SRANDMEMBER key [count]
When called with just the
key argument, return a random element from the set value stored at
Starting from Redis version 2.6, when called with the additional
count argument, return an array of
count distinct elements if
count is positive. If called with a negative
count the behavior changes and the command is allowed to return the same element multiple times. In this case the number of returned elements is the absolute value of the specified
When called with just the key argument, the operation is similar to SPOP, however while SPOP also removes the randomly selected element from the set, SRANDMEMBER will just return a random element without altering the original set in any way.
Bulk string reply: without the additional
count argument the command returns a Bulk Reply with the randomly selected element, or
key does not exist. Array reply: when the additional
count argument is passed the command returns an array of elements, or an empty array when
key does not exist.
(integer) 3redis> SRANDMEMBER myset
"one"redis> SRANDMEMBER myset 2
1) "one" 2) "three"redis> SRANDMEMBER myset -5
1) "one" 2) "two" 3) "three" 4) "one" 5) "three"
When a count argument is passed and is positive, the elements are returned as if every selected element is removed from the set (like the extraction of numbers in the game of Bingo). However elements are not removed from the Set. So basically:
When instead the count is negative, the behavior changes and the extraction happens as if you put the extracted element inside the bag again after every extraction, so repeated elements are possible, and the number of elements requested is always returned as we can repeat the same elements again and again, with the exception of an empty Set (non existing key) that will always produce an empty array as a result.
The distribution of the returned elements is far from perfect when the number of elements in the set is small, this is due to the fact that we used an approximated random element function that does not really guarantees good distribution.
The algorithm used, that is implemented inside dict.c, samples the hash table buckets to find a non-empty one. Once a non empty bucket is found, since we use chaining in our hash table implementation, the number of elements inside the bucket is checked and a random element is selected.
This means that if you have two non-empty buckets in the entire hash table, and one has three elements while one has just one, the element that is alone in its bucket will be returned with much higher probability.
© 2009–2018 Salvatore Sanfilippo
Licensed under the Creative Commons Attribution-ShareAlike License 4.0.