Codecov Report

❗ No coverage uploaded for pull request base (master@58c753e). Click here to learn what that means.
The diff coverage is 100%.

@@            Coverage Diff            @@
##             master    #2517   +/-   ##
=========================================
  Coverage          ?   91.44%           
=========================================
  Files             ?       63           
  Lines             ?    12067           
  Branches          ?        0           
=========================================
  Hits              ?    11035           
  Misses            ?     1032           
  Partials          ?        0

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 58c753e...67ae0ac. Read the comment docs.

Not sure about this.
This PR has this behaviour :

> fread("A\n")
Empty data.table (0 rows) of 1 col: A
> fread("A\n\n")
    A
1: NA
> fread("A\n\n\n")
    A
1: NA
2: NA
> fread("A,B\n")
Empty data.table (0 rows) of 2 cols: A,B
> fread("A,B\n\n")
Empty data.table (0 rows) of 2 cols: A,B
> fread("A,B\n\n\n")
Empty data.table (0 rows) of 2 cols: A,B
>

which would be a change from CRAN version which does this :

> fread("A\n")
Empty data.table (0 rows) of 1 col: A
> fread("A\n\n")
Empty data.table (0 rows) of 1 col: A
> fread("A\n\n\n")
Empty data.table (0 rows) of 1 col: A
> fread("A,B\n")
Empty data.table (0 rows) of 2 cols: A,B
> fread("A,B\n\n")
Empty data.table (0 rows) of 2 cols: A,B
> fread("A,B\n\n\n")
Empty data.table (0 rows) of 2 cols: A,B

Thoughts?

Consider this file:

Aa,Bb
10,20

30,40

Technically, this is not a valid CSV -- there shouldn't be a blank line there, but still, if this is the data, what do you do? You could choose a different coping strategy: (1) set fill=TRUE and fill the second line with NAs; (2) set blank.lines.skip=TRUE and ignore the second line; (3) the default action is to stop reading at the first blank line:

> fread("Aa,Bb\n10,20\n\n30,40\n")
    Aa Bb
1: 10 20
Warning message:
In fread("Aa,Bb\n10,20\n\n30,40\n") :
  Found the last consistent line but text exists afterwards. Consider fill=TRUE and/or blank.lines.skip=TRUE. First 200 characters of discarded line: <<30,40>>

Similarly, if the blank line occurs at the end, only this time fread doesn't display any warning:

> fread("Aa,Bb\n10,20\n30,40\n\n")
    Aa Bb
1: 10 20
2: 30 40

In both cases the CSV is still invalid, and fread deals with it in a way that it thinks is "most reasonable".

Now let's go back to the case in question. Consider the following file:

Aa
10

30

Unlike the first case, this is already a valid CSV: there is only one way to read it:

> fread("Aa\n10\n\n20\n")
   Aa
1: 10
2: NA
3: 20

Similarly, if the empty line occurs at the end of the file, it still remains a valid CSV file, and it should be read as such:

> fread("Aa\n10\n20\n\n")  # expected output:
   Aa
1: 10
2: 20
3: NA

The fact that it doesn't I view as a bug:

> fread("Aa\n10\n20\n\n")  # actual output
   Aa
1: 10
2: 20

To recap, in the first case (2+ -column file) blank lines make it an invalid CSV file, and there could be difference in opinion as to what coping strategy is best. At the same time, in the second case (1-column file) blank lines are allowed in a valid CSV file, and therefore they should be treated accordingly.