simple DelayedMatrix benchmark showing access time of n rows growing as O(n^3)

methbenchmark.R

if( Biobase::package.version("curatedTCGAData") < "1.5.6" ){
  BiocManager::install("waldronlab/curatedTCGAData")
}
stopifnot(BiocManager::version() >= "3.9")

library(curatedTCGAData) #requires >=1.5.6 and bioc-devel
mae <- curatedTCGAData("UCEC", "Methylation_methyl27", dry.run = FALSE) #~2 seconds from cache
dm <- assay(mae, 1)

# benchmarking showing cubic increase with # rows
n <- c(100, 200, 400, 600, 800, 1000, 1250, 1500, 1750, seq(2000, 4000, 500), 5000, 6000, 8000, 10000)
res1 <- sapply(n, function(i) system.time(as.matrix(dm[seq(1, nrow(dm), length.out=i), ]))) #i rows sequentially from 1:all
res2 <- sapply(n, function(i) system.time(as.matrix(dm[sample(1:nrow(dm), i), ]))) #i rows randomly from all
res3 <- sapply(n, function(i) system.time(as.matrix(dm[1:i, ]))) #first i rows

plot(n, res1["elapsed", ], xlab="# of rows selected", ylab="elapsed time for as.matrix", 
     type="l", ylim=c(0, 15))
lines(n, res2["elapsed", ], lty=2, lwd=2)
lines(n, res3["elapsed", ], lty=3, lwd=2)
legend("topleft", lty=1:3, lwd=c(1, 2, 2),
       legend=c("sequential sample", "random sample", "first n rows"))

## try to infer polynomial degree
getr2 <- function(deg, res, n){
  summary(lm(res["elapsed", ] ~ I(n^deg)))$r.squared
}

r2 <- data.frame(sequential=sapply(1:4, getr2, res1, n),
           random=sapply(1:4, getr2, res2, n),
           firstn=sapply(1:4, getr2, res3, n))
r2

Previous version had some ACC crud I forgot to remove

And changed to UCEC because it has only 118 columns (27578 rows)