Data.table: fread input random lines. Sampling

[skanskan]

Hello.
It would be great if data.table's function fread (or other import functions) were able to read random rows from a file, I think this is also called "sampling" lines (nonconsecutive rows).

As far as I know it can only read a single contiguous range.

There could be two options:

• The user could pass it a vector of numbers specifying the lines to be read.
• Or, the easiest way, just leave fread to read the rows he wants, for example using uniform random number generator to choose them.

Reading a file in this way would allow the user to have a quick glimpse of very big files.

[mgahan]

The shuf command already allows for this.

fread("shuf -n 5 iris.csv")

[RanaivosonHerimanitra]

+1

[skanskan]

@mgahan
I get this error:
'shuf' is not recognized as an internal or external command,

Maybe that's only available on linux. Isn't it?

[mgahan]

Good point @skanskan. I am not sure how to do this on Windows.

[skanskan]

That's why I think it could be implemented directly on fread, I think it shouldn't be difficult.

[MichaelChirico]

This is basically #583 in disguise.