write_csv should have option to quote character variables

[rmcd1024]

Perhaps I've missed an option, but I don't see how to have write_csv quote strings. This creates problems when strings look like integers.

library(readr)
write_csv(data.frame(a=c(rep("10", 1000), "01")), 'test1.csv')
x1 = read_csv('test1.csv')  ## a imported as integer, breaking last observation
write_csv(data.frame(a=c("01", rep("10", 1000))), 'test2.csv')
x2 = read_csv('test2.csv')  ## a imported correctly as character

Obviously in the read statements one could set guess_max=Inf, but it would be helpful to write quoted strings in the first place.

[jimhester]

Ok, I agree it should be possible to force quoting if desired.

[jxu]

Agreed. See quote argument of write.table.
I think there should be three options: Always quote character vectors, quote when needed (the current default), and never quote (dangerous, but here to mimic the functionality of write.table).

[bbolker]

+1000. Protecting text fields with commas in them is super-important! Falling back to write.csv() now ...

[jimhester]

@bbolker readr already quotes text fields with commas automatically, this issue is about forcing quoting even if it isn't strictly needed.

[bbolker]

sorry, I ran into some issues last night that made me think otherwise, but it was late. (Actually I think the problem was OpenOffice deciding it should split fields on semicolons as well as commas ...)

[mmuurr]

I'm curious if this open issue is truly open, or if there's a workaround that's commonly used (apart from write.csv and family). I, too, would like to strictly enforce quote-wrapping of certain character columns while writing those columns...

[jimhester]

One straightforward way to do this is to call base::dQuote() or glue::double_quote() on the columns you want to quote before passing to readr.

[philpav]

Is there still no built-in way for forcing quotes on all columns?

[sjackman]

Is there still no built-in way for forcing quotes on all columns?

write_csv(x, file, quote = "all")