In Puzzle 2431, I repeatedly got zero results from the Compound Library. Now that is entirely possible (though disappointing, of course).
But then I submitted a compound that already has the CL bonus ("Compound(s) found in library" when I hover that objective). Guess what: the reply from the Library says "0" again.
Shared this compound with scientists: Puzzle 2431, 23100 points.
The same here with three different compounds, all with the CL bonus and all return zero results.
Same problem occurs in Puzzle 2437, when submitting the same compound. It returns zero hits, and it does so immediately, which means that it is still in the server cache.
After changing one atom (from -OH to -NH3+), I can submit this new compound, and get 31 hits, including an Exact Match for the newly submitted one, but also an entry that matches the original (i.e. the one that returned zero hits).
My conclusion: the fact that both compounds get the CL bonus is not a hiccup of my local cache. Both of them are in the Library, but for some reason the search algorithm fails when fed with one of them. Or maybe I should say it failed eleven days ago, and that result is still in the server cache?
Some feedback would be appreciated. As it is now, it looks to me like this bug report has not been read by Foldit staff at all. I hope this is not really the case.
Apologies for the radio silence. We are aware of this feedback, and have been trying to track it down, unfortunately without much success.
Best we can tell, it's due to a temporary issue with the ZINC server we use to search for compounds. We've addressed a few potential failure cases and added extra logging, but it's not something that has replicated itself. We'll continue to monitor things.
The fact that a previously submitted compound returns the same result when resubmitted is expected. To limit how often we need to query the ZINC server, we cache the search results. Resubmitting the same compound will return the same results. There's potentially ways of manually resetting the compounds, and I'll see if there's a decent way of figuring out which compounds are affected and see if we can reset them. (I was holding off on that, so as not to complicate the debugging) – Hopefully that's sufficient to address the issue, and it won't re-appear.
Thank you for your reply. I can understand that an issue like this is not easy to replicate, especially when it was a temporary issue on an external server.
In case the phenomenon re-appears, I will let you know. If there is anything else that I (or possibly other players) could do to help you in your effort to track it down, just let us know!
Today, I tried re-submitting two more compounds that did not yield any hits in Puzzle 2431.
The first one, which I named P2431 1.0, came back immediately, again with zero hits. As you explained, this is to be expected when the search is still in the server cache.
I then changed one -OH to -NH3+, and named it P2431 1.0N. It took a reasonable time to come back with 63 hits, including one Exact Match, but not including P2431 1.0.
Conclusion: that compound is indeed not in the Library. But still, the fact that the original search did not return any of the 63 hits I got for my quite similar P2431 1.0N, suggests that something went wrong two weeks ago.
When I submitted the second one, P2431 1.1, I noticed that my search for P2431 1.0 was Running again. And so was the one I had submitted last Sunday, 0.1 P2431 (see my post above). After quite a few minutes, all three came back with 52, 54, and 29 hits, respectively. As expected, P2431 1.1 and P2431 1.0 did not include an exact match, and 0.1 P2431 did.
Could it be that you have indeed reset these compounds manually? If that is the case, it seems to me that this reset did what it was meant to do.
Thank you for your efforts!
Yes, we just submitted a reset of the compounds which were incorrectly marked as having zero hits. Some of these should now have compound results, but some of them should now be correctly marked as "Error" (as the ZINC server is spitting back errors for trying the search). – Hopefully things should now be set such that if there's an issue with the search when resubmitting a compound library result, that search will be marked as "Error", rather than being incorrectly marked as a zero result.
What should we do if the Foldit Compound Library returns "Error" for a search? In Puzzle 2437 at 535am CDT April 4, when I sent the compound I call JH 17138.393 +3853.8 to the Foldit Library, it returned "Error". JH is supposed to be a compound from the ZINC20 database. Usually such compounds give at least 1 hit in the Foldit Library. I shared this compound with scientists at least.
Unfortunately, the ZINC server search results doesn't seem to give a clear difference between "we're unable to process this compound, and resubmitting it won't help" and "we've had a temporary issue, and if you try submitting the compound again later it may work". This complicates the processing, so if something is listed as "Error" it will stay that way until there's some sort of manual intervention. We'll keep working on better ways of handling things.
I'll also note that for Puzzle 2437 we're using the Enamine REAL set. There's some overlap between the ZINC20 set, but there's some compounds in ZINC20 which aren't in the REAL set (and vice versa). If you've found the compound externally, it may or may not show up in the library we're using within Foldit.