I'm more or less teaching myself big O notation, so please forgive me if this is a duplicate of a question which applied to my question without me having the wisdom to realise it. For my own amusement/personal development, I'm trying to express the complexity of a function which works on 2 different datasets which relate to each other: I need to iterate over a list, and for each item in a list, iterate over another list checking if the items are related.


for things in list_a {
  for things in list_b {
    if (list_a.thing relates to list_b.thing) go ping

What I have on paper currently is O(n) * O(m), but I'm wondering if it should be expressed as O(n*m) instead, where n = size of set b and m = size of set a? Or is this something else entirely? Again, apologies if this is a stupid/duplicate question, but I couldn't find anyone specifically discussing a nested loop over two different datasets. This answer would suggest that it's O(n^2), but that feels wrong to me, since the sizes of the two datasets themselves are different and independent.

1 Answer 1


It'd be O(n*m) where the worst case is n = m or n*n thus O(n^2). We are interested in worst case run time for Big O. If the data sets sizes are different then it will still be in O(n^2) since we can't guarantee that, for example, the second dataset will be a logarithmic relation to the first. If we could, they wouldn't necessarily be independent. They might be but again, we are looking at worst case and O(nlogn) is within O(n^2).

An example would be an algorithm involving edges and nodes in a graph. Worst case, you might have every node connected to every other node but you'll never be able to be certain that there is a specific relation between the two beforehand without some kind of preprocessing.

  • 1
    Why is the reduction from O(n*m) to O(n^2) valid? You lose the important information that the complexity depends on two independent inputs. Feb 8, 2016 at 7:35
  • For example, you might now that in some situation m = O (log n), so O (n*m) becomes O (n log n).
    – gnasher729
    Feb 8, 2016 at 17:39
  • @BartvanIngenSchenau I believe since Big O defines the upper bound of an algorithm (which is looking for the worst case) and does not care about explaining information about how the inputs are related to each other, the notation requires n^2 since n could equal m.
    – jetset
    May 8, 2020 at 17:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.