It seems that all the row filters nodes don’t return the results the same way. When you use the new version of the row filter with condition <>, it removes also the missing values on the target column. It is not the case for the others row filter versions working on Knime table input ports. It reacts like the database row filter node.
formation ETL.knwf (32.5 KB)
1 Like
Hello @Fabien_Couprie,
thank you for your feedback and sorry for the long delay in addressing it.
Your post actually sparked a lot of discussion internally. We came to the conclusion that offering two versions of the “<>” operator should improve the situation and gives each user easy control over what happens with missing values, without resorting to composite predicates. The new variant will propagate missing values, the old one will stay as is (modulo labeling improvements). The internal ID is AP-23433. We’ll make sure it is easy to understand which of the two propagates missing values and which does not.
You were also correct in that currently the Row Filter follows the semantics of SQL: if a := missing, then a <> 5 evaluates to missing, and missing never matches same as false (following three-valued logic). The new variant will evaluate a <> 5 to true in this case, hence it will be output.
Best,
Manuel
4 Likes
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.