The body_add_flextable functions slows down each time it is called

[moreron]

Been using the excellent officer and flextable libraries for a while now, so helpful. Have recently seen what I think is a new behaviour when inserting a large(ish) number of flextables into the doc object. Basically the time required to run body_add_flextable function increases each time you call it.

This behaviour seems to be true whether calling the function in a for loop, lapply function, or manually writing each line. My current workaround is to print then read_docx after every few body_add_flextable calls otherwise the run time is in the hours not minutes. This only appears to be an issue with flextables body_add_gg and body_add_img are fine. Not sure if this is an officer or flextable package issue?

Plot from the code below to iteratively run body_add_flextable 20 times in the same doc object.

The code below gives representative version of the flextable, though the actual document has more complex and heavily formated ones (so can't use body_add_table). Is anyone else seeing this behaviour? Thanks for any pointers/assistence!

Session info
R version 4.3.0 (2023-04-21 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19045)

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

other attached packages:
[1] flextable_0.9.1 officer_0.6.2

rm(list = ls())

require(officer)
require(flextable)

Create sample data.frame

yr <- as.integer(strftime(Sys.Date(), "%Y"))
yrs <- 11
mns <- 12
cnts <- sample(0:20, size = yrs * mns, replace = T)

df <- data.frame(matrix(cnts,
nrow = yrs,
ncol = mns,
byrow = T,
dimnames = list(seq(yr - 10, yr),
head(month.abb, 12))))

df <- cbind(Year = rownames(df), data.frame(df, row.names = NULL))
df$Total <- rowSums(df[, 2:13], )
rm(cnts, mns, yr, yrs)

Create flextable

std_b <- fp_border(color="grey40")
ft <- flextable(df)
ft <- theme_zebra(ft,
odd_header = "white",
odd_body = "#7bccc4",
even_body = "white" )

ft <- border_inner_h(ft, border = std_b, part = "header")
ft <- hline(ft, i = 1, border = std_b, part = "header")
ft <- vline(ft, j = c("Year", "Dec"), border = std_b, part = "all")
ft <- width(ft, width = c(1.35, rep(1.2, 12), 1.35), unit = "cm")
rm(std_b)

Create new doc with titles, bookmarks and page breaks

titles <- paste("Flextable", 1:15)

if(exists("doc")) rm(doc)

doc <- read_docx()

for(t in titles) {

doc <- body_add_par(doc, t, style = "heading 3")
doc <- body_bookmark(doc, tolower(sub("Flextable ", "ft", t)))
doc <- body_add_break(doc)

}

rm(t)

Add flextables to doc bookmarks using for loop

tt_ft <- data.frame()
starter <- Sys.time()

for(t in titles) {

b <- Sys.time()
doc <- body_add_flextable(cursor_bookmark(doc, tolower(sub("Flextable ",
"ft", t))),
value = ft,
align = "left",
pos = "after")
s <- Sys.time() - b
tt <- data.frame(obj = t, secs = as.numeric(s))
tt_ft <- rbind(tt_ft, tt)
print(s)
}
Sys.time() - starter

Plot time taken for each iteration

plot(tt_ft$secs, xlab = "iteration: body_add_flextable", ylab = "seconds taken")

print(doc, target = "example_flextable.docx")

rm(list = ls())

[davidgohel]

we are using package xml2 and xml manipulation operate on RAM, tables in Word are big XML structures. You can use R Markdown as a substitute if you need huge reports, the behavior will be slower for small doc but faster for huge doc with tables