col_number() drops negatives

[gregmacfarlane]

If I don't request specific column types, then read_csv treats negative numbers appropriately.

readr::read_csv("A,B\n-4,5G")

##    A B
## 1 -4 5

If I specify that the column is numeric, then read_csv drops the negative signs.

readr::read_csv("A,B\n-4,5G", col_types = "nn")

##   A B
## 1 4 5

[dyndna]

Specifying col_types= "d" will fix it, although I thought that n should be numeric and should not convert negative to positive values.

@hadley: Thanks for detailed documentation which I did not read until I noticed col_types issue in my data set. Following helped me to understand col_types.

?read_tsv and vignette("column-types")

[rkarimi]

Had the same problem, and was not sure what is the difference between n and d!

[ctmann]

This introduces a new problem; d converts a number with commas to NA, but at least it warns the user. To confirm, please try the code below using the data on the attached spreadsheet, importing it using three different techniques (base R, readr without explicitly decaring coltypes, and finally, using your solution)

unzip_folder <- "~/R/TestBed/"
file <- "testbed.csv"
filename <- paste0(unzip_folder,file)


read.csv(filename)


library(readr)
read_csv(filename)


read_csv(filename, col_types = cols(
  Name = "c",
  Amounts = "d"))

testbed.zip

[hadley]

This issue actually illustrates two problems:

1. parse_number("-1) returns 1
2. collector_guess("1G") returns "number"