7

Today, I was just told about the seed() function from a programmer much more experienced than me. I normally just call choice() with a list as an argument, as I don't need anymore random number functionality than that.

My programmer friend told me that calling seed is necessary because otherwise Python always begins random number operations with zero as the default seed. This means that although the numbers appear random, we are really getting the same sequence every time.

This strikes me as rather odd. Does the choice() function, for example, really not call seed before it does its thing? Or is the reason it can't programmatically change its seed because that in of itself would involve picking a random number, and obviously that it is a bit of a problem if our end goal is also to pick a random number!

I'm ranting a bit here, but I'm wondering if anyone has a clear idea of how all this was implemented.

Karl Knechtel
  • 62,466
  • 11
  • 102
  • 153
Josh Imhoff
  • 6,746
  • 3
  • 13
  • 10
  • 2
    These are all things you can learn in the documentation ( http://docs.python.org/library/random.html), or by testing it out yourself in the interpreter. Try running Python and doing `random.choice(range(1000))` twice. Do you get the same answer? – Michael Hoffman Dec 07 '11 at 22:10
  • @MichaelHoffman you've missed the point. Of course you wouldn't get the same answer by doing it twice in the same Python process. The correct test is to open two processes and try it in each. – Karl Knechtel Dec 07 '11 at 22:19
  • I wrote that poorly. I meant that the whole operation (running Python and performing the function) should be done twice. – Michael Hoffman Dec 07 '11 at 22:22
  • 1
    Am I doing something wrong? Win7x64 Py2.7 Step 1: `Ctrl-R` -> `Python` `Enter`. Step 2: `import random; random.choice(range(1000))` Step 3: repeat steps 1 and 2. I thought it would be different every time, but it seems like it's the same thing. update: Oh, if I set the seed to `0` everytime, I get the same answers. Guess it chooses random seeds each time. – TankorSmash Jul 27 '12 at 16:46

1 Answers1

29

Your friend is dead wrong, and would know so if he read the documentation for the seed() function:

Initialize the basic random number generator. Optional argument x can be any hashable object. If x is omitted or None, current system time is used; current system time is also used to initialize the generator when the module is first imported. If randomness sources are provided by the operating system, they are used instead of the system time (see the os.urandom() function for details on availability).

(Emphasis mine.)

He's guessing based on his knowledge of how it works in other languages. The seed() function is mainly provided so that you can get a reproducible stream of pseudorandom numbers (which is necessary for some specific applications).

The functions you call directly from the random module are actually aliases to methods of a hidden instance of the random.Random class. Each instance, at least in effect, calls seed() within its __init__().

The choice() function obviously does not call seed() before operation, because that would mean re-seeding before every choice, which defeats the purpose of seeding.

Karl Knechtel
  • 62,466
  • 11
  • 102
  • 153