It's known that Randomize() is not effective when called rapidly... I tried to solve this using the code sample below. I can cat /dev/null (without becoming root) but when I run the code below I get an access denied RTE at 'Reset(f,1);'.. this code works fine when run as root... anyone have a suggestion?

-Alan

  procedure SuperRandomize;
    var
      f: file of cardinal;
      Card: Cardinal;
    begin
        Assign(f,'/dev/random');
        Reset(f,1);
        read(f,Card);
        RandSeed := Card;
        close(f);
    end;

Alan Mead wrote

It's known that Randomize() is not effective when called rapidly...

If you call randomize in the same process, I don't know, why you should call randomize more than once. But maybe you are calling randomize in seperate processes ...

Vincent.

Alan Mead wrote

It's known that Randomize() is not effective when called rapidly... I tried to solve this using the code sample below. I can cat /dev/null (without becoming root) but when I run the code below I get an access denied RTE at 'Reset(f,1);'.. this code works fine when run as

/dev/random isn't effective either when called permanently. The entropy pool of /dev/random is limited and /dev/random is a blocking device thus it will block after several thousand calls. /dev/urandom doesn't block but it's pseudo random as soon as /dev/random runs out of entropy.

root... anyone have a suggestion?

Do a filemode:=0; before.

Alan Mead wrote

It's known that Randomize() is not effective when called rapidly...

Why do you want to call randomize rapidly? Reading a numbers from /dev/random is semantically closer to calling random() a lot than to calling randomize() a lot.

Jonas

Hi All,
Does somebody kown the algorithm to convert a date from the double format to an explicit format?
Example: the function "now" givers me: 3.825956615197917E+004 I want to transform it in: year-month-day....

Thanks!

Jilani

Why not use DateToStr() or FormatDateTime() from sysutils ?

Michael.

Thanks for all the suggestions.

Right, if my program were to execute cntinuously while it generated these random numbers, it would be easy to just call randomize once. But my program runs very quickly and I'd like it to produce very random numbers even upon repeated execution of the program. If nothing else, it's important for my testing of it. As it stands, if I run it repeatedy without much time between runs I don't just get similar numbers, I seem to get the exact same sequence. I didn't time it precisely, but I'd say you want about a second or more to elapse between calls to Randomize() in order to get different sequences.

Hence my attempt to use /dev/urandom. I had suggested this in a discussion on this list about a year ago and I thought I would try to implement it now... I acknowledge the limitations of this approach.. maybe it would be easier to just stick a 1 sec delay before the randomize :)

BTW, I meant to ask: IIRC, about a year ago someone mentioned that 2.0 would use the Mersenne Twister method. Is that what's in 1.9.4? Last time I compared FPC's random() to MT, MT was a clear winner.

-Alan

Or, if you need the values for some sort of calculation, you have to

- find out the reference date, i. e. DateToStr(0)
- and the scale, (I think 1.0 = 1 day)
- consider the month lengths, with february changing in leap years
- for historic dates, consider calendar switches

There's a conversion function somewhere on my hard disk. If I find it, I'll post it later.

Or you might scan the strings created by DateToStr. That'd be quick-n-dirty, though :-)

Anton.

It's a lot simpler to use simply SysUtils.DecodeDate and DeccdeTime.

Uses SysUtils;
var Year, Month, Day, hour,minute,second,msec : word;
begin
  decodedate(trunc(thetdatetime),year,month,day);
  decodetime(frac(thetdatetime),hour,minute,second,msec);
end;

Or, simpler, use DecodeDateTime function from DateUtils unit. It's even Delphi-compatible. I'm not sure whether it exists in FPC 1.0.x (even if not, it's probably possible to just compile DateUtils unit sources from FPC 1.9.4 using FPC 1.0.10), but it exists in FPC 1.9.x for sure.

Correct, except for the detail that 1.9(.0), and maybe 1.9.2 don't have dateutils. It was a later addition to the 1.9.x branch.

At first sight, dateutils seems to not use 1.9.x features, so it could be that it indeed compiles under 1.0.10

The dateutils unit that comes with FPC 1.9.x contains a bazillion conversion routines.

Michael

Or, simpler, use DecodeDateTime function from DateUtils unit. It's

even Delphi-compatible. I'm not sure whether it exists in FPC 1.0.x (even if not, it's probably

possible to just compile DateUtils unit sources from FPC 1.9.4 using

FPC 1.0.10), but it exists in FPC 1.9.x for sure.
I know I can use DecodeDateTime, but I am just curious about the algorithm behind this conversion.

Jilani

Hi All,
In the next issue of Linux&C n.42 (http://www.oltrelinux.com) there will be the first article (Special) about Free Pascal (8 pages for this article, never happened before with other cool languages). Linux&C is the number one magazione in Italy dedicated to Linux with nearly 50.000 readers monthly. I am looking for a cool application (already written or to be written) to show to our readers (in the next articles) the power and the elegance of Free Pascal for serious development. Please let me know if you know about any. Let's spear Free Pascal! Jilani

Last time I compared FPC's random() to MT, MT was a clear winner.

Have you tried this ?
http://daniel.taickim.net/development/mtwister.pp

__________________________________
Do you Yahoo!?
Yahoo! Mail is new and improved - Check it out!
http://promotions.yahoo.com/new_mail

Well, I want to use the double value as a key to control a session in a web application, then I want to be able to get the explicit date from this value. So the surfer when he/she sees the number couldn't understand its meaning ;-). How ever, even if I can get the same result using other hints, I am really curious to understand the algorithm (all the frac part!).
Thank you the same.

Jilani

The TDateTime format is explained in the docs. It is a double:
- The integer part is the number of days since 00:00 1.1.1900
- The fractional part is the number of milliseconds since 00:00 of the day divided by the total number of milliseconds in a day.

Michael.

Jeff,

Thanks for this link. I found a Pascal implementation of MT somewhere .. not this one I think. Is it your code? Do you know the difference between the Real1() and Real2() etc. methods?

I was hoping that the algorithms underlying the RTL.Randomize and RTL.Random would switch to using MT in fpc version 2.

-Alan

Alan Mead wrote

I was hoping that the algorithms underlying the RTL.Randomize and RTL.Random would switch to using MT in fpc version 2.

Here's the post I recalled:

Is this still the case that 1.9.x implements MT? Hmm. As, I recall, MT beat FPC's 1.x random() in my testing using "ent"... I'll have to take a look again.

-Alan

Alan Mead wrote:
...

Reading comments in rtl/inc/system.inc FPC 1.9.x uses Mersenne Twister. Commited to CVS on 2003/10/26 by Jonas.

You could use a number from /dev/random or /dev/urandom to initialize randseed (that's just a global variable in the system unit, which is the only thing that randomize() changes when you call it), instead of randomize(). Then you'll get the best of both worlds.

Jonas

Alan Mead wrote

Thanks for this link. I found a Pascal implementation of MT somewhere .. not this one I think. Is it your code? Do you know the difference between the Real1() and Real2() etc. methods?

The 1.9.x FPC rtl already uses the Mersenne Twister. However, that doesn't change the need of having to get an initial random seed from somewhere.

Jonas

FPC itself ;)?
Lazarus =) ?
Tony's CBT?

Please let me know if you know about any. Let's spear Free Pascal! Jilani

Jilani Khaldi,

Hi All, In the next issue of Linux&C n.42 (http://www.oltrelinux.com) there will be the first article (Special) about Free Pascal (8 pages for this article, never happened before with other cool languages)

Good news, I've known a lot of fpc fans here in Italy (I'm from Trieste).

Linux&C is the number one magazine in Italy dedicated to Linux with nearly 50.000 readers monthly.

I've read your articles on FirebirdSQL you wrote for Linux&C, very clear and well written.

I am looking for a cool application (already written or to be written) to show to our readers (in the next articles) the power and the elegance of Free Pascal for serious development.

I agree with Florian, FPC itself! imho is sufficient to look at the fpc documentation to find enough examples of elegance and power. Anyway, a bunch of good Delphi libraries have been successfully ported to fpc (Synapse, ICS, Decal, Abbrevia,...), almost anything you can do with Delphi could be done in fpc as well.
I've converted the EOS code (http://sourceforge.net/projects/camelos) to fpc in a couple of days with only minor changes from Delphi and translated the Junit core testing framework from Java in a day or too: http://camelos.sourceforge.net/fpcUnit.html , so, no doubt for me that FPC is mature for serious programming.

If you need help for your article or you need a review, just ask (you can get in contact with me in private).
Ciao,
Dean Zobec

Am I allowed to punt my own apps ?

How about direqcafe ?
It's entirely written in lazarus, and is the most widely used thin-client internet cafe management tool available on Linux.
http://akinimod.sf.net

Ciao
A.J.

Okay, appology - I didn't look closely and thought this came from the lazarus mailing list. I'll leave it to the requester to decide if a lazarus app should also count as kewl fpc app :)

A.J.

Dean Zobec wrote

I've converted the EOS code (http://sourceforge.net/projects/camelos) to fpc in a couple of days with only minor changes from Delphi and translated the Junit core testing framework from Java in a day or too: http://camelos.sourceforge.net/fpcUnit.html , so,

I am interested in fpcUnit. Where can I download the source for fpcUnit. The tar.gz on http://camelos.sourceforge.net/fpcUnit.html only contains the gui test runner, not the base classes, like TestCase. I tried to find it at cvs, but couldn't.

Regards,
Vincent.

I think it would be even good to include it with FPC itself, in the FCL. As far as I know, the test classes themselves don't need any GUI functionality ?

Michael.

It's included in the EOS package:
http://prdownloads.sourceforge.net/camelos/EOS-0.9.3.tar.gz?download (in the cvs you'll find it in /EOS/Apps/fpcUnit/ ) as it uses the EOS objects and containers for xml output. Please feel free to contact me for any question about it and any suggestion is highly welcomed.
I'm planning to resume the work on the code to make it EOS independent, and it could be a good idea IMO to include it as a plugin in Lazarus like JUnit in Eclipse, but I don't have enough experience on Lazarus yet. I should work on it for a while, meanwhile any help would be greatly appreciated.

Regards,
Dean

Michael,

You're right, like in JUnit, no GUI functionality is really needed for the tests, I usually run the tests from the command line, by writing a simple console program like this:

  RegisterTests([TMyFirstTest, TMySecondTest, TMyThirdTest ....]);
  RunRegisteredTests;
  writeln(GetResultsAsXML);

see the tests of the fpcUnit itself for a complete example: http://cvs.sourceforge.net/viewcvs.py/camelos/EOS/Apps/fpcUnit/tests/runtests.dpr?rev=1.10&view=log

I'll be very happy to see it included in the FCL as a humble contribution to my preferred compiler. For the occasion I could write a command line template program for the running of the tests and insert some command line arguments to be able to choose different formats for the results (xml or plain text for example).
The code ha a dependency on the EOS library for the automatic xml serialization that is probably unnecessary and I think I can easily remove it. I can send you a streamlined FPC version by the end of the month for an evaluation.

Thank you again for your constant effort in the FPC development. Regards,
Dean

I would be grateful for that.

The code ha a dependency on the EOS library for the automatic xml serialization that is probably unnecessary and I think I can easily remove it. I can send you a streamlined FPC version by the end of the month for an evaluation.

Please do !
The XML can be generated by our own XML units, if you want XML.

As soon as I get that, I will add it to the FCL.

Thank you again for your constant effort in the FPC development.

And thank you for contributing !

Michael.