Pattern to keep only certain column

Fred_ae · ‎02-28-2022

Hello, I'm new on dataiku, and I have to use for first time pattern to keep column.

I have this database with 6 column :

CA 2020 CA 2021 FOOD SERVICE EFFECTIF 2020 EFFECTIF 2021

2000 300000 MC DO HOTEL 220 240

And I go on prepare ( recipe) with keep only columns.

My aim is to have 3 dataset :

1- only column with number

2- only column who start per CA

3- only column who start per effectif.

I tried a lot of things to have it , but I have always no result.

So I would like to know how I can do it :

thanks in advance

tgb417 · ‎03-01-2022

@Fred_ae

Welcome to the Dataiku community. We are glad that you have joined us.

If I was in a situation like I think your are describing I’d likely use the remove delete column step using a regular expression or Regex as the filter pattern. Here is a thread from the community about this general topic.

https://community.dataiku.com/t5/Using-Dataiku/Remove-columns-by-pattern/m-p/15817

Regarding the regular expressions they can be a bit of a beast.

The thing to know is that the pipe symbol “|” is the alternative operator. You can in a signal regular expression find each of the kinds of your columns in one step. Separating each alternative with a “|”

here is a bit about the alternative operator.

https://www.regular-expressions.info/alternation.html

There are some tools out there that you can play with regular expressions. They have been of some help to me over the years.

https://regexr.com/

others please jump in and lend a hand.

@Fred_ae let us know how you are getting along.

--Tom

Fred_ae · ‎03-01-2022

Thanks, I feed solution to have only column who start per 'CA'

I dit :

^CA.*

But now i'm trying to have only column who contain the number 20

to have all column with 2021 2022 2023 ...

thanks again

tgb417 · ‎03-02-2022

For the numerical column names, You might try something like

^\d*

\d is the symbol for a numerical digit.

You then might combine the two as

^CA.*|^\d*

I’m not at a computer with DSS access so I have not been able to test the above.

in the past there was a bug where sometimes one had to use a double backslash in some cases. If you are having problems making this work try \\ rather than a single \.

hope this helps a bit. Let us know how you are getting on with this.

--Tom