Reading CSV with row.names by fread()

Question

I ran into an issue with fread() function from data.table package. I know it is still experimental, but maybe I am making some mistake somewhere.

Here is reproducible example:

library(data.table)
test <- data.frame(a=rnorm(300), b=rnorm(300))
write.csv(test,"a.csv")
fread("a.csv")

Gives an error:

Error in rbindlist(allargs) : 
  Item 2 has 2 columns, inconsistent with item 1 which has 3 columns

And side question: Why should I ever leave row.names=TRUE in write.csv question? So far I had only problems with it because it adds one unnamed column to the data.

Thx.

Answer 1

As a workaround you can read the rownames as a new column, by setting header=FALSE

fread("a.csv",header=FALSE)
header' changed by user from 'auto' to FALSE
      V1                 V2                 V3
  1:                      a                  b
  2:   1  -1.55640470495795    -1.344760319214
  3:   2   2.89752713867643   2.48413035874463
  4:   3 -0.493990961968582  0.119727513514055
  5:   4  0.559770137546773   1.07420769675405
 ---                                          
297: 296  0.585750601363698  -1.59845801200953
298: 297 -0.867339301988422  0.776738489388772
299: 298 0.0942821874550108 -0.649440075398178
300: 299 -0.308039637386426 -0.840171787291445
301: 300  0.358526722813896    -1.362322309472

From the help of fread, it looks that all the examples are using row.names=FALSE, so as you have mentioned this works fine:

write.csv(test,"b.csv",row.names=FALSE)
fread("b.csv")