Filter should retain value

If a filter transform loses its uplink, all its settings (column, value) disappear. That’s fair enough for the column but it can be a right pain if you haven’t made a not of the value.

By contrast, if the replace transform loses its uplink, all the information (apart for column) remains intact so just connecting it to something and picking a column gets it going again. With the filter transform, it might as well be deleted and a new one inserted.

When someone deletes a column we have the choice in downstream Filter transforms to either:

  1. delete all the filter terms associated with it, or
  2. change the column to ‘’.

Currently we do 1. This seems the most logical thing to do. But it does lose associated filter terms.

Doing 2 would keep the associated filter terms, but significantly changes the meaning of the Filter and seems very unintuitive.

The relationship between replace terms and columns is a bit different to filter terms and columns.

Note that you can copy and paste filter terms:

I will give it some thought.

OK, I can understand the rationale. However, this pops up whenever the columns are changed. When I was developing the transform it was often the case that I needed to re-order or subtly change the columns right at the start, and then this issue would propagate through every filter in the chain. Trying to remember what a filter was that I set a day or so previously could be a struggle, and the clipboard wouldn’t help.

Also, as pointed out, there is inconsistency in that the replace transform doesn’t lose everything. That seems more sensible to me, and actually what I expected. Having said that, I hope consistency isn’t achieved by making replace act the same as filter, now!

If you reorder or name a column, you shouldn’t lose any corresponding filter terms (if you do, let us know). Try to avoid deleting connections, if you can, as this nukes all column related options downstream.

There is a difference between Filter and Replace in each filter terms can be tied to a different column, whereas all replace terms apply to all selected columns. That is why they are treated differently.

OK, thanks. If it crops up again I’ll let you know the precise circumstances.