[datatable-help] suggestion for transform.data.table.Rd

[Matthew Dowle]

Ah yes, will do. Thanks again. Automatically part of #2316 by dint of link
back to this thread.

> One other small thing in the docs along these lines:  Now that
> assignment by reference does not print by default, the last few
> examples in ?data.table kind of lose something.
>
> e.g. instead of
>     DT[, z:=42L]
>
> maybe make it one of these?
>     DT[, z:=42][]
>     (DT[, z:=42L])
>     print(DT[, z:=42L])
>     DT[, z:=42L]; DT
>
> Best,
> Garrett
>
> On Tue, Oct 16, 2012 at 4:40 AM, Matthew Dowle <mdowle at mdowle.plus.com>
> wrote:
>>
>> Well spotted, thanks. Will do. Filed as #2316 linking back to this
>> thread.
>>
>> And will include this in ?transform.data.table :
>>
>>     dt[, transform(.SD, c=max(b)), by="a"]
>>     dt[, c:=max(b), by="a"]   # same, but much much faster
>>
>> Btw, just to highlight to all that `:=`() is working, and by group, in
>> 1.8.3 now. This makes multiple := by group particularly elegant. Thanks
>> to
>> Thell Fowler for the suggestion.
>>
>> DT[, `:=`(newcol1:=max(a),
>>           newcol2:=min(b),
>>           existingcol:=max(c),
>>           newcol3:=mean(a)), by=grp]
>>
>> Rather than multiple LHS and having to set with=FALSE, which also now
>> works by group and you no longer need with=FALSE. But `:=`(...) gets
>> each
>> LHS and RHS together on one line in pairs, so makes it easier to read.
>>
>> For people who have joined the list recently, the latest
>> info/documentation is always in NEWS first. It usually takes a version
>> or
>> two for it to be fleshed out in the help files. "NEWS" is considered
>> documentation :
>>
>> https://r-forge.r-project.org/scm/viewvc.php/pkg/NEWS?view=markup&root=datatable
>>
>>
>>> It was not clear to me from the help page when I should use
>>> `transform` or `within` instead of `:=`.
>>>
>>> May I suggest copying the following comment (found in data.table.R) to
>>> the help page?
>>>
>>> # `within` and other similar functions in data.table are
>>> # not just provided for users who expect them to work, or
>>> # prefer syntax they're used to, but for non-data.table-aware
>>> # packages to retain keys (for example). Hopefully users and
>>> # code will use the faster data.table syntax in time.
>>>
>>> Also, the examples section of ?transform.data.table has the following
>>> commented out "example"
>>>
>>>     # dt[, transform, c = max(b), by="a"]  # like "ave"
>>>
>>> If run, this will give an error
>>>     Error in `[.data.table`(dt, , transform, c = max(b), by = "a") :
>>>       unused argument(s) (c = max(b))
>>>
>>> Based on test 104, I think that should be changed to
>>>
>>>     dt[, transform(.SD, c=max(b)), by="a"]
>>>
>>> Best,
>>> Garrett
>>> _______________________________________________
>>> datatable-help mailing list
>>> datatable-help at lists.r-forge.r-project.org
>>> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/datatable-help
>>>
>>
>>
>