Override column types when importing data using readr::read_csv() when there are many columns

Question

I am trying to read a csv file using readr::read_csv in R. The csv file that I am importing has about 150 columns, I am just including the first few columns for the example. I am looking to override the second column from the default type (which is date when I do read_csv) to character, or other date format.

GIS Join Match Code Data File Year  State Name  State Code  County Name County   Code   Area Name   Persons: Total G0100010    2008-2012   Alabama 1   Autauga County  1   Autauga County, Alabama 54590  df <- data.frame("GIS Join Match Code"="G0100010", "Data File" = "2008-2012", "State" = "Alabama", "County" = "Autauga County", "Population" = 54590) 

The issue is that when I use readr::read_csv, it seems I may have to use all variables while overriding in the col_types (see error below). That is need to specify overriding all the 150 columns individually(?).. The question is that : Is there a way to specify overriding the col_type of just specific columns, or a named list of objects? In my case, it would be just overriding the column "Data File Year".

I understand that any omitted columns will be automatically parsed, which is fine for my analysis. I think it gets further complex as the column names have a space in them in the file I downloaded (For e.g., "Data File Year", "State Code") etc.

tempdata <- read_csv(df, col_types = "cc") Error: You have 135 column names, but 2 columns 

The Other option I guess, if possible, is to just skip reading the second column all together?

Answer 1

Here follows a more generic answer to this question if someone happens to stumble upon this in the future. It is less advisable to use "skip" to jump columns as this will fail to work if the imported data source structure is changed.

It could be easier in your example to simply set a default column type, and then define any columns that differ from the default.

E.g., if all columns typically are "d", but the date column should be "D", load the data as follows:

  read_csv(df, col_types = cols(.default = "d", date = "D")) 

or if, e.g., column date should be "D" and column "xxx" be "i", do so as follows:

  read_csv(df, col_types = cols(.default = "d", date = "D", xxx = "i")) 

The use of "default" above is powerful if you have multiple columns and only specific exceptions (such as "date" and "xxx").

Answer 2

Yes. For example to force numeric data to be treated as characters:

examplecsv = "a,b,c\n1,2,a\n3,4,d" read_csv(examplecsv) # A tibble: 2 x 3 #      a     b     c #  <int> <int> <chr> #1     1     2     a #2     3     4     d read_csv(examplecsv, col_types = cols(b = col_character())) # A tibble: 2 x 3 #      a     b     c #  <int> <chr> <chr> #1     1     2     a #2     3     4     d 

Choices are:

col_character()  col_date() col_time()  col_datetime()  col_double()  col_factor() # to enforce, will never be guessed col_integer()  col_logical()  col_number()  col_skip() # to force skip column 

More: http://readr.tidyverse.org/articles/readr.html