.escape_xml() performance tweak

[TimTaylor]

If stringi::stri_replace_all_fixed() cycles through patterns and replacements in the order given then you could change

x_utf8 <- stringi::stri_enc_toutf8(x)
x_no_amp <- stringi::stri_replace_all_fixed(str = x_utf8, pattern = c("&"), replacement = c("&amp;"), vectorize_all = FALSE)
stringi::stri_replace_all_fixed(str = x_no_amp, pattern = c("\"", "<", ">", "'"), replacement = c("&quot;", "&lt;", "&gt;", "&apos;"), vectorize_all = FALSE)

to

x_utf8 <- stringi::stri_enc_toutf8(x)
stringi::stri_replace_all_fixed(str = x_utf8, pattern = c("&","\"", "<", ">", "'"), replacement = c("&amp;", "&quot;", "&lt;", "&gt;", "&apos;"), vectorize_all = FALSE)

and eek out a little more performance (For me 10% on top of change you propose in #126). Looking at the stringi documentation and code this seems to be the case.

Hope this makes sense.

[chainsawriot]

Thanks for the suggestion, @TimTaylor . It seems that it can really prevent some wasteful steps.

.escape_xml <- function(x) {
    stringi::stri_replace_all_fixed(str = stringi::stri_enc_toutf8(x), pattern = c("&", "\"", "<", ">", "'"), replacement = c("&amp;", "&quot;", "&lt;", "&gt;", "&apos;"), vectorize_all = FALSE)
}

Without #126

system.time(write_ods(nycflights13::flights))
##   user  system elapsed 
## 162.434   3.269 165.810 

Again versus current HEAD

system.time(write_ods(nycflights13::flights))
##   user  system elapsed 
## 203.636   4.029 207.797 

[chainsawriot]

added 3833747 thank you!