Using := multiple times in data.table

Question

I frequently find myself doing a long series of chained calculations using := on the same data table. For example, something like this

test = data.table(1:10, 1:10, 1:10, 1:10)

test[, V1 := V1^2]
test[, V2 := V1*V2]
test[, V3 := V2/V3]
test[, V4 := sqrt(V3)]
test[, new := letters[V4]]

Having to write test[, ...] on every line 1) takes longer to type (not a big issue I can cope with that). But, more importantly, also distracts visually from the flow and content of the calculation. I would much rather write something like

test[, {
  V1 := V1^2
  V2 := V1*V2
  V3 := V2/V3
  V4 := sqrt(V3)
  new := letters[V4]
}]

But this throws an error of You have wrapped := with {} which is ok but then := must be the only thing inside {}.

I know that I can write

within(test, {
  V1 = V1^2
  V2 = V1*V2
  V3 = V2/V3
  V4 = sqrt(V3)
  new = letters[V4]
  })

But this loses the efficiency of using :=

I tried writing a function to provide this ability

with.dt = function(dt, expressions){
  e = strsplit(expressions,'\n')
  for (i in 1:length(e)){    
    dt[, eval(parse(text = e[i]))]
  }
  dt
  }

with.dt(test, "
  V1 := V1^2;
  V2 := V1*V2;
  V3 := V2/V3;
  V4 := sqrt(V3);
  new := letters[V4];
  ")

But this does not change the content of the data.table

Is there any syntax that permits the neat look of the within version, while retaining the assign by reference property of :=?

If you don't need them done sequentially you can do something like this: test[,:=(v1 = v1^2, v2 = v1*v2)] There are supposed to be backticks around the := — Kristofersen
– Kristofersen, Commented Dec 7, 2016 at 17:33
Extending what Kristoferson said, I think you can do it sequentially with a hack like DT[, `:=`(V1 = V1 <- 10, V2 = V2 <- V1^2, ...)] — Frank
– Frank, Commented Dec 7, 2016 at 17:37
@Frank Thanks for the suggestion. I updated the q with another attempt using a function definition for with.dt. Any idea why this does not work? — dww
– dww, Commented Dec 7, 2016 at 17:57
@clemlaflemme One of the features of data.table and := in particular is that the object is modified by reference so you don't need to reassign the result. — Frank
– Frank, Commented Dec 7, 2016 at 18:11
@Kristofersen it took me forever to realize this but you can escape backticks in comments within an inline code block by backslash: \` will leave it as is: `:=` — MichaelChirico
– MichaelChirico, Commented Dec 7, 2016 at 19:25

Kristofersen · Accepted Answer · 2016-12-07 18:57:06Z

2

There are a couple different ways to do this.

Original test matrix used here:

   v1 v2
1:  1  3
2:  2  4

First, we can do something like this:

test[,`:=`(v1 = v1^2,
          v2 = v1*v2)

Output:

v1 v2
 1  3
 4  8

Or, if we want it done sequentially we can use Frank's hack.

test[, `:=`(v1 = v1 <- v1^2, v2 = v2 * v1)]

Output:

v1 v2
 1  3
 4 16

Or, similarly we could run this:

test[,c("v1","v2") := list(v1^2,v1^2*v2)]

Output:

v1 v2
 1  3
 4 16

edited Dec 7, 2016 at 18:57

answered Dec 7, 2016 at 17:56

Kristofersen

2,8261 gold badge18 silver badges33 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

eddi Over a year ago

or test[, v1 := v1^2][, v2 := v2*v1]

dww · Accepted Answer · 2016-12-07 21:31:40Z

1

We can use a function that takes a list of expressions:

with.dt = function(dt, expr){
  for (j in 1:length(expr)) set(dt, , names(expr)[j], dt[, eval(expr[[j]])])
}

test = data.table(1:10, 1:10, 1:10, 1:10)
with.dt(test, expression(
  V1 = V1^2,
  V2 = V1*V2,
  V3 = V2/V3,
  V4 = sqrt(V3),
  new = letters[V4]
))

#     V1   V2  V3 V4 new
# 1:   1    1   1  1   a
# 2:   4    8   4  2   b
# 3:   9   27   9  3   c
# 4:  16   64  16  4   d
# 5:  25  125  25  5   e
# 6:  36  216  36  6   f
# 7:  49  343  49  7   g
# 8:  64  512  64  8   h
# 9:  81  729  81  9   i
#10: 100 1000 100 10   j

edited Dec 7, 2016 at 21:31

answered Dec 7, 2016 at 19:15

dww

31.6k8 gold badges75 silver badges126 bronze badges

Collectives™ on Stack Overflow

Using := multiple times in data.table

2 Answers 2

1 Comment

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related