specific extended characters are ignored in tag names when /tags/search

There’s something kind of odd going on with this.

1. When I do a search for “Darkleer ♦ The Disciple”, I get exactly one matching tag.

2. Similarly, when I do a search for “Dave ♦ Sawbuck”, I get exactly one matching tag.

3. When I search for “Darkleer ♦”, I get no results whatsoever.

4. When I search for just “♦”, I get 100 tags, all of which have a ♦ in them. But none of the matching tags is either “Darkleer ♦ The Disciple” or “Dave ♦ Sawbuck”. In fact, none of the tags at all have the word “Darkleer” in them.

5. When I search for “Dave ♦”, I get an assortment of 12 tags, all of which have both the word “Dave” and the ♦ symbol in them. But none of those 12 tags are the “Dave ♦ Sawbuck” tag.

I’m wondering if one of the Elasticsearch upgrades changed the way the tokenizer works, so that any tags indexed after the upgrade work fine, and any tags indexed before the upgrade don’t handle the ♦ character properly.

EDIT: On staging, unlike in production, searching for “Dave ♦” does include “Dave ♦ Sawbuck” in the search results. So if this isn’t an indexing issue, there’s something different about the ES configuration for staging vs. production.

EDIT 2: But the “Darkleer ♦ The Disciple” tag has the exact same behavior on staging as on production. So it’s not a staging vs. production difference.

EDIT 3: Yup, this is an indexing issue. Sarken reindexed “Dave ♦ Sawbuck” in production, and it’s now showing up in the “♦” and “Dave ♦” searches (in addition to the “Dave ♦ Sawbuck” search, where it always appeared).

Looks like this is back to the original brokenness thanks to https://github.com/otwcode/otwarchive/pull/3224 \o/

This appears to have gotten more broken with the Elasticsearch upgrade: now searching for "Dave ♦ Sawbuck" returns zero results rather than too many.

• *Labels added*: Roadmap-TagWrangling, Keep-SJ

• *Labels added*: Milestone-Internal0.9