bug: unexpected handling of http 400 error

[talagluck]

Description

I'm trying the following query:

select * From read_json('https://data.cityofnewyork.us/resource/2npr-yv2b.json?$limit=1000');

And I'm getting the following error:

ExecutionException: External error: Exec error: Remote session error: unable to dispatch table access: status: Unknown, message: "Missing session for id c3031c45-efcd-4491-b863-60f9570a88fe", details: [], metadata: MetadataMap { headers: {"content-type": "application/grpc", "content-length": "0", "date": "Tue, 26 Mar 2024 16:32:02 GMT"} }

When I go directly to the url, I can see the data.

And when I run this query, it works:

select * From read_json('https://opdb.org/api/search/typeahead?q=*');

[tychoish]

Some debugging notes here:

• the $ in the URL seems to be potentially an issue. or at least that's a thing that's different between the thing that works and the thing that doesn't (this is unlikely to be the thing)
• the session error seems like a "running in cloud" problem, so I ran it locally in the CLI off of (approximately) main and got the following response

    > select * From read_json('https://data.cityofnewyork.us/resource/2npr-yv2b.json?$limit=1000');
    Error: External error: Unexpected status code '400 Bad Request' for url: 'https://data.cityofnewyork.us/resource/2npr-yv2b.json?$limit=1000'
  

  I don't think this is a cloud problem, or at least it's not a cloud problem worth looking at yet.
• I did some poking at this more and curl gives me a 400 when I attempt to do a HEAD request, so that's probably worth looking into.
• The error reporting from remote glaredb instances probably could be better.

We definitely should investigate more here if there's something easy to get these to work (like HEAD request avoidance) but some web servers/CDNs might not be compatible for a bit.

[talagluck]

Totally fair - and to be clear, this isn't necessarily a "let's fix now" but an "I see something that doesn't work and it would be nice if it did."