Remove and de-duplicate tags with leading/trailing whitespace

[AetherUnbound]

Description

We have some records in our data where there are duplicate tags, only the duplicate tag has leading or trailing whitespace. Here's an example: https://api.openverse.engineering/v1/images/2d454032-0cc1-48a5-8f40-e9235f1a4f12/

"tags": [
        {
            "name": "abigfave",
            "accuracy": null
        },
        {
            "name": " abigfave",
            "accuracy": null
        },
        {
            "name": "artisanal",
            "accuracy": null
        },
        {
            "name": " artisanal",
            "accuracy": null
        },
        {
            "name": "color",
            "accuracy": null
        },
        {
            "name": " color",
            "accuracy": null
        },
        ...
    ],

This might need to be tackled in two steps, or a least an operation which covers both cases:

1. Remove leading/trailing whitespace from existing tags
2. Deduplicate tags which might previously have been separate ones

We will also want to check, similar to #1566, that any new tags added always have extra whitespace stripped.

Additional context

Related to #430

[sarayourfriend]

@WordPress/openverse-catalog I'd like to take a shot at this issue. Am I correct to assume this should use the batched update DAG? And if so, I think I'd like to try it in two steps, as suggested, basically doing something like this:

1. Batch update to strip leading/trailing whitespace
2. Batch update using something like select distinct for each set of tags?

Is such a thing possible with the batched update DAG? Are there any potentially helpful examples of how we've used that recently I could work off of?

[krysal]

@sarayourfriend You're correct. It's possible to do it with the batched_update DAG. The deletion of duplicates was resolved in #1566 (comment) and Postgres has string functions for trimming.

[AetherUnbound]

If possible, it'd be best to combine both of those steps into a single batched update, that way we don't have to do two passes on the data! Might make for a tricky query, but then we only have to run it once 😄

[sarayourfriend]

I guess select distinct trimming_function(tag.name), or something along those lines would work?

Thanks for the input, y'all.

[krysal]

Reopening for the pending execution of the trim_and_deduplicate_tags DAG.

[krysal]

Solved in #4557.