math/rand: document which bits of seed are used #15788
Comments
|
Even though functions in rand take int64 seed parameter the actual implemented algorithm uses only 31 bit of the seed ( In your example: Probably this should be documented. |
|
Aha, well this information allows me to rewrite my program to fix it. Thanks :) I'll leave this open and let others decide what to do with it, maybe use it to file a documentation bug. |
ianlancetaylor
changed the title
|
Out of curiosity, why does it take an int64 opposed to an int32 when it only uses 31bits of information? |
|
@BlackBaronsTux The package is written in terms of generating random numbers from a |
|
Why is this a doc bug and "let's just change the impl to match what a reasonable person would read in the current docs?" I'm happy to pick this up as a doc bug, but it seems like it might also be reasonable to collapse the given int64 into an int32 by xor'ing the top 32bits with the bottom 32bits, so that a caller who expects all 64bits to matter will get what they expect. (It also won't change the current behavior of the default seed 1, because 0 ^ 1 == 1.) |
|
@jeffallen Changing the random number sequences for a given seed is painful for anybody with existing tests that expect a particular sequence. It's a change we would make to fix a clear bug, such as an unexpected lack of randomness, but it's not a change we will make for a problem like this. |
|
CL https://golang.org/cl/23352 mentions this issue. |
|
CL https://golang.org/cl/23461 mentions this issue. |
Fixes #15788 Change-Id: I5a1fd1e5992f1c16cf8d8437d742bf02e1653b9c Reviewed-on: https://go-review.googlesource.com/23461 Reviewed-by: Ian Lance Taylor <iant@golang.org> Run-TryBot: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
go version)?1.6.2 (1.6.0 has the same issue)
go env)?GOHOSTARCH="amd64"
GOHOSTOS="linux"
Ubuntu 14.04.02 Trusty
If possible, provide a recipe for reproducing the error.
A complete runnable program is good.
A link on play.golang.org is best.
https://play.golang.org/p/JSKCIMotEE
A different output string, and a different sequence of integers in the array.
The exact same output string, and the exact same sequence of integers in the array.
Basically using a different seed (especially one that's numerically quite different) should probably generate a different random sequence. It's very possible that that's just a limitation of the algorithm and that's life but just in case this is actually a bug I wanted to file it so at least someone with more knowledge than me could confirm it either way.
The text was updated successfully, but these errors were encountered: