imagine we want to find all rows in our input table matching a certain number in a given column. Let's say the column to test is a Double type column and our value of interest is an Integer type. The Row Filter node now creates an empty table.
While this might be some Java specific issue, it is a) not very intuitive for some of the KNIME users and b) can quickly be overseen in a huge workflow (which happened to me actually). And from my modest Java knowledge it could easily be solved.
What do you think?
P.S.: The Rule-based Row Filter for example behaves as desired and does not need a ".0" to do what he should do. ;-)
You shouldn't use pattern matching for numeric columns (maybe this options shouldn't even be available). Use range checking instead. The pattern matching will convert it into a string and double values will always have a separator character.
Just to clarify this:
"Pattern matching" always converts the input from "Column to test" into a String before comparison?
Obviously it is allowed to use range checking for String comparison. When doing that, I get a warning message, that this may not work as expected. Can anyone provide me with an example where it may work? Just out of interest...
However, I think a more "intelligent" Row Filter node might me somewhat more intuitive and neat.