search

numberscope / frontscope

Numberscope's front end and user interface: responsible for specifying sequences and defining and displaying visualizers
MIT License
7 stars 15 forks source link

Network errors on OEIS sequences lead to bad sequences being cached #407

Open katestange opened 1 month ago

katestange commented 1 month ago

As far as I can tell, sequence A002487 is returning this.seq.first as +Infinity which is probably not good.

https://oeis.org/A002487

gwhitney commented 1 month ago

As with #409, this sequence is seeming fine here. Changing the title of the bug to reflect that it is likely a caching error.

katestange commented 1 month ago

Ok, I opened a different browser and cleared all browser history for all time, and then this sequence worked. But A003462 doesn't. Even with totally cleared browser history. Am I going nuts?

katestange commented 1 month ago

Also A000225 and A007583 fail.

But these new-to-me ones work: A115067, A000326, A003463

This is after repeatedly clearing all browser history and restarting frontscope.

The error is as reported: seq.first is Infinity and seq.last is NaN

gwhitney commented 1 month ago

Did you pull my latest PR? It supposedly has a fix, which is to say I made a change I think will help.Let me know if it does or does not.

katestange commented 1 month ago

Tried -- same result.

gwhitney commented 1 month ago

OK, on new browser with cleared cache and my latest PR, when you get a sequence mess up like this, open a new tab and try the same sequence and let me know if the same sequence fails twice in a row in different tabs. Thanks!

katestange commented 1 month ago

Yes, still happening in multiple tabs (with cleared cache, reloaded browser and latest PR).

katestange commented 1 month ago

FWIW, it's not working at the old scope at numberscope.colorado.edu either.

katestange commented 1 month ago

Is it working for you?!

katestange commented 1 month ago

For example, go to numberscope.colorado.edu, add OEIS sequence A000225 and then try Differences -- it reports too few terms. Which is not the case on the OEIS. Or try ModFill, it pops up an error.

gwhitney commented 1 month ago

Ah, I see that https://numberscope.colorado.edu/api/get_oeis_name_and_values/A000225 is not working. So this one is a backend problem and needs to be diagnosed as such.

gwhitney commented 1 month ago

Same thing with A003462 abd A007583.

Vectornaut commented 1 month ago

Of the sequences flagged above, the ones that have "crawling too fast" errors logged are:

These are precisely the ones that still have problems as of today. Calling get_oeis_names_and_values on them returns a "value fetching in progress" error.

There are many "crawling too fast" errors in the logs, but these are the only sequences that got a "crawling too fast" error while fetching the B-file. We think this means that the error protection for metadata requests needs to be extended to value requests.

gwhitney commented 2 weeks ago

This needs https://github.com/numberscope/backscope/issues/154 to be resolved. Hence assigning to milestone beta, so as not to block alpha on backscope. But we could move it up if the team prefers and if @Vectornaut will have availability to work on the backscope part.