Column value abbreviation

[Prag_Mor]

Dear Knimer,

Please need help to get the node details in below format

Column	Expected result in Knime
swizterland-april-monday-america	Sw-Ap-Mo-Am

[takbb]

Hi @Prag_Mor

There are a number of ways this could be achieved. Attached is a method that springs to mind.

KNIME_abbreviate.knwf (10.4 KB)

[Prag_Mor]

Thanks for the solution. @takbb
But the below problem arises:

swizterland-april-monday-america Sw-Ap-Mo-Am
swizterland-april Sw-Ap-- (Expected value is Sw-Ap)

In column Combiner how to prevent hyphens to be added if there is no value.

Column	Expected result in Knime
swizterland-april-monday-america	Sw-Ap-Mo-Am
swizterland-april	Sw-Ap	Sw-Ap–

[takbb]

Hi @Prag_Mor

I think maybe you are going to need to state all of your requirements … how many different variations of terms does this need to allow for?

[takbb]

Maybe one of these solutions. It really depends on what other examples you have.KNIME_abbreviate 2.knwf (21.4 KB)

[Prag_Mor]

Thank you @takbb
It worked

[Daniel_Weikert]

For regex you could also try ^\w{2}|(?<=-)\w{2} and then a groupby because the extractor splits to multiple rows. The result is not capitalized however. Still fun challenge.

[takbb]

Hi @Daniel_Weikert , is that using the Palladian Regex Extractor?

OK, yes. Nice regex and great idea. Building on that… what if we find a way to capitalize the “words” prior to the regex? I just discovered that String Manipulation’s capitalize function allows specification of a delimiter:

capitalize($$CURRENTCOLUMN$$,"-")

Then we apply your Regex, but modified slightly so we also take the leading hyphen along for the ride too.
^\w{2}|-\w{2}

(I got a bit lost using the extractor to split to rows, when I had more than one row as input data, as I then had a group of rows with no obvious (to me) way of assembling them back into their original row-groups. So I tried splitting to columns instead.

If we tell it to split to columns we get this:

Replacing missing values with zero-length (empty) strings, we can then combine them back with the Column Combiner. So that gets it down to 4 nodes from 5, without writing the logic “in code” (as I did in the three-node option2)

Thought it would be nice though to find away of avoiding having to deal with those annoying “missing values”. Doing the Regex extraction to a list (instead of to columns) looked promising because it automatically skipped missing values from the lists it generated

Only trouble was… finding a way to convert the collection back into a string! I thought that ought to be straightforward. Then I discovered what others had discovered before me… and it seemed to cause more trouble than I was trying to avoid!

My search led me here:

Convert List to String

To be more close to solution it is necessary to see your input data and desirable result. Or current WF with data.

Which then led me here, with a node by Vernalis:

Collection to String

and that got it down to three nodes with no coded logic (albeit needing to use what appears to be a useful community node). Any other takers?

Yes I agree, a fun challenge!

KNIME_abbreviate 3.knwf (35.3 KB)

[Daniel_Weikert]

yes I used the palladian regex extractor.
br