write_csv silently fails when directory doesn't exist

[dchudz]

write_csv(cars, "this/directory/doesnt/exist/cars.csv")

... doesn't succeed, but also doesn't give an error (latest master, and the version on CRAN). Seems like it would be better to have an error (like write.csv does)?

My system details, in case that matters:

                                                                                           sysname 
                                                                                          "Darwin" 
                                                                                           release 
                                                                                          "13.1.0" 
                                                                                           version 
"Darwin Kernel Version 13.1.0: Thu Jan 16 19:40:37 PST 2014; root:xnu-2422.90.20~2/RELEASE_X86_64" 
                                                                                          nodename 
                                                                                    "davids-mbp-4" 
                                                                                           machine 
                                                                                          "x86_64" 
                                                                                             login 
                                                                                           "david" 
                                                                                              user 
                                                                                           "david" 
                                                                                    effective_user 
                                                                                           "david" 
> 

[dchudz]

Hmm, I looks like maybe this is intentional? (mustWork=FALSE in normalizePath: https://github.com/hadley/readr/blob/master/R/write.R#L33).

But I don't understand why.

[hadley]

The file won't exist yet, so normalizePath() will warn otherwise. I think the error checking needs to happen in the C++

[dchudz]

Ah right, sorry!

(It wouldn't be good to handle it with an assertion that file.exists(dirname(path))?)