bruns added inline comments. INLINE COMMENTS
> dfaure wrote in kcoredirlister.cpp:852 > That was the case already.... hmm but not in my version of kio. It looks like > this patch is on top of other patches. D10742 <https://phabricator.kde.org/D10742> - this makes it even harder to review ... > mwolff wrote in kcoredirlister.cpp:1970 > O(N) iteration over a large hash is extremely slow, is this done elsewhere? > if so, then you may need to find an alternative approach - potentially via > multiple containers or by using a sorted vector after all like you proposed > initially My proposal: A few lines earlier, a `KFileItemList` (~`QList<KFileItem`) is created. Sort this list by name. Use a ordered container (e.g. `QMap`) instead of `QHash`. You can iterate the `QMap` and the `KFileItemList` in parallel, which makes erasing an O(m + n) iteration, instead of O(m * n). (Of course, on top you have sorting of the KFileItemList - O(n log n) - and the overhead when inserting into the QMap<QUrl, KFileItem>) REPOSITORY R241 KIO REVISION DETAIL https://phabricator.kde.org/D11282 To: jtamate, #frameworks, dfaure Cc: kde-frameworks-devel, bruns, mwolff, michaelh, ngraham