[datatable-help] Preserving comments in [.data.table?

[Matthew Dowle]

> On 11/09/12 12:32, Matthew Dowle wrote:
>>> [...]
>>> I think we want column attributes. As in: all of them. The comment is
>>> just one attribute. R uses attributes for all sorts of sneaky things so
>>> I think it is a bad idea if we lose some of them some of the time.
>> Oh I see. Please file a bug.report(package="data.table") something like:
>> "Column attributes (such as 'comment') are sometimes lost."
>
> Done as #2270

Thanks, will do.

>
>> [...]
>> we'd need setcomment() [or
>> setlabel(), or both] to avoid that copy; adding to the set* family.
>>> A set attributes that allows setting column attributes may suffice.
>> Already exists: try setattr(). Works on anything.
>
> Documentation says it only works for x= of type data.table. The column
> of a data.table will rarely be of this type....? I think we need
> setcolumnattr() or similar to do what you want. But copying a single
> column is not *that* bad [1], so I suggest fixing the lost attributes is
> the first priority.

Documentation is misleading/incomplete. setattr() accepts x of any type,
as well as data.table. I'll modify #2270 to include documenting this.
Thanks.

> DT = data.table(a=1:3,b=4:6)
> .Internal(inspect(DT))
@0x000000001e8487e0 19 VECSXP g0c7 [OBJ,NAM(2),ATT] (len=2, tl=100)
  @0x000000000cc6a668 13 INTSXP g0c2 [NAM(2)] (len=3, tl=0) 1,2,3
  @0x000000000cc687d0 13 INTSXP g0c2 [NAM(2)] (len=3, tl=0) 4,5,6
[ ... snip attribs ... ]
> setattr(DT$b,"comment","Test comment")
> str(DT)
Classes ‘data.table’ and 'data.frame':  3 obs. of  2 variables:
 $ a: int  1 2 3
 $ b: atomic  4 5 6
  ..- attr(*, "comment")= chr "Test comment"
 - attr(*, ".internal.selfref")=<externalptr>
> .Internal(inspect(DT))
@0x000000001e8487e0 19 VECSXP g0c7 [OBJ,NAM(2),ATT] (len=2, tl=100)
  @0x000000000cc6a668 13 INTSXP g0c2 [NAM(2)] (len=3, tl=0) 1,2,3
  @0x000000000cc687d0 13 INTSXP g0c2 [NAM(2),ATT] (len=3, tl=0) 4,5,6
[ ... snip attribs ... ]

The hex addresses are unchanged and therefore show no copy of any of the
vectors was taken, but the column's attribute was changed by reference by
setattr as needed.

>
> Allan
> [1] You would typically set the attributes when you create the column
> and then leave them well alone. It is only because you keep deleting
> them that I keep going back and re-setting them.

Ok, understood. Thanks.