Improve query parsing performance #351
Conversation
|
Thanks for the working on this improvement! In general, the change looks good to me. 👍 From the very beginning, I strived to keep the project build as simple as possible. This has a number of advantages and makes it easier for newcomers to learn and contribute to the project. So I have a concern about the introduction of the multi-module build. I would suggest to keep the PR simpler and focus only on the actual code changes. At the moment I would prefer to have a benchmark as a separate project. If the benchmark would become bigger and there would be a strong argument in favor of its inclusion in than the main build, I think we can reconsider it and discuss it as a separate issue. What do you think about it? Would it be ok for you to just keep changes in the |
Use the fact that `lineIdx` is a sorted array to more efficiently calculate the previous cursor and the number of cursors between it and 0. This gives an approximate 3.5x speedup when parsing the kitchen sink query.
|
Sure, if you don't want to a have a multi-module build. I added the benchmark module so the performance of the change could be verified. It is a separate commit so I can easily remove it. |
|
Thanks again! |
Use the fact that
lineIdxis a sorted array to more efficiently calculate the previous cursor and the number of cursors between it and 0. This gives an approximate 3.5x speedup when parsing the kitchen sink query.