don't boost street if parse is ambiguous

[blackmad]

This is a redo of #1430 that's on the pelias repo rather than on mine, and fixed for place->venue rename

It fixes mccarren park and wrigley field as well as mccarren park brooklyn and wrigley field chicago

this was the original approach suggested by @missinglink for solving "wrigley field" - I don't see a great way to solve it in pelias/parser at the moment, though I'm open to suggestions.

doesn't change anything in a 1k query set.

[orangejulius]

Cool, this makes a lot of sense. I'll test it out next week!

[blackmad]

something else this should probably do is that if the parse is ambiguous, consume the whole query

it would fix queries like
Kinetic Arts Center
"parsed_text": {
"subject": "Arts Center",
"street": "Arts Center"
}

and

KT Auto Repair
"parsed_text": {
"subject": "Auto Repair",
"street": "Auto Repair"
}

[orangejulius]

I just gave this PR a full run through all our tests, and it looks good. There are some improvements to POI related tests, such as the following:

It looks similar, but not quite as good as #1469 in that regard (#1469 is a more aggressive change with more downsides, so...tradeoffs)

There were very minimal changes to address queries that are not worth worrying about. Here's an example of the magnitude of the changes

As far as I'm concerned from a behavior perspective this PR is good to merge.

@missinglink do you have any thoughts on the code? It does add even more logic to our already complex sanitizer/_text_pelias_parser.js file, but that's probably unavoidable.

[Joxit]

I think you can put this before const restOfSolutionIsTheSame =.... ? something like

if (!streetOnlySolution || !venueOnlySolution || streetOnlySolution.pair[0].span.body !== venueOnlySolution.pair[0].span.body) {
  return false
}

[orangejulius]

We just merged #1308, so I'd be curious to see if/when this PR provides additional improvements on top of this. @blackmad if you get a chance to rebase and check it out, let me know, otherwise I'll add it to my queue for testing.

[blackmad]

I think this is basically a no-op on top of 1308, I'm happy to just close it

[orangejulius]

@blackmad Its a no-op for autocomplete, but not necessarily search, so it might still be useful.

[blackmad]

okay true you're right. This is still important for "wrigley field" in search