Set the watermarked property to false in all images where it's NULL

[obulat]

Current Situation

Currently, there are 1 105 608 images that have watermarked property set to NULL.

Suggested Improvement

It should be safe to assume that these images are not watermarked, and set the value to false.

Benefit

Data consistency.

Additional context

Part of #244

Implementation

• 🙋 I would be interested in implementing this feature.

[krysal]

I wonder what is the meaning of this column and why it should be considered false for all those images 🤔
Can you elaborate on the reasoning, Olga? Do we have providers that return watermarked images?

Also, I think we would have to modify this for each provider script instead of doing a one-time UPDATE in the database, or the refresh process will convert the column to null again.

[obulat]

I wonder what is the meaning of this column and why it should be considered false for all those images 🤔 Can you elaborate on the reasoning, Olga? Do we have providers that return watermarked images?

Sorry, I haven't noticed your reply until today, @krysal.
The reasoning for this issue was that by default, we should probably consider images to be not watermarked and only set watermarked to true if we know that the provider returns watermarked images.

There were some providers who set the watermarked to true in the CommonCrawl days: IHA, McCord Museum, FloraOn, but now no providers set watermarked value, and so it falls back to false: https://github.com/WordPress/openverse-catalog/blob/9cd0b97b844c7442a00e8c056fc7246e83db06d7/openverse_catalog/dags/common/storage/image.py#L61

Because of this fallback, currently, we cannot be sure what false or null in the watermarked column means: not watermarked or not known. So, I think it makes sense to simplify and only use false here.

Also, I think we would have to modify this for each provider script instead of doing a one-time UPDATE in the database, or the refresh process will convert the column to null again.

I was thinking of running the UPDATE in the catalog (upstream, the source-of-truth) database, which is not touched by the refresh process, so the values will not be re-set to null. The data would be updated during the re-ingestion, but some providers are not re-ingested, so the data would never be updated.