Walt,

have you tried to use "math.randomseed(x)" in order to set the seed of the built-in pseudo-random generator? You may use the current date/time in order to get a "random" "x".

Kind rgeards,

Andreas Rozek

math.random will result in a "random" number every time you do a command-R in the simulator, but if you actually relaunch the simulator, the number will always be the same. This has to do with the seed on the system being the same. Doing something like math.randomseed(os.time()) or randomseed(os.clock()) will yield a more random number.

I am having the same problem with my random not being random. What is wrong with my code?

1
2

math.randomseed(os.clock());
canToPop=math.random(numOfCans);

Same problem..

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.

Email Sent.

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?

Still waiting for a response...

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.

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;

I did some further testing and found some odd things:

1) Over a period of about 10 minutes, the first number went from 25 to 26 to 27. I was not able to reproduce what this coincided with.
2) It appears to only be related to actually calling math.randomseed and then running through a sequence. When I do not seed, I get a different sequence (and first number) ever time I run the test. So it appears this bug is directly in connection with performing a seed.

I completely agree with menace960 -- while there is a workaround, this is definitely a bug and one that comes up each time you seed. This should definitely be flagged as a priority bug since anyone using randomseed will get bitten.

Scott

(Logged as bug #134)

I just did some further testing and I can almost confirm this is a bug on compile. I took an app that we have published in the apple app store and recompiled it using the new version of Corona. On the iphone4 - math.random is no longer working as it did when I compiled the published app early December 2010.

Arg, very frustrating as I was almost finished with a new app when I spotted this bug.

Help -

@IgnacioIturra - Of course I looked through the thread, I try to practice studying a problem well enough before posting blindly into the forum and filling everyones inbox with nonsense..
I tried all methods I could find to get the function to work but it's still not working on the iphone side.
Did you try and duplicate the problem for yourself by compiling for iphone4 or are you just posting blindly without "bothering" to do the work ;)? Sorry -couldn't resist!
I just want the bug fixed -

"I've tried the math.randomseed(os.clock()); before attemting any randomization but it doesn't help."

That alone tells me you haven't. As a matter of fact I HAVE dealt with this bug during development of my game and it was fixed by following the instructions in this very thread (which you haven't properly read if you still can't fix it).

I'll reiterate just for your benefit.

math.randomseed(os.time())

at the beginning of your code. Just once (make sure it's not in a loop or called again).
then

math.random()

Once. That number will always be the same. As long as you discard it, you can use math.random again and the rest of the numbers will be "random".

If you don't do this, it'll work fine on the simulator (which is why I had so much trouble finding the bug), but won't work right on your iphone.

ignacio and/or mike, could you please post sample code on how you propose we approach this workaround? It could benefit the rest of us newbies ;)

Thanks,
RD

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;

As others have pointed out, all right from here in this very thread. :)

All you need to do is seed once with os.time() and then call random once and ignore it's value. From that point on the correct random number sequence is generated.

Scott

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

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