Request for review: "gh-119702: New dynamic algorithm selection for string search (+ rfind alignment)"

I have done some work on removing hard-boundaries in algorithm selection for string search.

As I went through the code carefully and ended up understanding most of it, couple of side effects of this happened to be:

  1. Non-trivial performance improvements
  2. Direction agnostic algorithm implementations (rfind uses exactly same routines and does not suffer quadratic complexity anymore)

Besides benefits listed above, this also results in better code structure which supports sensible initial implementation of ā€œfinite set of patternsā€ string search.

This would ideally be merged before I moved on to other string extensions. See: String Search Overview, Coverage and API

It has been reviewed and iterated many times. I gave it a month or so to make sure it settled and it is ready for a coredev review.

I would appreciate if someone looked at it.

4 Likes

If anyone is interested in understanding string search algorithms better, this is still needed.

There havenā€™t been any objections to this yet and feedback is mostly positive.

However, I appreciate that it might take a while until the right person appears to review this. So I will bump this from time to time if it is ok. I posted in String search improvements Ā· Issue #691 Ā· faster-cpython/ideas Ā· GitHub, but been advised to ping a bit more here in discourse.

Itā€™s not. This doesnā€™t scale, and is noisy. Maintainers are already aware of their backlog.

3 Likes

But in this case, it would be good to receive such confirmation. I.e. that it is in someones backlog.

Otherwise, if no response was given, I am left under impression that no one has acknowledged it / picked it up.

In which case bumping it again is a pretty natural following action IMO.

Or is there something that I am not aware of? E.g. once I post here I have assurance by default that it is in someones backlog? Or something similarā€¦

1 Like

If itā€™s in the tracker, itā€™s in peopleā€™s backlogs. That may or may not mean it gets someoneā€™s attention soon, but thatā€™s for the maintainers to manage. Nagging people in an attempt to get action sooner is not helpful.

So are there any actions that I am allowed to do to prevent this from being forgotten after posting it once?

Is this your interpretation of my actions so far regarding this topic?

1 Like

Wait patiently. Iā€™m sorry if that isnā€™t what you want to hear, but everyone involved here is a volunteer and youā€™re not entitled to any particular timeframe for responses. As I said, it wonā€™t be forgotten, itā€™s on the tracker list. People will review that list when they have the time and inclination.

Not so far, no. Iā€™m hoping my comments help you avoid falling into the trap of appearing to behave like that, is all. Not being willing to accept the advice given by one of the moderators is concerning, though.

4 Likes

Letā€™s take a step back.

Is bumping PR-review-request forbidden in this forum?

I would say bumping it once in 3-6 months could be reasonable. Once a year?

I am completely fine if the rules are that it is forbidden, as long as it is clear to me and everyone agrees.

1 Like

The overriding rule is ā€œbe reasonableā€. Iā€™m not aware of specific rules that give an exact number of months/weeks or anything like that. Continuing to bump when requested not to by a moderator, does not match my idea of ā€œbeing reasonableā€. Continuing to argue the point when youā€™ve been given advice on what to do doesnā€™t seem reasonable to me.

Iā€™m not going to comment further. I feel like if I continue to give you comments to argue against, youā€™re going to end up in code of conduct territory, and Iā€™d prefer not to see that happen.

2 Likes

I tend to be as reasonable as I can, all of your concerns are much appreciated.

2 Likes