Allow 0 and 1 singletons to coerce to logical without warning in set() and := ?

Highlighted via http://stackoverflow.com/questions/7235657/fastest-way-to-replace-nas-in-a-large-data-table/7249454?noredirect=1#comment43831879_7249454

DT = data.table(a=1:4, b=c(FALSE, TRUE, NA, FALSE))
DT
#    a     b
#1: 1 FALSE
#2: 2  TRUE
#3: 3    NA
#4: 4 FALSE
set(DT,3L,1L,0)  # ok, no warning
set(DT,3L,2L,0)  # should be ok too without warning
# Warning message:
# In set(DT, 3L, 2L, 0) :
#   Coerced 'double' RHS to 'logical' to match the column's type; may have truncated precision. Either 
# change the target column to 'double' first (by creating a new 'double' vector length 4 (nrows of entire 
# table) and assign that; i.e. 'replace' column), or coerce RHS to 'logical' (e.g. 1L, NA_[real|integer]_, as.*, 
# etc) to make your intent clear and for speed. Or, set the column type correctly up front when you create 
# the table and stick to it, please.
DT
#    a     b
#1: 1 FALSE
#2: 2  TRUE
#3: 0 FALSE
#4: 4 FALSE

Apply to 0, 1, 0L and 1L. Just length-1 vectors (i.e. singletons), not longer, as already in place for 0 and 1 to integer.

Maybe leave the warning in place for speed (we want to be warned about unnecessary repetitive coercing). Or just leave the warning in place for set() (whose purpose is to be iterated) and drop just for :=.

I prefer to leave the warning, because usually (for me at least) this warning is only produced when I've been sloppy & tells me I need to go back and use 0L, F, etc. as appropriate.

I would be fine with an option to trigger this, of course.

Has our experience with logical01 in fread/fwrite informed this issue at all?

IMO because set is meant to be low overhead it is best to avoid coercion and leave the warning

on the other hand := should be more relaxed and not raise warnings in cases like

DT1 = data.table(id = c(rep(1L,4L), 2L, 2L, 1L, 2L, 1L),
               code = c("c","b","c","c","a","a","b","a","c"),
               valA = 1:9, valB = 10:18,
               stringsAsFactors = FALSE)
DT1[code=="c", valB := valB+1]

Warning message:
In `[.data.table`(DT1, code == "c", `:=`(valB, valB + 1)) :
  Coerced double RHS to integer to match the type of the target column (column 4 named 'valB'). The RHS values contain no fractions so would be more efficiently created as integer. Consider using R's 'L' postfix (typeof(0L) vs typeof(0)) to create constants as integer and avoid this warning. Wrapping the RHS with as.integer() will avoid this warning too but it's better if possible to create the RHS as integer in the first place so that the cost of the coercion can be avoided.

Navigation Menu

Repository navigation

Allow 0 and 1 singletons to coerce to logical without warning in `set()` and `:=` ? #996

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Allow 0 and 1 singletons to coerce to logical without warning in set() and := ? #996

Description

Activity

mattdowle commented on Jan 5, 2015

MichaelChirico commented on Jul 27, 2015

MichaelChirico commented on May 23, 2019

jangorecki commented on May 24, 2019

jangorecki commented on Sep 20, 2019

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Allow 0 and 1 singletons to coerce to logical without warning in `set()` and `:=` ? #996