Mel,

The usage would be something like:

DT[, from:to, with=FALSE]
# or
DT[, .SD, .SDcols = from:to]
where from and to are the start and end column names. I agree there’s no real 
advantage in terms of typing/prone to errors.

There might be some merit in readability, as people normally remember column 
names and not numbers… And this allows you to refer to the names directly 
without having to type DT and then look up the column or use a match() to find 
out the column programatically or do:

DT[, .SD, .SDcols = names(DT)[some_idx]]


-- 
Arun

On 10 Feb 2015 at 22:39:14, Bacou, Melanie ([email protected]) wrote:

Everyone,

The varA...varZ construct is borrowed from STATA syntax. Probably a reason why 
it got into subset() in the first place, though definitely not very R-like. In 
fact I’ve never come across this construct in R before and had no idea it was 
actually working either!

I’m not sure dt[, .SD, .SDcols=list(varA...varZ)] is less typing, less prone to 
error, or more readable than dt[, .SD, .SDcols=names(dt)[1:24] and using 
indices is also more flexible (what about if we want more complex sequences). I 
can see one use case for this syntax though if dt might change over time but 
variables always come in known sequences.

Not sure we should really encourage it — but agreed with Arun, if it’s in 
base::subset() then no reason why not.

—Mel.

On 2/10/2015 1:50 PM, Arunkumar Srinivasan wrote:

I had the same reaction when I found out ‘subset’ already did this :-).

I’ve the same impression that it’s a bit odd, even though some people prefer 
it..

Arun

On 10 Feb 2015 at 19:39:29, Chris Neff ([email protected]) wrote:

Wow, didn’t realize that worked! So there is precedent then. It just looks 
funny to me, but you are right it is easily avoided. I just didn’t want to see 
more divergence from subset and data.frame logic, but since this already works 
with subset that’s fine.
On Tue Feb 10 2015 at 1:34:03 PM Arunkumar Srinivasan [email protected] 
wrote:

Chris,
But what’s the problem? You can simply not use it?
It’s not that uncommon. `base::subset()` does this.
--  
Arun

On 10 Feb 2015 at 19:31:43, Chris Neff ([email protected]) wrote:

I don't like this idea. It adds extra that it doesn't need to.  Doing it with 
column numbers is more straightforward, and if all you have is names you can 
get numbers by doing match() or whatever and then getting the sequence with 
seq(). Having a sequence of column names is odd.
On Tue Feb 10 2015 at 1:28:25 PM Arunkumar Srinivasan <[email protected]> 
wrote:

    Farrel,
    It could be useful. Please file an issue on the github project page. Thanks.
    --  
    Arun

    On 10 Feb 2015 at 01:08:46, Farrel Buchinsky ([email protected]) wrote:

    So lets say one has a data.table with the following columns
    first.name, last.name, height, weight, shoe.size, eye.color, hair.length, 
appendage.size, ear.length
    If one wanted to just include weight through hair.length one would have to 
go something such as this
    dt[,list(weight, shoe.size, eye.color, hair.length)]
    Is there a way to do something along the lines of
    dt[,list(weight...hair.length)]
    If so, can you direct me to the documentation? If not can you build it? Is 
it difficult? Some data.tables have many columns.
    Thanking you in anticipation.
    Farrel
    _______________________________________________
    datatable-help mailing list
    [email protected]
    https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
    _______________________________________________
    datatable-help mailing list
    [email protected]
    https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help

datatable-help mailing list
[email protected]
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help

--
Melanie BACOU
International Food Policy Research Institute
Snr. Program Manager, HarvestChoice
Work +1(202)862-5699
E-mail [email protected]
Visit www.harvestchoice.org
_______________________________________________
datatable-help mailing list
[email protected]
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help

Reply via email to