inst/benchmarks/benchmark-identical.R

In robertzk/objectdiff: Objectdiff

# What is the fastest way to tell if two objects are identical?

library(microbenchmark)

# Note: base::identical uses memcmp under the hood for atomic comparisons,
# which is *very* fast!
# https://github.com/SurajGupta/r-source/blob/46102b91b35a7befa0cf6cc6abd0da09b86f1621/src/main/identical.c#L160

x <- sample(seq_len(1000000)); x2 <- x
y <- runif(1000000, 0, 1); y2 <- y

print(microbenchmark(
  "base::identical on numerics" = identical(y, y2),
  "base::identical on integers" = identical(x, x2)
))

# Note the max below is artificial and a result of garbage collection.
#
# Unit: nanoseconds
#                         expr min    lq median   uq   max neval
#  base::identical on integers 804 905.5  958.0 1062 12934   100
#  base::identical on numerics 812 906.0  964.5 1058  2669   100
#

# Vectorized identicality
x <- as.list(1:20000)
y <- as.list(1:20000)
y[[1000]] <- 1
print(microbenchmark(
  "Identical check using Map"    = Map(identical, x, y),
  "Identical check using mapply" = mapply(identical, x, y)
))

# Unit: milliseconds
#                          expr      min       lq   median       uq       max neval
#     Identical check using Map 37.20443 42.76116 47.18948 51.38830 116.86836   100
#  Identical check using mapply 43.61590 50.57991 56.57245 60.28356  86.80098   100