Bug: `.GRP` should create column named `GRP`, not `.GRP` (a la `.I`, `.N`)

[MichaelChirico]

That .GRP returns a variable named .GRP by default is causing some errors.

Consider:

data.table(id=rep(1:3,each=5),yr=rep(rep(1:3,3),rep(c(2,1,2),3)))[,.GRP,by=.(id,yr)][,!.GRP%in%range(.GRP),by=id]

(Basically, trying to eliminate the first and last yr of observation from each id)

This is an error: cannot change value of locked binding for '.GRP'. I can think of a number of workarounds, but it seems the simpler solution is to simply change the naming behavior.

I imagine this is the same reason why .I and .N return columns named I and N by default.

[eantonya]

I haven't thought this through very carefully, but I assume whatever argument applies to .GRP also applies to .BY.

[jangorecki]

Not sure but maybe it is only a matter of adjusting regex in two lines of [.data.table:
jvnames = gsub("^[.]([NI])$","\\1",as.character(jsub)) in data.table.R#L896
jvnames[jj-1L] = gsub("^[.]([NI])$","\\1",deparse(jsubl[[jj]])) in data.table.R#L907

[MichaelChirico]

@jangorecki perhaps these lines as well

[arunsrinivasan]

@MichaelChirico no, just the ones Jan's pointed. Would be great to wrap this up..

[MichaelChirico]

@arunsrinivasan gonna take care of it now. Is there anything that needs to be changed in the tests / manual for this?