Implement Fisher Yates algorithm#172
Conversation
kayvank
force-pushed
the
170/implement-Fisher-Yates-algorithm
branch
from
402f667 to
7eaf4e7
Compare
|
Thanks for submitting a PR! I'll have a good look at it later but if you're open to it, you could look at a one-time test for randomness here: https://cnut1648.github.io/files/posts/Test_for_rand.pdf Basically, take the indices that are output by shuffleVec and take a look at the distribution. But that's might be overkill. |
| where | ||
| shuffleVec :: (RandomGen g) => g -> VU.Vector Int -> VU.Vector Int | ||
| shuffleVec g v = runST $ do | ||
| vm <- VU.thaw v |
There was a problem hiding this comment.
Instead of declaring the vector from list just to thaw it we can just create a new one in the shuffle vec function.
kayvank
force-pushed
the
170/implement-Fisher-Yates-algorithm
branch
from
fc6f38a to
cc83918
Compare
Issue 170, implement PR comments and clean up build warnings
kayvank
force-pushed
the
170/implement-Fisher-Yates-algorithm
branch
from
cc83918 to
3e69bd5
Compare
|
@kayvank this looks good. Rustin on some simple tests. Mostly that indices aren't dropped or duplicated and that it doesn't fail on empty. |
| shuffledIndices :: (RandomGen g) => g -> Int -> VU.Vector Int | ||
| shuffledIndices pureGen k = VU.fromList (shuffle' [0 .. (k - 1)] k pureGen) | ||
| shuffledIndices pureGen k | ||
| | k <= 0 = VU.empty |
There was a problem hiding this comment.
We return empty vector even when k is a negative number, which does not seen correct.
Should we error inf the rare event that k < 0? @mchav
There was a problem hiding this comment.
I think that's fine since the number is derived from the size of the dataframe. And the shuffle of an empty dataframe is an empty dataframe.
kayvank
force-pushed
the
170/implement-Fisher-Yates-algorithm
branch
from
ff07a4e to
0b00734
Compare
Added two new unit tests. |
3 participants
No description provided.