CODEWITHSUNDEEP
rdata.table
Pepacz
Pepacz

Reputation: 959

R - Pass column names into data.table formula - difference between get and eval

I have names of columns of data.table stored in variables. I need to refer to the columns via these variables. I made the code working (example below), but I have no clue why I need to use sometimes get() and sometime eval(). Can somebody clarify this, please?

# generate some data
foo <- rep(1:2,each = 3)
bar <- rep(c("A","B","C"),2)
baz <- rep(1:5,2)[1:6]
df <- data.frame(foo,bar,baz)
setDT(df)

# refer to columns directly by their names
df[, "qux":=baz-baz[bar=="C"], by=foo]

# save column names into variables and call columns via these variables
var1 <- "foo"
var2 <- "bar"
var3 <- "baz"
varNew <- "qux2"

df[, eval(varNew) := get(var3) - get(var3)[get(var2) == "C"], by = get(var1)]

df
   foo bar baz qux qux2
1:   1   A   1  -2   -2
2:   1   B   2  -1   -1
3:   1   C   3   0    0
4:   2   A   4   3    3
5:   2   B   5   4    4
6:   2   C   1   0    0

Upvotes: 4

Views: 751

Answers (1)

steveb
steveb

Reputation: 5532

This example shows the difference between how eval and get differ in function. Using a data.table object is not needed to show what each does.

iVec     <- c(123, 456)
iVarName <- "iVec"

# Returns the contents of 'iVarName' (a string).  This happens
# to be the name of a variable but doesn't have to.
eval(iVarName)
##> [1] "iVec"

# Returns the contents of what 'iVarName' refers to (it
# refers to the variable "iVec" in this case, which
# is a variable which contains a vector of integers).
get(iVarName)
##> [1] 123 456

### #########################################
### Similar to above but where the variable
### 'iVec2' does not exist.
### #########################################
rm(iVec2)
# The variable "iVec2" does not exist.
iVarName2 <- 'iVec2'

# Returns the contents of 'iVarName2' (a string).  This is not
# the name of an existing variable in this context.
eval(iVarName2)
## [1] "iVec2"
get(iVarName2)  # Returns an error because 'iVec2' doesn't exist.
## Error in get(iVarName2) : object 'iVec2' not found

Since this question is more about eval vs. get, I will leave the data.table specifics out. The way data.table handles strings and variable names is very likely answered in a different SO post.

Upvotes: 5

Related Questions