Openrefine: Fill down and blank down per record

[Downchuck]

As mentioned in the article "Fill down the right and secure way" it would be helpful to have a blank down and fill down with record awareness. Currently, a blank down would cross record boundaries. In some cases, this isn't the desired behavior and a various temporary transforms may need to be applied to get the desired behavior (of just running blank down on a per record basis).

[tfmorris]

Record mode is only partially supported in Refine and is generally not used at all by transformations. Having per record variants of these functions would be useful, but I don't think we want to get rid of the current behavior.

[shawnholt]

FYI - would be great to alert user if record mode is on though!

[ettorerizza]

Issue still relevant for many users : https://groups.google.com/forum/#!topic/openrefine/TjxiExVpJ6w

[ettorerizza]

There should be a way to create a _safe_ "Fill/Blank Down", since we know the number of records and the length of each in the previous column (I do not know if I'm clear?)

Another important improvement would be a "Fill/Blank" ALL, which would allow to get in a click a clean denormalized dataset.

[wetneb]

I think we should make sure Fill down works as expected in Records mode. I would be in favor of fixing that before 2.9.

[ettorerizza]

Just a little visual reminder of the problem, for the sake of documentation. In this case, the 1 in the first record spreads to the second record.

[thadguidry]

@wetneb Can you work on a fix for this over the next 2 weeks ? (invoice-able work)

[tfmorris]

There is extensive discussion of what actually got implemented in #1573 and #1803, although I was unable to find a concise description.