Length one character vector always guessed as character

[grandtiger]

I have trouble attaching files, so I just put the content of the files here since they are simple.

I found a surprising behavior using read_csv function in readr package. The format of date is different when the input file has only one row vs. multiple rows.

Could you please take a look and see if this is a bug or user error?

Thanks a lot!

file1.csv

"COL11","COL21","12/09/15"

file2.csv

"COL11","COL21","12/09/15"
"COL12","COL22","12/10/15"
"COL13","COL23","12/11/15"

readr_test.R

library("readr")
default.locale <- locale(date_names = "en", date_format = "%m/%d/%y",
                         time_format = "%H:%M:%S", decimal_mark = ".", grouping_mark = NULL,
                         tz = "UTC", encoding = "UTF-8", asciify = FALSE)
file1 <- read_csv('file1.csv', locale = default.locale, col_names = FALSE)
colnames(file1) <- c("COL1","COL2","DATE")
View(file1)
class(file1$DATE)
file2 <- read_csv('file2.csv', locale = default.locale, col_names = FALSE)
colnames(file2) <- c("COL1","COL2","DATE")
View(file2)
class(file2$DATE)
print(class(file1$DATE) == class(file2$DATE))

[jennybc]

I tried to figure out if this is related to a problem I'm having. It is not.

But since I have it, here's the example above, stripped way down:

library("readr")
(a <- read_csv("2015-12-09\n", col_names = FALSE))
#> Source: local data frame [1 x 1]
#> 
#>           X1
#>        (chr)
#> 1 2015-12-09
(b <- read_csv("2015-12-09\n2015-12-10", col_names = FALSE))
#> Source: local data frame [2 x 1]
#> 
#>           X1
#>       (date)
#> 1 2015-12-09
#> 2 2015-12-10
class(a$X1)
#> [1] "character"
class(b$X1)
#> [1] "Date"

[hadley]

Even more minimal:

read_csv("1\n", col_names = FALSE)
#> Source: local data frame [1 x 1]
#> 
#>      X1
#>   <chr>
#> 1     1

Likely to be an off-by-one error

[jimhester]

After bisecting 0cb2d0b is the commit that introduced this bug