Suggested new convenience function: recast()

[foundinblank]

Tweeted about this earlier this week. Essentially, I do this a lot, and imagine many others do, too.

... %>% mutate(col1 = as.character(col1)) %>% ... 
... %>% mutate(col2 = as.integer(col2)) %>% ... 

It's clunky typing the column name twice, and while there's also mutate_at(vars(column_name), as.character), that also feels clunky too. I propose a convenience wrapper function, recast() such that:

... %>% recast(col1,col2, as.character) %>% ...
... %>% recast(col3,col4, as.numeric) %>% ...

(Thanks to JLS_Datascience for the idea.)

I'd like to try writing a function for this and then do a PR. I can model it off a similar convenience wrapper (which one?).

Can I?

[krlmlr]

Thanks. Have you seen mutate_at()?

[krlmlr]

I missed the part where you wrote about mutate_at() .

My gut feeling is that such a function should live elsewhere, partly because this functionality already can be achieved with existing tools. The vars() is a technical necessity for supporting more than one function, but doesn't add too much cognitive burden IMO.

#3433 is about providing a better interface for passing the functions.

[hadley]

And generally, I think you should solve type problems further up the pipeline (e.g. during import)

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/