I'm collecting files in this spreadsheet for now: https://docs.google.com/spreadsheets/d/1SIqjRdDh7G4KRBi47kz5WV3Ejl6YISiUrq3GhERnBSU/edit#gid=0

Other testing scenarios we need to cover?

I have collected examples of files pointing to all relevant datatypes in their SDC.

Data types which are not used on Commons, or still very exotic, seem to be the following ones. I found no Commons files pointing to...

• Mathematical expression
• Geo shape
• Musical Notation
• Tabular data

So I'm ignoring these (also not doing sandbox experiments with them). As these are basically strings in the backend too, that shouldn't really pose problems, as briefly discussed with @Pintoch.

I will now proceed to further collect files that cover all supported file types on Commons - these should give us examples to work with for the preview function of the Commons Reconciliation service for instance (cfr T292526: Add preview service to reconciliation API).

I have added a selection of files that should cover all supported file types on Wikimedia Commons. In addition, I chose files that will be interesting for testing the preview function per T292526: Add preview service to reconciliation API and inserted a variety of formats of the file path itself so that we can stress-test T290088: The Structured Data on Commons reconciliation service recognizes the most widely used Commons file name formats.

The dataset is here. You can simply import this in OpenRefine and start/test reconciliation and data extension with this set.

Not done yet: make sure that a similar set of interesting testing files is available on Beta Commons.