www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - Epoch time + msecs

reply Handyman <handyman handyworld.tv> writes:
How to get current time as a float (or a double or a real) as a 
Unix epoch + milliseconds (e.g, 1447437383.465, or even 
1447437383.46512 with finer resolution)?   I read 
http://dlang.org/intro-to-datetime.html and the docs of course.  
I came this far

    auto ct = Clock.currTime();
    auto milliseconds = ct.fracSec.msecs;
    auto epoch = ct.toUnixTime();

But I think this is not the shortest way and I don't know how to 
combine and cast into a float (or a double or a real).

My goal was to multiply by 1000 and feed as this number as a seed 
to D's random number generator.  But there may be beter ways to 
make a seed.
Nov 13 2015
next sibling parent reply Steven Schveighoffer <schveiguy yahoo.com> writes:
On 11/13/15 1:00 PM, Handyman wrote:

 How to get current time as a float (or a double or a real) as a Unix
 epoch + milliseconds (e.g, 1447437383.465, or even 1447437383.46512 with
 finer resolution)?   I read http://dlang.org/intro-to-datetime.html and
 the docs of course. I came this far

     auto ct = Clock.currTime();
     auto milliseconds = ct.fracSec.msecs;
     auto epoch = ct.toUnixTime();


The toUnixTime call is going to truncate all the subseconds off.

What I would do is this:

auto t = (Clock.currTime() - SysTime.fromUnixTime(0)).total!"msecs" / 1000.0

Which is essentially what toUnixTime would do (but the fromUnixTime(0) 
is a constant internally).

-Steve
Nov 13 2015
parent reply Handyman <handyman handyworld.tv> writes:
On Friday, 13 November 2015 at 18:27:42 UTC, Steven Schveighoffer 
wrote:

 On 11/13/15 1:00 PM, Handyman wrote:
 What I would do is this:

 auto t = (Clock.currTime() - 
 SysTime.fromUnixTime(0)).total!"msecs" / 1000.0


Thanks, Steve.

    import std.stdio;
    import std.datetime;

    void main() {
       auto t = (Clock.currTime() - 
SysTime.fromUnixTime(0)).total!"msecs" / 1000.0;
       writeln(t);
    }

Gives an error: Error: no property 'fromUnixTime' for type 
'SysTime'.

When I do

    auto t = (Clock.currTime() - 
unixTimeToStdTime(0)).total!"msecs" / 1000.0;

I get
Error: incompatible types for ((currTime(opCall())) - 
(unixTimeToStdTime(0))): 'SysTime' and 'long'

and when I do

auto t = (Clock.currTime() - 
SysTime.fromUnixTime(0)).total!"msecs" / 1000.0;

I get:
Error: no property 'fromUnixTime' for type 'SysTime'

I think I am spoiled by 'typeless' scriping languages.
Nov 13 2015
parent Steven Schveighoffer <schveiguy yahoo.com> writes:
On 11/13/15 3:15 PM, Handyman wrote:

 On Friday, 13 November 2015 at 18:27:42 UTC, Steven Schveighoffer wrote:

 On 11/13/15 1:00 PM, Handyman wrote:
 What I would do is this:

 auto t = (Clock.currTime() - SysTime.fromUnixTime(0)).total!"msecs" /
 1000.0


 Thanks, Steve.

     import std.stdio;
     import std.datetime;

     void main() {
        auto t = (Clock.currTime() -
 SysTime.fromUnixTime(0)).total!"msecs" / 1000.0;
        writeln(t);
     }

 Gives an error: Error: no property 'fromUnixTime' for type 'SysTime'.


gah, fromUnixTime seems to have been added in 2.069. If you cannot 
update to the latest compiler, this should work:

auto t = (Clock.currTime() - 
SysTime(unixTimeToStdTime(0))).total!"msecs" / 1000.0

-Steve
Nov 13 2015
prev sibling parent reply Jonathan M Davis via Digitalmars-d-learn writes:
On Friday, November 13, 2015 18:00:13 Handyman via Digitalmars-d-learn wrote:

 How to get current time as a float (or a double or a real) as a
 Unix epoch + milliseconds (e.g, 1447437383.465, or even
 1447437383.46512 with finer resolution)?   I read
 http://dlang.org/intro-to-datetime.html and the docs of course.
 I came this far

     auto ct = Clock.currTime();
     auto milliseconds = ct.fracSec.msecs;
     auto epoch = ct.toUnixTime();

 But I think this is not the shortest way and I don't know how to
 combine and cast into a float (or a double or a real).

 My goal was to multiply by 1000 and feed as this number as a seed
 to D's random number generator.  But there may be beter ways to
 make a seed.


Well, there's what Steven showed you if you really want to use unix time,
but if all you care about is getting a value for a seed, then you could just
use stdTime. e.g.

auto seed = Clock.currTime().stdTime;

It'll be in hecto-nanoseconds, so you should get more variation than with
unix time as milliseconds, and you don't have to do any math.

Alternatively, std.random.unpredictable seed is there specifically to
provide an unpredictable seed. It's probably best to just use that if you
don't have a good reason not to.

- Jonathan M Davis
Nov 13 2015
parent reply Handyman <handyman handyworld.tv> writes:
On Friday, 13 November 2015 at 23:12:34 UTC, Steven Schveighoffer 
wrote:

 If you cannot update to the latest compiler


Didn't realised this feature is so new to the compiler.  In my 
case, it is no problem to update the compiler, and I will.  
Thanks!

On Saturday, 14 November 2015 at 03:10:17 UTC, Jonathan M Davis 
wrote:

 auto seed = Clock.currTime().stdTime;
 (...) and you don't have to do any math.


Perfect.

On Saturday, 14 November 2015 at 03:10:17 UTC, Jonathan M Davis 
wrote:

 Alternatively, std.random.unpredictable seed is there 
 specifically to provide an unpredictable seed. It's probably 
 best to just
 use that if you don't have a good reason not to.


Of course.  That's why I mentioned my purpose of using 
Clock.currTime(), in the hope I got corrected in using the right 
and offical seed method which I failed to find, which brings me 
to another point.

The D docs seem very thorough and complete to me but less 
accessible in comparison to, e.g., Perl docs, Php docs, or Ruby 
docs.  In particular I have difficulties in understanding the 
headers of the standard library function specifications (e.g.,: 
auto uniform(string boundaries = "[)", T1, T2, 
UniformRandomNumberGenerator)(T1 a, T2 b, ref 
UniformRandomNumberGenerator urng) if 
(isFloatingPoint!(CommonType!(T1, T2)) && 
isUniformRNG!UniformRandomNumberGenerator);)

These headers are necessary and give loads of information. D is 
no toy language, that's for sure.  Maybe a tutorial should be 
written on how to read such headers and how to read the standard 
library docs in general.  I am making notes for myself, but these 
are of too little quality to be used in the public domain.

Like I said: I am spoiled by scripting languages.

Thanks again, your answers have helped me a lot.
Nov 14 2015
next sibling parent Nicholas Wilson <iamthewilsonator hotmail.com> writes:
On Saturday, 14 November 2015 at 12:14:42 UTC, Handyman wrote:


 The D docs seem very thorough and complete to me but less 
 accessible in comparison to, e.g., Perl docs, Php docs, or Ruby 
 docs.  In particular I have difficulties in understanding the 
 headers of the standard library function specifications (e.g.,: 
 auto uniform(string boundaries = "[)", T1, T2, 
 UniformRandomNumberGenerator)(T1 a, T2 b, ref 
 UniformRandomNumberGenerator urng) if 
 (isFloatingPoint!(CommonType!(T1, T2)) && 
 isUniformRNG!UniformRandomNumberGenerator);


  auto // the return type is deduced
  uniform // the name of the function
(string boundaries = "[)", T1, T2, URNG) // the set if compile 
time parameters
// a string that defaults to "[)" denoting a half open interval 
and three named types T1, T2 and URNG
(T1 a, T2 b, ref URNG urng) // the set of runtime parameters a T1 
, a T2 and an URNG taken by reference
  if  (isFloatingPoint!(CommonType!(T1, T2)) && isUniformRNG!RNG); 
// a set of constraints to make sure you don't use the wrong 
types with it.
Nov 14 2015
prev sibling parent reply Kagamin <spam here.lot> writes:
On Saturday, 14 November 2015 at 12:14:42 UTC, Handyman wrote:

 Of course.  That's why I mentioned my purpose of using 
 Clock.currTime(), in the hope I got corrected in using the 
 right and offical seed method which I failed to find, which 
 brings me to another point.


Don't docs provide examples on how use (and seed) a generator?
Nov 16 2015
parent reply Marc =?UTF-8?B?U2Now7x0eg==?= <schuetzm gmx.net> writes:
On Monday, 16 November 2015 at 10:29:25 UTC, Kagamin wrote:

 On Saturday, 14 November 2015 at 12:14:42 UTC, Handyman wrote:

 Of course.  That's why I mentioned my purpose of using 
 Clock.currTime(), in the hope I got corrected in using the 
 right and offical seed method which I failed to find, which 
 brings me to another point.


 Don't docs provide examples on how use (and seed) a generator? 



Hmmm... why is `unpredictableSeed` only a `uint`? Surely most 
PRNGs have more than 32 bits of internal state?
Nov 16 2015
parent reply Kagamin <spam here.lot> writes:
On Monday, 16 November 2015 at 10:47:36 UTC, Marc Schütz wrote:

 Hmmm... why is `unpredictableSeed` only a `uint`? Surely most 
 PRNGs have more than 32 bits of internal state?


Maybe it's only worth 32 random bits? There's a rangified example 
too:
Nov 16 2015
parent reply Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Monday, 16 November 2015 at 13:34:41 UTC, Kagamin wrote:

 On Monday, 16 November 2015 at 10:47:36 UTC, Marc Schütz wrote:

 Hmmm... why is `unpredictableSeed` only a `uint`? Surely most 
 PRNGs have more than 32 bits of internal state?


 Maybe it's only worth 32 random bits? There's a rangified 
 example too: 



No, to get a period of 2^19937 you need 19937 bits or 2-3 KiB… ;)
Nov 16 2015
parent Kagamin <spam here.lot> writes:
On Monday, 16 November 2015 at 14:21:02 UTC, Ola Fosheim Grøstad 
wrote:

 No, to get a period of 2^19937 you need 19937 bits or 2-3 KiB… 
 ;)


This computes out of context :) but...
Nov 16 2015