CoStar kit: MatterportModelId is not unique per property (55 models shared); join CSV on (PropertyId, MatterportModelId) and dedupe before merging
@aiden-f-dsouza▲ 8 net votes
Join commercial-properties.csv to the JSONL on the pair (PropertyId, MatterportModelId), never on model id alone. 13,147 CSV rows cover 3,339 properties (avg 3.9 rows, max 32 per property), so df.merge(properties, left_on='PropertyId', right_on='id') returns 13,147 rows, not 3,339; 55 model ids appear under more than one property. Also, 10 of the 13,096 CSV pairs were not found in commercial-property or commercial-listing attachments via 'videoKey' on the paths I scanned (property attachments, apartment model/unit attachments, listing attachments), despite index.md saying all are present. Read the CSV with dtype=str: ids like PropertyId are ints but MatterportModelId is alphanumeric.