[R-Forge #2793] Allow subsets of LHS of := #571

Activity

1. jangorecki commented on Aug 20, 2014

   

   jangorecki

   on Aug 20, 2014

   Member

   More actions

   Also mentioned in nice use case here: http://stackoverflow.com/questions/21819253/how-to-change-the-last-value-in-each-group-by-reference-in-data-table
2. matthieugomez commented on Sep 1, 2014

   

   matthieugomez

   on Sep 1, 2014

   Contributor

   More actions

   Would not be the same thing as dt[b<2, a := 0] ?
   Maybe it is what is implied by the first post, but, indeed, I think that being able to use relative index for conditions could be nice
   For now, the syntax to do the same is cumbersome.

    DT[DT[,index=.I[expression],by=id]$index,a:=999]
   

   It would be simpler to be able to do things such as

       DT[, a[1] := 999,by=id]
       DT[, a[.N] := 999,by=id]
       DT[, a[which.max(b)]:= 999,by=id]
   
3. arunsrinivasan

   added

   High

   on Sep 24, 2014
4. franknarf1 commented on Sep 17, 2015

   

   franknarf1

   on Sep 17, 2015

   Contributor

   More actions

   I think this is a good example (in the comment): http://stackoverflow.com/questions/32638845/replacing-nas-in-dataframe-with-constants-in-another-in-r/32639171#comment53127039_32639171

   Instead of

   DF1[DF2, x := ifelse(is.na(x), i.x, x), on=c("y","z")]
   

   I'd like to do

   DF1[DF2, x[is.na(x)] := i.x, on=c("y","z")]
   

   👍React with 👍1Reacted by Michael Chirico
5. franknarf1 commented on Jan 18, 2016

   

   franknarf1

   on Jan 18, 2016

   Last edited by franknarf1

   Contributor

   More actions

   This is another good example, I think: http://stackoverflow.com/q/34845396

   And another not-bad example: http://stackoverflow.com/q/37862828/

   dat = setDT(structure(list(diameter = c(1L, 1L, 1L, 2L, 2L, 3L), thread = c(4L, 
   6L, 4L, 5L, 7L, 9L), size = c("small", "small", "small", "small", 
   "small", "small")), .Names = c("diameter", "thread", "size"), row.names = c(NA, 
   -6L), class = "data.frame"))
   
   dat[, size := "small"][, size[ thread == max(thread) ] := "large", by = diameter]
   

   And another not-bad example: http://stackoverflow.com/q/38083937/

   library(data.table)
   DT <- data.table(c(1, 1, 1, 1, 1, 2, 2, 2, 2, 2),
                    as.IDate(c("2015-12-31", "2014-12-31", "2013-12-31", "2012-12-31", 
                               "2011-12-31", "2015-12-31", "2014-12-31", "2014-01-25",
                               "2013-01-25", "2012-01-25")))
   setnames(DT, c("id", "date"))
   
   # desired syntax
   DT[, r[month(date) == month(date)[1L]] := year(date) - year(date)[1L], by=id]
   

   And another one: http://stackoverflow.com/q/38910869/

   library(data.table)
   dt_A <- data.table(key1 = LETTERS[1:6], key2 = (1:6), flag=NA_character_ )
   dt_A[4, flag:="exclude"]
   dt_B <- data.table(key1 = LETTERS[3:6], key2 = c(3:5,0L)) 
   
   # desired syntax
   dt_A[dt_B, flag[ flag != "exclude" ] := "include" ]
   
   # In this case, there's an alternative, 
   # but I'm not sure about performance, since it's doing like two joins ...
   
   dt_A[fsetdiff(dt_B, dt_A[flag=="exclude", names(dt_B), with=FALSE]), on=names(dt_B), 
     flag := "include"]
   

   And a bad example: https://stackoverflow.com/a/44525179/ like DT[, x[.N] := sum(v), by=g]

   Another x[.N] := one: https://stackoverflow.com/q/50821561/
6. DavidArenburg commented on Aug 19, 2016

   

   DavidArenburg

   on Aug 19, 2016

   Last edited by DavidArenburg

   Member

   More actions

   Another good example of a possibly useful use case on SO http://stackoverflow.com/a/39035748/3001626

   Also, this is labeled as "high" but it is not assigned to any milestone...
7. MichaelChirico commented on Sep 21, 2016

   

   MichaelChirico

   on Sep 21, 2016

   Member

   More actions

   Agree that the best use cases for this appear to be when combined with a merge in i, or when assigning to the first element of a by group, so that the conditioning is either much harder or impossible to do in i directly. Reverting to an rbindlapply approach to build a DT from scratch for now
8. jangorecki

   removed

   High

   on Jun 8, 2020