[freebase-gridworks] Making new rows with values from split column

[Jeanne Kramer-Smyth]

When I take a column and split the value into new rows, is there a way to have those new rows inherit all other values of the row? A way for them to not be dependent rows on the original parent?

Thanks!
Jeanne

[Raymond Yee]

Jeanne,

I think this would be a useful feature.

-Raymond

[David Huynh]

We don't have that yet. It didn't even occur to me actually but it makes
sense now that you suggested it.

In the meantime, you could go over each of the other columns and do this
transform:
row.record.cells[columnName].value[0]

David

[Jeanne Kramer-Smyth]

Along the same vein, I would like to be able to collapse rows together when I use clustering to update values. My specific case is one in which I want to be able to add values together when I discover they belong in the same row.

For example, the original data might look like this:

ABCDEFG                     35                   100
ABcdeFG                       20                     50

After clustering and finding the match, I would love to end up with 1 row that looks like this:

ABCDEFG                      55                   150

Plausible? Anyone else think this would be useful? I am not sure what happens to non-numeric values - perhaps just putting all the values in a single cell with a given value separator.

Jeanne

[Johan Sundström]

On Wed, Apr 28, 2010 at 23:09, Jeanne Kramer-Smyth
<jkrame...@gmail.com> wrote:
> Along the same vein, I would like to be able to collapse rows together when
> I use clustering to update values. My specific case is one in which I want
> to be able to add values together when I discover they belong in the same
> row.
>
> For example, the original data might look like this:
>
> ABCDEFG                     35                   100
> ABcdeFG                       20                     50
>
> After clustering and finding the match, I would love to end up with 1 row
> that looks like this:
>
> ABCDEFG                      55                   150
>
> Plausible? Anyone else think this would be useful? I am not sure what
> happens to non-numeric values - perhaps just putting all the values in a
> single cell with a given value separator.

Definitely -- this kind of thing is something I'd probably drop out to
a programming language to do most of the time. What operation to
perform on the numbers in the collapsed targets would vary by the
semantics of each column, though; for columns that are "counts of
ABCDEFG", summing is of course the interesting operation, whereas for
columns that are "average price of ABCDEFGs", it would be an
average(count(each type of reconciled ABCDEFG) * value)), however that
would or would not be expressible in something like GEL. Yet other
column semantics might apply too, of course.

I would love this kind of collapsing functionality, especially if
columns could be marked up to carry along some of their semantics,
guiding per-column default choices of operation like this, letting
Gridworks get smarter (without overruling power uses with some other
agenda).

--
/ Johan Sundström, http://ecmanaut.blogspot.com/

[Stefano Mazzocchi]

Yeah, very interesting idea.

I would note that there are two different tasks here:

1) find the sets of rows to conflate

2) how to conflate them

Right now I think Gridworks is pretty good at #1 but totally lacking on
#2... the designing the UI will be pretty interesting.

--
Stefano Mazzocchi Application Catalyst
Metaweb Technologies, Inc. ste...@metaweb.com
-------------------------------------------------------------------

[David Huynh]

I just tried a simple implementation of a "denormalize" command (which
is what I think Jeanne wanted), but my implementation didn't work as
expected. This command involves the record model--an under-designed
aspect of Gridworks currently. I'll need to think much more about this
issue, and I'm afraid it'll have to be a post-1.0 feature.

David