function.canonicalize: Canonicalize a Function

In thomasWeise/functionComposeR: R Package for Composing and Canonicalizing Functions

Description

function.canonicalize(f) tries to generate a canonical and compact version of the function f. It therefore tries to substitute all variables in the body of f with their value if the value is known and then tries to evaluate all sub-expressions whose results will be constant. In other words, the promises, variables, and expressions inside a function f are evaluated and replaced with constant values where possible.

We make use of the functions unenclose, modify_lang, and substitute_q of the package pryr, the function substituteDirect of the package methods and also of our recursive version of the ideas given by by schloerke (http://github.com/schloerke) at http://github.com/hadley/pryr/issues/43. We also apply expression.simplify to take care of the partial resolution of the function body and for replacing identical sub-expressions with the same object (e.g., to make the execution more cache friendly).

These measures are intented to produce a version of a function f which is both human-readable and as fast to evaluate as possible. The new function will reside in the same environment as f.

Usage

function.canonicalize(f)

Arguments

f	the function

Value

the canonicalized function

Examples

f <- function(x) { 5+3+x }
function.canonicalize(f)
# function (x)
# 8 + x
z <- 24;
g <- function(x) { tan(sin(z) + (z*27) / x) }
function.canonicalize(g)
# function (x)
# tan(-0.905578362006624 + 648/x)
a <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25)
b <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25)
c <- c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25)
f <- function(x) (c-(x*a)/b)*(a+b+c)
g <- function.canonicalize(f)
g
# function (x)
# (c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17,
# 18, 19, 20, 21, 22, 23, 24, 25) - (x * c(1, 2, 3, 4, 5, 6, 7,
# 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23,
# 24, 25))/c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
# 16, 17, 18, 19, 20, 21, 22, 23, 24, 25)) * c(3, 6, 9, 12, 15,
# 18, 21, 24, 27, 30, 33, 36, 39, 42, 45, 48, 51, 54, 57, 60, 63,
# 66, 69, 72, 75)
library(microbenchmark)
x <- runif(1000)
microbenchmark(f(x), g(x))
# Unit: microseconds
# expr   min    lq     mean median     uq    max neval
# f(x) 8.868 9.097 10.01950 9.2935 9.4805 35.340   100
# g(x) 8.285 8.506  9.38347 8.6730 8.7650 28.501   100

thomasWeise/functionComposeR documentation built on May 28, 2019, 4:03 p.m.