strings dependency

[elong0527]

In the convert function, I need to replace strings based on a mapping rule.

Currently, I rely on stringi in the convert function. Is there a efficient way in using base R?

r2rtf/R/conversion.R

Line 107 in c55513d

text[index] <- stringi::stri_replace_all_fixed(text[index], names(char_latex), char_latex,

Here is an example mappings I would need to transfer from left side to the right side in a vector of strings.

  char_rtf <- c(
    "^" = "\\super ",
    "_" = "\\sub ",
    ">=" = "\\geq ",
    "<=" = "\\leq ",
  )

cc: @nanxstats @yihui

[yihui]

I think what you are doing currently is quite reasonable:

r2rtf/R/conversion.R

Lines 106 to 114 in c55513d

	if (load_stringi) {
	text[index] <- stringi::stri_replace_all_fixed(text[index], names(char_latex), char_latex,
	vectorize_all = FALSE, opts_fixed = list(case_insensitive = FALSE)
	)
	} else {
	for (i in 1:length(char_latex)) {
	text[index] <- gsub(names(char_latex[i]), char_latex[i], text[index], fixed = TRUE)
	}
	}

I wouldn't worry too much about the performance of gsub() (with a for-loop). You can do some benchmarking for convert(load_stringi = TRUE) vs convert(load_stringi = FALSE) to have a clearer idea. I guess the latter is very likely to be slower, but if practically the difference is 10ms (gsub()) vs 1ms (stringi), I won't bother thinking about it at all and will just use gsub(). Of course, the time depends on the size of the character vector text. I guess the time difference won't be noticeable if length(text) is relatively small (e.g., less than 1000).

[elong0527]

Thanks for the suggestion, I am closing the issue with your blessing.

I added it because of a use case in clinical trial that requires to save all safety data as a listing in RTF format for EU (called ICH listing).

For a large trial, it can goes to > 100k records with > 10k pages in a RTF files.

[yihui]

For 100k records, the time difference may be noticeable but I guess gsub() should take no longer than one second, which should be fine. Again, benchmarking will give you a clearer idea.

BTW, for gsub(), using perl = TRUE (with default fixed = FALSE, since fixed = TRUE is incompatible with perl = TRUE) may give you some substantial speedup.