as.vector performance

[ethanbsmith]

Description

i noticed some odd timings on as.vector() on an xts object

Expected behavior

same performance as as.vector on a matrix

base r will dispatch to this

as.vector.xts <- function(x, mode = "any") {
  return(as.vector(coredata(x), mode = mode))
}

Minimal, reproducible example

z <- getSymbols("SPY", auto.assign = F)
microbenchmark::microbenchmark(as.vector(z), as.vector(coredata(z)), as.numeric(z), as.numeric(coredata(z)))
#Unit: microseconds
#                    expr   min    lq    mean median     uq   max neval
#            as.vector(z) 353.8 359.9 377.669 363.95 373.35 645.4   100
#  as.vector(coredata(z))   9.9  11.1  12.300  11.70  13.55  21.0   100
#           as.numeric(z)   7.9   8.7  10.574   9.15  11.65  77.1   100
# as.numeric(coredata(z))   7.1   7.7   8.670   8.00   9.10  15.3   100
identical(as.vector(z[,4]), as.vector(coredata(z[,4])))
#[1] TRUE

Session Info

R version 4.3.1 (2023-06-16 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 11 x64 (build 22621)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.utf8  LC_CTYPE=English_United States.utf8    LC_MONETARY=English_United States.utf8 LC_NUMERIC=C                           LC_TIME=English_United States.utf8    

time zone: America/Denver
tzcode source: internal

attached base packages:
[1] parallel  stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] kableExtra_1.3.4  jsonlite_1.8.7    readxl_1.4.3      xml2_1.3.5        curl_5.0.2        Rcpp_1.0.11       matrixStats_1.0.0 data.table_1.14.8 doFuture_1.0.0    future_1.33.0     doParallel_1.0.17 iterators_1.0.14 
[13] foreach_1.5.2     quantmod_0.4.25   TTR_0.24.3.1      xts_0.13.1.2      zoo_1.8-12        plotrix_3.8-2   

[joshuaulrich]

This is due to the as.matrix() call in as.vector.zoo():

as.vector.zoo <- function(x, mode = "any") {
    as.vector(as.matrix(x), mode = mode)
}

This function hasn't been altered since it was added in r18 "first version that passes R CMD check" on 2004-10-08. This was before coredata() was introduced.

@zeileis what do you think about changing the as.matrix() call to coredata() in the above function? Then it would be:

as.vector.zoo <- function(x, mode = "any") {
    as.vector(coredata(x), mode = mode)
}

After looking at as.matrix.zoo(), I don't think this would be a breaking change. The first line of that function is y <- as.matrix(coredata(x), ...). The rest of as.matrix.zoo() handles setting dims, colnames, and rownames on the result. But we're just going to call as.vector() on the result, which will drop those attributes anyway. Thoughts?

[zeileis]

Good idea! Thanks for pointing this out Ethan @ethanbsmith and for doing the additional digging Josh @joshuaulrich !

I agree and have just committed the fix to the R-Forge version of zoo, revision 1192.

[joshuaulrich]

Fixed upstream in zoo.