Data quality: rules & ideas

[raphodn]

We're starting to see new github issues related to data quality, and users asking for more flexibility on price addition.

So I was thinking we should start a list of possible rules that can be used in the future by moderators (nutripatrol) or automated (robotoff) to help flag errors or low quality prices

[raphodn]

Model	Rule	Description	Related issue(s)
Price	if with a product_code, then the category-related fields should be empty (category_tag, labels_tags, origins_tags, price_per)
Price	if with a category_tag, then the product-related fields should be empty (product_code)
Price	if with a category_tag, then price_per should be set
Price	the price cannot be negative	there is already a check in the backend on save i believe
Price	if price_without_discount is set, then price_is_discounted must be True
Price	if both price_without_discount & price are set, then price_without_discount > price
Price	location_osm_type should be set if location_osm_id is set (and vice-versa)
Price	a price date cannot be in the future
Product	check if the barcode is valid (EAN8, EAN13)		openfoodfacts/open-prices-frontend#533
Proof	a RECEIPT proof should have all it's prices with the same location (as the proof)		#244
Proof	a PRICE_TAG proof should have all it's prices with the same location (as the proof)		#244
Proof	a RECEIPT proof should have all it's prices with the same date (as the proof)		#314
Proof	a PRICE_TAG proof should have all it's prices with the same date (as the proof)		#314
Proof	a RECEIPT proof should have all it's prices with the same currency (as the proof)
Proof	a PRICE_TAG proof should have all it's prices with the same currency (as the proof)
Price	location_osm_type should be set if location_osm_id is set (and vice-versa)
Proof	a proof date cannot be in the future

[teolemon]

@raphodn

• there's the notion of error and warning (eg all prices on a receipt are the same, probably a warning, not obligatory an error)
• Proof: send to Google Cloud Vision, and store the matching JSON output #320