Req: List Reason on Machine pages

I have a Feature Request. Is it possible to list the Reason that a TM was classified the way it was on individual machine pages? This could be as simple as the Decider which “proved” it, but it would be even better to include details (like the number of backsteps needed for backtracking or the start time and period of the cycle for Cyclers, etc.). This seems necessary for reproducibility, understanding and collaboration.

I can imagine other metadata about individual TMs could also be useful (what Zoology they are part of, links to forum posts for that specific machine, etc.).

In my case, I am specifically looking at (Skelet #10) which y’all list as proven infinite! I have not been able to prove it infinite with my system and I’d love to understand how you did it! (Or check if there was, perhaps, a mistake).

YES!! It is a very much needed feature.

I have opened this feature request on the backend of the API a while ago: If a machine is decided returns the name/link of its decider · Issue #2 · bbchallenge/bbchallenge-api · GitHub

There is a technical inconvenience.

Decided machines are given in big index files:
These files contain the big endian 4-byte representations of IDs in the DB of the machines that were decided by the decider, in increasing order of ID.

As it is not affordable to load those files in memory in my VPS setup, in order to implement your feature one has to implement a dichotomy of file seek that looks if the requested ID is in one of these files and returns the file name. Although it is standard algorithmic I have not have had the time to implement and test it.

If you had the time to wrap your head around it I would be happy to talk about this in more details.

The particular machine that you point at was decided by [Decider] Backward reasoning. Basically, any machine that is not a Cycler/Translated Cycler so far has been decided by Backward Reasoning (depth 300) :smiley: .

Just a follow up as I wonder if you believe that the machine was mistakenly decided or if you agree that it can be decided by backward reasoning.

My Backtracking does not classify it. I’ve opened a new Thread: Skelet #10 Backtracking?

1 Like

I implemented the feature:
Screenshot from 2022-06-11 21-52-34

This required:

  1. Making a custom library for binary search in binary files:
  2. Adapting the backend for it: fa45d51
  3. Adapting the frontend for it: c8d094d

EDIT: There seems to be a bug, I have temporarily disabled the feature and will debug it tomorrow.

The bug is fixed! I learnt to that occasion the funny thing that had been decided by both [Decider] Backward reasoning and [Decider] Translated cyclers (machine index present in two files, this can happen in some rare instances where we rolled-back the undecided index). The feature shows just one decider (whichever the backend finds first while looking through index files) as being decided by multiple deciders will be a rare event.

However, contrarily to what I said here before, has not been decided by Backward Reasoning and only by Translated Cyclers.

This is great, thanks!

Is it intentional that Decider is shown for Undecided machines? Ex: I wouldn’t have expected that, what does it mean? That this is the last Decider that attempted to categorize this machine?

Screen Shot 2022-06-12 at 11.48.12 AM

I’m seeing Undecided paired with different Deciders … which seems odd:

Screen Shot 2022-06-12 at 11.49.36 AM

Hm, if I reload those links they don’t show Decider … but if I click “Random” All the examples do show Undecided and a Decider. Seems like some sort of Caching / First Load issue …

There was indeed a caching issue, thank you! I believe that I corrected the bug and added more safeguards to protect against this: dd37f5d, can you confirm you can’t reproduce it on your end?

You should never see a decider show up for an undecided machine, or else there is a another bug :smiley:

Thank you very much!

Looks fixed, thanks!

1 Like