Possible Row Filter Node Behavior with Missing Values in "Does Not Equal" Condition

[lanfengye]

row filter bug.knwf (72.6 KB)
I encountered an unexpected behavior when using the Row Filter node with “Does not equal” conditions. It appears that Missing values are being systematically excluded from the output, even when explicitly using “Does not equal” filtering. This occurs in both interactive configuration and expression-based modes (tested on KNIME 5.4.3/Local Win11).

[Attached screenshot/Workflow] demonstrates the issue:

1. Input table contains mixed valid values and missing cells
2. Applying col != "a" removes all rows with “a” AND nulls

Is this intentional design or a potential edge case handling? Appreciate any clarification.

[FlorentM]

Hello,

This should be the intentional behaviour. To filter both string values and missing values, two ways are possible:

• Add another criterion to keep missing values and set the Match row if matched by option to “Any criterion”
• The old way, select the operator “Equals” and set the Filter behaviour to “Output non-matching rows”

[lanfengye]

Thank you very much. However, I’m inclined to think this is a bug, and it would be best to receive confirmation from the KNIME official team.

[hotzm]

Hi @lanfengye,

the operator operates with semantics similar to SQL (see previous topic here), where Row2 “column1” could be an “a” in reality – it might just be missing in the dataset so the operator filters it out. The DB Filter node would do so, too, but the old Row Filter did not.

For the next feature release AP 5.5 (due in Summer) we added a variant of “does not equal” that retains missing values to make it easier to use in complex filters:

The existing operator will be relabeled to “Is not equal (nor missing)” to make it clear what it does and the new operator will be called “Is not equal”, which should make it more intuitive for users of the old Row Filter.

[lanfengye]

Thank you very much. I got it ~