Subsetting data.trames uses a syntax similar to tibble, or formulas for i,
j, and possibly by or keyby to use the data.table syntax instead.
# S3 method for class 'data.trame'
x[i, j, by, keyby, with = TRUE, drop = FALSE, ...]
set_(x, i, j, value, byref = FALSE)
let_(x, i = NULL, j = seq_along(x), value)Arguments
- x
A data.trame object.
- i
Selection of rows by indices, negative indices, logical or a formula
- j
Selection of columns by indices, negative indices, logical, names or a formula (both i and j must be formulas simultaneously). If
:=is used in the formula to create one or more new variables by reference, the expression must be placed between{}to avoid operators precedence issues, or better::=could be just replaced by~.- by
Grouping columns (must be a formula and
jmust be also provided as a formula)- keyby
Either
TRUE/FALSEifbyis provided, or a formula (andjmust also be provided as a formula)- with
Logical, whether to evaluate
jin the data.trame ifTRUEor in the calling environment ifFALSE(default isTRUE).with = FALSEis similar to tibble subsetting and it is forced wheniorjare not formulas.- drop
Coerce to a vector if the returned data.trame only has one column
- ...
Further arguments passed to the underlying
data.tablesubsetting- value
The value to insert as subassignment in a data.trame object.
- byref
Logical, whether to use by reference or not (
FALSEby default).
Value
A data.trame object, or a vector if drop = TRUE and the result has
only one column.
Examples
dtrm <- data.trame(
a = 1:3,
b = letters[1:3],
c = factor(LETTERS[1:3])
)
# Subsetting rows, the tibble-way
dtrm[1:2, ]
#> # A data.trame: [2 × 3]
#> a b c
#> <int> <chr> <fct>
#> 1 1 a A
#> 2 2 b B
dtrm[-1, ]
#> # A data.trame: [2 × 3]
#> a b c
#> <int> <chr> <fct>
#> 1 2 b B
#> 2 3 c C
dtrm[c(TRUE, FALSE, TRUE), ]
#> # A data.trame: [2 × 3]
#> a b c
#> <int> <chr> <fct>
#> 1 1 a A
#> 2 3 c C
# On the contrary to data.table, providing only one arg, means subsetting
# columns (like for data.frame or tibble)
dtrm[c(TRUE, FALSE, TRUE)]
#> # A data.trame: [3 × 2]
#> a c
#> <int> <fct>
#> 1 1 A
#> 2 2 B
#> 3 3 C
dtrm[dtrm$a > 1, ] # Must fully qualify the column name
#> # A data.trame: [2 × 3]
#> a b c
#> <int> <chr> <fct>
#> 1 2 b B
#> 2 3 c C
# Subsetting the data.table way, with formulas: no fully qualification needed
dtrm[~a > 1, ]
#> # A data.trame: [2 × 3]
#> a b c
#> <int> <chr> <fct>
#> 1 2 b B
#> 2 3 c C
# Subsetting the columns, the tibble way
dtrm[, 1:2]
#> # A data.trame: [3 × 2]
#> a b
#> <int> <chr>
#> 1 1 a
#> 2 2 b
#> 3 3 c
dtrm[, -1]
#> # A data.trame: [3 × 2]
#> b c
#> <chr> <fct>
#> 1 a A
#> 2 b B
#> 3 c C
dtrm[, c(TRUE, FALSE, TRUE)]
#> # A data.trame: [3 × 2]
#> a c
#> <int> <fct>
#> 1 1 A
#> 2 2 B
#> 3 3 C
dtrm[, c("a", "b")]
#> # A data.trame: [3 × 2]
#> a b
#> <int> <chr>
#> 1 1 a
#> 2 2 b
#> 3 3 c
# You must set drop = TRUE explicitly to return a vector
dtrm[, 2] # Still a data.trame, like tibble, but unlike the data.frame method
#> # A data.trame: [3 × 1]
#> b
#> <chr>
#> 1 a
#> 2 b
#> 3 c
dtrm[, 2, drop = TRUE] # Now a vector
#> [1] "a" "b" "c"
# The selection is referentially transparent, i.e., you can do:
sel <- c("c", "b")
dtrm[, sel]
#> # A data.trame: [3 × 2]
#> c b
#> <fct> <chr>
#> 1 A a
#> 2 B b
#> 3 C c
# Subsetting the columns, the data.table way, with formulas
dtrm[~1:2, ~.(b)]
#> b
#> <char>
#> 1: a
#> 2: b
dtrm[~1:2, ~b] # If not enclosed in .(), returns a vector instead
#> [1] "a" "b"
# Precautions are needed here because it is NOT referentially transparent:
dtrm[, ~..sel] # In data.table language, this is how you access `sel`
#> # A data.trame: [3 × 2]
#> c b
#> <fct> <chr>
#> 1 A a
#> 2 B b
#> 3 C c
# Extended data.table syntax using i, j, by, or keyby with formulas
# Warning: due to precedence of operators, you must use braces here!
dtrm[, ~{d := paste0(b, c)}] # Changed in place (by reference!)
#> # A data.trame: [3 × 4]
#> a b c d
#> <int> <chr> <fct> <chr>
#> 1 1 a A aA
#> 2 2 b B bB
#> 3 3 c C cC
# Another form that does not need braces, but is less readable:
dtrm[, ~`:=`(e, paste0(b, a))]
#> # A data.trame: [3 × 5]
#> a b c d e
#> <int> <chr> <fct> <chr> <chr>
#> 1 1 a A aA a1
#> 2 2 b B bB b2
#> 3 3 c C cC c3
# or equivalently:
dtrm[, ~let(e = paste0(b, a))]
#> # A data.trame: [3 × 5]
#> a b c d e
#> <int> <chr> <fct> <chr> <chr>
#> 1 1 a A aA a1
#> 2 2 b B bB b2
#> 3 3 c C cC c3
# In this case, it is much better to just replace `:=` by `~`, but internally
# it uses set(). It is faster, but much more limited and cannot use by or
# or keyby:
dtrm[, f ~ paste0(c, a)]
#> # A data.trame: [3 × 6]
#> a b c d e f
#> <int> <chr> <fct> <chr> <chr> <chr>
#> 1 1 a A aA a1 A1
#> 2 2 b B bB b2 B2
#> 3 3 c C cC c3 C3
# One can also use standard evaluation in that case using with = FALSE
dtrm[, f ~ paste0(dtrm$c, dtrm$a), with = FALSE]
#> # A data.trame: [3 × 6]
#> a b c d e f
#> <int> <chr> <fct> <chr> <chr> <chr>
#> 1 1 a A aA a1 A1
#> 2 2 b B bB b2 B2
#> 3 3 c C cC c3 C3
#
# Take care when you provide only one argument:
# If it is a formula, the data.table syntax is used (select rows)
# otherwise, the data.frame syntax applies, and columns are selected!
dtrm[1:2] # All rows and 2 first columns
#> # A data.trame: [3 × 2]
#> a b
#> <int> <chr>
#> 1 1 a
#> 2 2 b
#> 3 3 c
dtrm[~1:2] # All columns and 2 first rows!
#> # A data.trame: [2 × 6]
#> a b c d e f
#> <int> <chr> <fct> <chr> <chr> <chr>
#> 1 1 a A aA a1 A1
#> 2 2 b B bB b2 B2
# For $, on the contrary to data.frame/data.table, but like tibble,
# no partial match is allowed (returns NULL with a warning)
dtrm$count <- dtrm$c
names(dtrm)
#> [1] "a" "b" "c" "d" "e" "f" "count"
dtrm$count #OK
#> [1] A B C
#> Levels: A B C
#dtrm$co # Not OK, no partial match allowed