Write list comprehensions in R!
The eList package allows users to write vectorized for
loops and contains a variety of tools for working with lists and other
vectors. Just wrap a normal for
loop within one of the
comprehension functions, such as List()
, and let the
package do the rest.
Features include, but are not limited to:
"."
to
separate names.if
, else
statements to filter
results.=
., clust =
and the
cluster.enum
and items
to
access the index or name within the loop, or zip
objects
together.for
loops.List(...)
or List[...]
for the
comprehension..
notation for
variables) or calls.You can install the released version of eList from CRAN with:
install.packages("eList")
A simple “list” comprehension that accumulates all integer sequences
to 4 using the List
function. Though it looks like a
for
loop, it is actually using lapply
behind
the scenes.
library(eList)
#>
#> Attaching package: 'eList'
#> The following object is masked from 'package:stats':
#>
#> filter
#> The following object is masked from 'package:utils':
#>
#> zip
List(for (i in 1:4) 1:i)
#> [[1]]
#> [1] 1
#>
#> [[2]]
#> [1] 1 2
#>
#> [[3]]
#> [1] 1 2 3
#>
#> [[4]]
#> [1] 1 2 3 4
Loops can be nested and filtered using if
statements.
The example below uses Num
to produce a numeric vector
rather than a list. Other comprehensions include Chr
for
character vectors, Logical
for logical vectors,
Vec
for flat (non-list) vectors, etc.
Num(for (i in 1:4) for (j in 3:6) if (i == j) i^2)
#> [1] 9 16
Use the “dot” notation to use multiple variables within the loop.
Chr(for (i.j in zip(1:4, 5:8)) paste0(i, j))
#> [1] "15" "26" "37" "48"
Use =
within the loop to assign a name to each item
within the list, or other item.
<- zip(letters[1:4], 5:8)
values
List(for (i.j in values) i = j)
#> $a
#> [1] "5"
#>
#> $b
#> [1] "6"
#>
#> $c
#> [1] "7"
#>
#> $d
#> [1] "8"
Parallelization is also very easy. Just create a cluster and add it
to the comprehension with the clust
argument.
<- auto_cluster()
my_cluster
<- Num(for (i in sample(1:100, 50)) sqrt(i), clust = my_cluster)
x
# Close the cluster if not needed!
close_cluster(my_cluster)
x#> [1] 3.872983 9.949874 3.316625 6.633250 7.000000 9.165151 9.899495 9.539392
#> [9] 9.055385 3.464102 5.830952 9.486833 3.741657 8.660254 4.898979 8.000000
#> [17] 4.690416 5.477226 5.385165 5.000000 8.485281 8.246211 4.358899 7.416198
#> [25] 8.717798 9.591663 9.433981 7.549834 3.162278 9.797959 7.615773 3.000000
#> [33] 9.746794 2.236068 5.291503 8.124038 9.219544 9.695360 7.483315 9.643651
#> [41] 3.605551 4.000000 1.732051 5.196152 2.828427 6.855655 7.681146 7.348469
#> [49] 7.211103 7.280110
Want a statistical summary using a comprehension? eList contains a
variety of summary functions for that purpose. Stats
is a
general summary comprehension that computes many different values.
Stats(for (i in sample(1:100, 50)) sqrt(i))
#> $min
#> [1] 2
#>
#> $q1
#> [1] 6.143211
#>
#> $med
#> [1] 7.483016
#>
#> $q3
#> [1] 8.587745
#>
#> $max
#> [1] 10
#>
#> $mean
#> [1] 7.172598
#>
#> $sd
#> [1] 2.028831
eList also contains functional programming style functions for working with lists and other vectors. These functions perform an operation using a function on another object. They are similar to the higher order functions in Base R, but are pipe-friendly, handle a wide ranger of object types, and allow for different methods of specifying functions.
<- list(1:4, 5:8, 9:12)
x map(x, mean)
#> [[1]]
#> [1] 2.5
#>
#> [[2]]
#> [1] 6.5
#>
#> [[3]]
#> [1] 10.5
This can also be calculated using formula notation. Formulas can be be written as either a two-sided formula or a one-sided formula by prefixing variables with dots.
# Two-sided Formula
map(x, i ~ sqrt(i) + 1)
#> [[1]]
#> [1] 2.000000 2.414214 2.732051 3.000000
#>
#> [[2]]
#> [1] 3.236068 3.449490 3.645751 3.828427
#>
#> [[3]]
#> [1] 4.000000 4.162278 4.316625 4.464102
# One-sided Formula
map(x, ~ sqrt(.i) + 1)
#> [[1]]
#> [1] 2.000000 2.414214 2.732051 3.000000
#>
#> [[2]]
#> [1] 3.236068 3.449490 3.645751 3.828427
#>
#> [[3]]
#> [1] 4.000000 4.162278 4.316625 4.464102
The higher order functions also accept unevaluated “calls”.
<- substitute(round(digits=2))
round2 map(rnorm(5), round2)
#> [[1]]
#> [1] 0.16
#>
#> [[2]]
#> [1] -3.18
#>
#> [[3]]
#> [1] -0.17
#>
#> [[4]]
#> [1] -0.13
#>
#> [[5]]
#> [1] -0.27