Multiple elastic search terms: COSMOS search API about cosmos

[iross]

https://github.com/UW-COSMOS/Cosmos/blob/master/cosmos/birdnest_backend/birdnest_backend/app.py#L268

Looks like it does a "match" as opposed to a "match phrase" (which is what we use in the xDD API). So multiple words are treated as an OR.

[cambro]

That's what I thought. We should probably try to incorporate a switch for how multiple terms are handled in ES, which requires both front end and back end work. The first step is to distinguish multiple terms from a search phrase. Comma-separation would probably be best. Want to coordinate with @davenquinn about this @iross?

[davenquinn]

If we default to "and" (match phrase I guess) and split the search string on commas to generate "or" relationships, could this be done without involving the frontend?

[iross]

"match phrase" isn't just an "and", it's a "search for this string as a complete phrase" So really there's three modes possible: "novel coronavirus" (the phrase) "novel" and "coronavirus" (both in the document, but not necessarily adjacent) "novel" or "coronavirus" I think a default of phrase matching, with phrases separated by comma makes sense. But I'm less sure about whether a comma-separated list should be ORed or ANDed by default.

…

On Tue, Jun 16, 2020 at 12:01 PM Daven Quinn ***@***.***> wrote: If we default to "and" (match phrase I guess) and split the search string on commas to generate "or" relationships, could this be done without involving the frontend? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#108 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AALAW7N3IV3SPFJEPM7CZGLRW6QQLANCNFSM4N7FE7ZA> .

[cambro]

Exactly as stated.

1. multiple terms no commas is a phrase
2. multiple terms, comma separated: default to inclusive "and", optional "or" would be good

Edit: @iross this search is not over the document, it is over individual cosmos table, figure, equation elements.

[davenquinn]

OK, so we would need an and/or switch in the API and frontend.

[iross]

Logic has been updated in the live instance. on the search_es_objects endpoint:

• Search now uses match_phrase by default, so multiple terms with no commas is searched as a full phrase
• Comma-separated lists are treated as separate terms, defaulting to and
• &search_logic=any will switch comma-separated term lists to use OR logic.

Keeping this open until I propagate the changes back into master.

[cambro]

See this issue for possible lingering problem:
UW-COSMOS/cosmos-visualizer#48

[davenquinn]

Added search_logic to frontend in COSMOS/cosmos-visualizer#4e81d1d.

[ankur-gos]

@iross are the changes in master for AND/OR logic?

[iross]

Merged in ages ago. Closing.