Fixes #24
Fixes #24
Conversation
arrays are more difficult to build and use but they are faster.
but why? should revisit to see if it is simplifiable
Since we suppress lints, document how we could recover the lost accuracy when necessary
this is used at `sufficient_match_count >= 3` so does not have to be u64
- no need to iterate over array of 2 items. - inserting in front of a vec is each an O(n) operation for vec. VecDeque it is O(1).
|
more drops in accuracy. I should reassess |
|
ah after pulling in 9217701 the accuracy is the same again. |
|
Hm... strange but in original lib these changes didn't impact accuracy. |
|
considering they are 0.3% difference, we could maybe lay out every single example and compare which is processed which
if you want that prioritized, we can do it, but since these fixes result in same as current main, they can be merged regardless. Of course if you'd rather stop any other fixes before the root cause for the 1,2,3 discrepency has been found, we can do that as well. |
|
I've checked. All cases have zeros (compare 0 vs 0). I believe we should return this code but eliminate this case (if multibyte usages are 0 and 0 we should compare mess). Please try to return it with this condition. |
|
Do you want me to do that on top of here or on a separate PR focused on only this issue? |
|
Hm, you're right it will be better to have it in separate PR. |
|
I cannot see the questions for this PR. Did you leave them as code comments? |
Being more strict with
cargo clippyandcargo testlooking more into
cargo clippy -- -Wclippy::pedanticAs said before, pedantic is just that. Pedantic. There are many false positives and we don't need to fix all of them.
But it still would be valuable to understand why they are false positives and document them if necessary.