producing a column specification for type_convert()

[jennybc]

I'm learning the hard way that readr::type_convert() no longer takes column specifications as a character string, like cccnnn. Apparently I do this in googlesheets examples, tests, etc. Is there a specific reason to require the long form in type_convert()? If I have to workaround this myself, is there a better way than what I've got here? This looks a bit nutty.

x <- "cccnnn"
do.call(readr::cols, as.list(strsplit(x, "")[[1]]))
#> cols(
#>    = col_character(),
#>    = col_character(),
#>    = col_character(),
#>    = col_number(),
#>    = col_number(),
#>    = col_number()
#> )

[hadley]

Likely just a mistake

[jimhester]

This was changed in response to #369 and #431, where we explicitly disabled support for character vectors in type_convert(). There is even a blurb in ?type_convert (including a typo)

Unlike other functions ‘type_convert()’ does not allow
character specificatinos of ‘col_types’.

@jennybc Probably the easiest way to do what you want is to use as.col_spec(), but it is unexported :(.

readr::as.col_spec("cccnnn")
#> Error: 'as.col_spec' is not an exported object from 'namespace:readr'
readr:::as.col_spec("cccnnn")
#> cols(
#>    = col_character(),
#>    = col_character(),
#>    = col_character(),
#>    = col_number(),
#>    = col_number(),
#>    = col_number()
#> )

[hadley]

@jennybc what do you think we should do in this scenario?

library(tidyverse)
df <- tibble(x = 1:2, y = 1:2)
type_convert(df, col_types = "dd")

[jennybc]

I would expect x and y to be double.

[hadley]

What about type_convert(df, col_types = "cc") ?

[jennybc]

Then character?

[hadley]

Hmmm, that's considerably expanding the scope — currently it only operates on character columns.

[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/