use os.time() instead of os.clock()

math.randomseed( os.time() )

carlos

Same problem..

what is value for numOfCans?

Here is the semantics for math.random

math.random(fooValue) generates integer numbers between 1 and upper. (numOfCans?)
math.random(foolowerValue, fooupperValue) generates integer numbers between lower and upper.

numOfCans is a tonumber(event.id) from a button press. tpye(numOfCans) is number when tested.

The result is supposed to be a number between 1 and numOfCans.

The result is in the correct range, its just always the same pattern of numbers. It seems the randomseed is not working.

send me some code...

info@anscamobile.com

c

Email Sent.

Got it, replied.

math.randomseed( os.time() ) -- needed to be placed at the top of the file.

The issue was randomseed was called everytime right before math.random was being called, thus not seeding the right random number.

c

Hmm.. this bug is back.. Fixed another piece of code that was masking this problem (buttons were being called multiple times on single button presses)

Can I send you the code again?

yes, please send code to info@anscamobile.com, or suppor@anscamobile.com

carlos

Still waiting for a response...

This may sound crazy, but what happens if you call a few random numbers (and, say, print them to the terminal). Is the SEQUENCE the same each time?

I ask because I think I've hit cases where the first "random" number tended to come out the same each time, even with a random seed, but then later numbers in the series would have more random-looking values. So I just discarded the first number and used the second one.

However, if you're getting entire identical sequences, then something else is wrong.

Since I've seen a few examples (not here, but in other places) where folks set the random seed multiple times, I wanted to point out you only want to set it *once*. Setting it more than once just screws with the algorithm. Typically, one would set the seed with a call to "time(0)" as that is, on most systems, the number of seconds since the "epoch" (when unix first came to life). Now, I could be wrong, but this is, in my experience, still a very standard library call and in lua it's available with "os.time()". It does depend on the hardware you're on, but the iPhone definitely supports this convention and I would suggest that any platform we would be using Corona on will as well. Using os.clock() is not a reliable thing since it's very conceivable you'll get the same number (or very close to the same number) each time you run your app since it's the number of CPU seconds used since starting your app... whereas "os.time()" will always (unless called again within the same second) be different. And since you only seed *once*, this should never be an issue.

The "pseudo-random" algorithm, simply put, generates a pseudo-random sequence of numbers. Calling random just grabs the next one in the sequence. Setting seed more than once will just reset the sequence and actually give you less randomness.

What a lot of folks also don't know (and why the OP considered the behavior to be a bug) is the same seed will generate the *same* sequence of numbers! This is a very good thing! :) Say you randomly generate a dungeon or a set of obstacles in your game. Well, you can now "save" that level simply by storing your random seed you used to generate it... instead of saving the entire level data. To recreate it, just set your seed and run through the same algorithm. Of course, this will only work if you don't change how you generate your level, but for certain things it's quite nice.

At any rate, set the seed once, then call random multiple times.

In the latter part of this thread the indication was it generated the same result. This is most likely exactly what evank was saying: You're likely just encountering the same starting numbers. If you are taking random(1,5) you have a very small sample set and will end up with similar sequences (or at least the feel of similar sequences) no matter what was being used to generate "random" numbers since you literally only have 5 numbers to choose from.

Hope this helps a little bit in understanding how random works.

I am well aware of how random numbers and seeds work. There was a bug in my code that was changing the seed each time, but that still should not have given me the same "random" number every time in the original version I had thrown together for testing. Even when the random seed was put in the right spot (out of the loop), the first random number was always the same, and as my game loop never needed a second random, I never got anything but the first number, which was always the same. After discussing with the developers, it seems that a simple discard of the first random number will allow the following randoms to be correct.

I still consider this a bug, but one I can work around.

(I am not some beginner programmer, I am just new to Corona and LUA. I have my BS and MS in Computer Science and work as a developer for business applications)

No need to get defensive, more than just you read these bug reports and there was enough conjecture here (as well as in other places) that I wanted to post about not seeding multiple times... just to be sure. I don't know you, I couldn't tell your level of expertise.

I, also, recently ran into the random bug you point out (it is indeed a bug) and it was very simple to reproduce.

So, devs, here you go: main.lua ::

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

-- without seeding, you get a different sequence and different starting number every run
print("No Seed Set");
for i = 1, 10 do
  print(math.random(1, 100));
end;
<code>
-- as soon as you seed, you get the same starting number
_seed = os.time();
print("Seed: ".._seed);
math.randomseed(_seed);
for i = 1, 10 do
  print(math.random(1, 100));
end;
<code>
-- and again to demonstrate the same problem each re-seed
_seed = _seed + 1;
print("Seed: ".._seed);
math.randomseed(_seed);
for i = 1, 10 do
  print(math.random(1, 100));
end;
<code>
-- finally, the workaround
_seed = _seed + 1;
print("Seed: ".._seed);
math.randomseed(_seed);
math.random(); -- <<<<< this is the work around...ick.
for i = 1, 10 do
  print(math.random(1, 100));
end;

But, but, but.., I just did that.

It's really no more complicated than having this at the start of your main.lua

1 2	math.randomseed(os.time()) math.random()

Post #17 from this thread gives a decent run down on how to actually seed and use random correctly:
http://developer.anscamobile.com/forum/2010/01/05/mathrandom-bug#comment-1417

Post #19 gives example code which will show the problem and the last snippet shows the work around. This is also specified in the comment right before yours.
http://developer.anscamobile.com/forum/2010/01/05/mathrandom-bug#comment-1439

For clarification:

1 2 3 4 5 6 7 8 9

_seed = os.time(); -- finally, the workaround -- _seed = _seed + 1; commented out as it is not germane to this snippet print("Seed: ".._seed); math.randomseed(_seed); math.random(); -- <<<<< this is the work around...ick. for i = 1, 10 do   print(math.random(1, 100)); end;

Yeah, I saw (glimpse) at the posts you mentioned, but it didn't clicked.
Sometimes devs don't realize that new blood is coming to the forums and are learning (even experience ones).

New devs are also looking for 'friendly forums'; some are afraid of asking question for fear of sounding "dumb" and leave.
Take a look at what menace690 posted, he left because he thought it was bug that was never going to be fixed.

Keep in mind that if potential corona users don't buy the software because they don't feel "right", that's money that Ansca is losing; and if Ansca doesn't make money, corona could well be extinct (I'm exaggerating, but it has happened )

All this long thread could've been avoided (perhaps at post #17) if something as simple as the (but, but, but) was posted earlier.

Sorry for the rumble.

Thanks and cheers ;)

RD

Yes, this problem has has been around for years. Anyone who's relied on the rand() function in libc (which is what Lua relies on) has seen this problem. Here's a note from one of the creators of Lua documenting the issue almost 4 years ago:

http://lua-users.org/lists/lua-l/2007-03/msg00564.html

According to this post:
http://lua-users.org/wiki/MathLibraryTutorial

... OSX and FreeBSD have issues.

The solution proposed was based on the idea that os.time() returns a relatively similar number each time. For example, here are two os.time() values separated by roughly one day (A day has 86400 seconds)

1 2	1327302026 -- today 1327388297 -- some time tomorrow

Thanks for that, I too have seen that random numbers aren't quite as random as I'd like.

As an aside, I've been using this library for my games which require procedural generation - it will give the same set of random numbers for a given seed across all platforms.

http://developer.anscamobile.com/code/portable-seedable-random-number-generator

views:2121 update：2012/2/10 9:26:24