Enhancement: Option to write the keys used for the join

[datatraveller1]

I'd like to suggest a join and joinp option to write the keys used for a join to a file:
--keys-output <file> Write keys to <file>

e.g. if the join is:
qsv joinp --keys-output=keys.csv category,class file1.csv category,class file2.csv -o values.csv

then --keys-output=keys.csv would write the keys (category,class) for which the join was successfull to the file keys.csv.

I assume this could happen "for free" (without much additional coding), because these keys are used to build the join result.

[jqnatividad]

Hi @datatraveller1 ,
That's a good idea.
I implemented it in join.

I can't implement it in joinp however, as qsv just builds the join definition using the Polars join_builder and hands it off to the Polars engine.

[datatraveller1]

Hi @jqnatividad Thank you very much. 👍
It's a pity that it doesn't work for joinp, but I can use something like the following commands instead:

qsv joinp category,class file1.csv category,class file2.csv -o values.csv
qsv dedup --select category,class values.csv | qsv select category,class -o keys.csv

[jqnatividad]

I was actually thinking of removing join right after I added --ignore-case to joinp.

But now, it's clear that they serve different use cases.

Speaking of which, I just added non-equi join support to joinp. 😄

#2409

And as to your "recipe", that's great! That's why one of qsv's goal is composability.

My hope is that over time, we build up a recipe library (mixing in other common CLI commands), preferably a large part of it from the community that can be readily shared and reused.

It will take time, but we're already seeding it with qsv pro...