Restore previous sorting method about nameresolution HOT 3 CLOSED

Had this discussion in slack, but putting here as well:

Looking. back at v 1.2, I think that what happens is that the first query does some sorting, but it doesn't really matter because the second query is the one that goes at gets all the names for the curies found in query 1:
7:41
{
"query": f"({curie_filter}) AND ({name_filters})",
"limit": 1000000,
"sort": "length ASC",
"fields": "curie,name",
}
7:42
So now you have this sorted by length of the name, but with the curie on mutliple rows and all the curies intermixed together.
So then the results get parsed out like this:
output = defaultdict(list)
for doc in response.json()["response"]["docs"]:
output[doc["curie"]].append(doc["name"])
(edited)
7:44
And this output dict is magically preserving the order. curie with the shortest doc gets added first. Now the sorting of the first query does actually matter, because you want to get the curies that have the shortest name first.
7:45
So for us, either we need to add another parameter on the doc which is the length of the shortest element in the synonym list, or we need a way to figure it out in the query

Chris Bizon
8:02 PM
I think we could do it on the query with https://solr.apache.org/guide/7_6/function-queries.html#sort-by-function, we just need to figure out the right function. I think it's the length of the first synonym

from nameresolution.

So for us, either we need to add another parameter on the doc which is the length of the shortest element in the synonym list, or we need a way to figure it out in the query

I’ve been looking in https://solr.apache.org/guide/7_6/function-queries.html and I don’t see a length function. I tried Googling for it and couldn’t find anything either (apart from ChatGPT which helpfully suggested the field_length() function that nobody else has heard of). I think you’re right that probably the best short-term solution is to reinstate the length field in Solr that we used to have previously. I’m going to call this shortest_name_length for now so it’s clear what it means, but I wonder if we should try to factor the information content value in here as well. I should be able to implement the shortest_name_length field in Babel pretty quickly, so I'll take back ownership of this ticket.

from nameresolution.

This has now been significantly improved with CURIE suffix sorting, and it's working well enough that is what is being used by Translator UI. Closing.

from nameresolution.