());
 pragma(msg, d); // Duration(990L)
 ---


Rebindable should work in CTFE eventually:

http://d.puremagic.com/issues/show_bug.cgi?id=10035

However, LocalTime will never work in CTFE, because it has to call tzset. So, 
it may work eventually to create a SysTime with UTC or a SimpleTimeZone but 
not LocalTime, and unfortunately, that means that I can never make 
SysTime.init valid without adding a bunch of null checks to SysTime (which I 
don't want to do) or making it so that LocalTime always has to check whether 
tzset has been called and call it if it hasn't instead of calling it when you 
get the singleton (which I don't really want to do either). So, you might be 
able to get some of what you want eventually, but SysTime will never fully 
work at compile time with the local time zone or with getting the current 
time.

- Jonathan M Davis
Aug 23 2013