A transformation casts malformed numeric text and needs to identify failures without aborting the query. Which approach is suitable?

1. Assume CAST always returns valid numbers
2. Remove every text column from the source
3. Replace the whole dataset with one default row
4. Use try_cast and evaluate resulting nulls