Allow args to opts_regex etc. to be passed with "..."

[gagolews]

Currently calling stri_opts_collator, stri_opts_regex, and stri_opts_brkiter e.g. in stringi-search functions is a "little bit" inconvenient. What about adding a "..." param to all the functions so that all the matching args are passed to stri_opts?

Something like:

> test <- function (str, pattern, simplify=FALSE, opts_regex = NULL, ...) 
+ {
+    if (!missing(...))
+       opts_regex <- do.call(stri_opts_regex, as.list(c(opts_regex, ...)))
+    .Call("stri_extract_all_regex", str, pattern, simplify, opts_regex, PACKAGE = "stringi")
+ }
> 
> print(test("ala", "ALA"))
[[1]]
[1] NA

> print(test("ala", "ALA", case_insensitive=TRUE))
[[1]]
[1] "ala"

> print(test("ala", "ALA", case=TRUE))
[[1]]
[1] "ala"

> print(test("ala", "ALA", case_insensitive=TRUE, nosucharg=FALSE))
 Error in (function (case_insensitive, comments, dotall, literal, multiline,  : 
  unused argument (nosucharg = FALSE) 

> print(test("ala", "ALA", opts_regex=stri_opts_regex(case_insensitive=FALSE), case_insensitive=TRUE))

 Error in (function (case_insensitive, comments, dotall, literal, multiline,  : 
  formal argument "case_insensitive" matched by multiple actual arguments 

Benchmarks:

> microbenchmark::microbenchmark(
+    stri_extract_all_regex("ala", "ala"),
+    stri_extract_all_regex("ala", "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
+    stri_extract_all_regex("ala", "ala", opts_regex=list(case_insensitive=FALSE)),
+    test("ala", "ala"),
+    test("ala", "ala", case_insensitive=FALSE),
+    test("ala", "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
+    test("ala", "ala", opts_regex=list(case_insensitive=FALSE))
+ )
Unit: microseconds
                                                                                         expr    min      lq     mean  median      uq     max neval
                                                         stri_extract_all_regex("ala", "ala") 15.649 16.5875 19.09487 17.0950 18.0355  97.610   100
 stri_extract_all_regex("ala", "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 24.647 25.7270 27.77733 26.3455 28.0155  54.031   100
            stri_extract_all_regex("ala", "ala", opts_regex = list(case_insensitive = FALSE)) 16.800 17.9585 18.81852 18.5280 19.4885  27.298   100
                                                                           test("ala", "ala") 15.643 16.7275 17.91048 17.5255 18.2935  38.547   100
                                                 test("ala", "ala", case_insensitive = FALSE) 56.032 57.9775 61.66630 58.9165 60.0255 103.298   100
                   test("ala", "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 25.667 26.9000 28.57848 27.3160 28.1750  56.455   100
                              test("ala", "ala", opts_regex = list(case_insensitive = FALSE)) 16.599 17.9690 19.41654 18.8200 19.5445  38.198   100

> x <- stri_rand_strings(10000, 3)
> microbenchmark::microbenchmark(
+    stri_extract_all_regex(x, "ala"),
+    stri_extract_all_regex(x, "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
+    stri_extract_all_regex(x, "ala", opts_regex=list(case_insensitive=FALSE)),
+    test(x, "ala"),
+    test(x, "ala", case_insensitive=FALSE),
+    test(x, "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
+    test(x, "ala", opts_regex=list(case_insensitive=FALSE)),
+    unit="relative"
+ )
Unit: relative
                                                                                     expr       min        lq      mean   median        uq       max neval
                                                         stri_extract_all_regex(x, "ala") 0.9981751 0.9961943 0.9624526 1.001169 0.9908091 0.9326833   100
 stri_extract_all_regex(x, "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 1.0000000 1.0000000 1.0000000 1.000000 1.0000000 1.0000000   100
            stri_extract_all_regex(x, "ala", opts_regex = list(case_insensitive = FALSE)) 0.9976754 1.0013583 0.9913800 1.003427 1.0021834 0.9551786   100
                                                                           test(x, "ala") 0.9982575 1.0048992 0.9965834 1.010263 1.0219637 0.9833843   100
                                                 test(x, "ala", case_insensitive = FALSE) 1.0232501 1.0275143 1.0000415 1.033012 1.0288242 1.0404019   100
                   test(x, "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 1.0049166 1.0225035 1.0088873 1.024782 1.0277768 1.0289821   100
                              test(x, "ala", opts_regex = list(case_insensitive = FALSE)) 1.0047261 1.0019214 1.0068571 1.005476 1.0114852 2.1843647   100

Do we accept such a performance "drop"? That'd be really user-friendly, IMO. :)

[gagolews]

However, I'm unsure if missing(...) is valid at all, see this SO question

[gagolews]

An alternative solution posted by GSee@SO:

library("stringi")
test1 <- compiler::cmpfun(function(str, pattern, simplify=FALSE, opts_regex = NULL, ...) 
{
    if (!missing(...))
       opts_regex <- do.call(stri_opts_regex, as.list(c(opts_regex, ...)))
    .Call("stri_extract_all_regex", str, pattern, simplify, opts_regex, PACKAGE = "stringi")
})

test2 <- compiler::cmpfun(function(str, pattern, simplify=FALSE, opts_regex = NULL, ...) 
{
   .dotargs <- list(...)
   if (length(.dotargs))
       opts_regex <- do.call(stri_opts_regex, as.list(c(opts_regex, .dotargs)))
    .Call("stri_extract_all_regex", str, pattern, simplify, opts_regex, PACKAGE = "stringi")
})

microbenchmark::microbenchmark(
    stri_extract_all_regex("ala", "ala"),
    test1("ala", "ala"),
    test2("ala", "ala"),
    stri_extract_all_regex("ala", "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
    test1("ala", "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
    test2("ala", "ala", opts_regex=stri_opts_regex(case_insensitive=FALSE)),
    stri_extract_all_regex("ala", "ala", opts_regex=list(case_insensitive=FALSE)),
    test1("ala", "ala", opts_regex=list(case_insensitive=FALSE)),
    test2("ala", "ala", opts_regex=list(case_insensitive=FALSE)),    
    test1("ala", "ala", case_insensitive=FALSE),
    test2("ala", "ala", case_insensitive=FALSE),
   unit="relative"
 )

give

Unit: relative
                                                                                         expr      min       lq     mean   median       uq       max neval
                                                         stri_extract_all_regex("ala", "ala") 1.000000 1.000000 1.000000 1.000000 1.000000 1.0000000   100
                                                                          test1("ala", "ala") 1.075398 1.052826 1.017465 1.053447 1.029103 0.8134295   100
                                                                          test2("ala", "ala") 1.103345 1.084199 1.148588 1.075272 1.161981 3.0893550   100
 stri_extract_all_regex("ala", "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 1.650189 1.578858 1.436439 1.565039 1.501811 1.1126940   100
                  test1("ala", "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 1.685809 1.609878 1.525307 1.585138 1.563039 1.3008204   100
                  test2("ala", "ala", opts_regex = stri_opts_regex(case_insensitive = FALSE)) 1.731662 1.652550 1.611348 1.645803 1.775540 1.7175679   100
            stri_extract_all_regex("ala", "ala", opts_regex = list(case_insensitive = FALSE)) 1.117478 1.091565 1.057609 1.093022 1.062635 1.2145773   100
                             test1("ala", "ala", opts_regex = list(case_insensitive = FALSE)) 1.204451 1.166843 1.147524 1.171154 1.259060 1.1053589   100
                             test2("ala", "ala", opts_regex = list(case_insensitive = FALSE)) 1.208672 1.186623 1.117606 1.179420 1.158620 1.1505213   100
                                                test1("ala", "ala", case_insensitive = FALSE) 3.606190 3.386835 3.133516 3.336023 3.373023 3.4719326   100
                                                test2("ala", "ala", case_insensitive = FALSE) 3.400652 3.238217 3.011531 3.175456 3.116777 2.8689177   100

[gsee]

Try it where you pass in something without defining it first such that you get an error like in the SO post.

[gsee]

Instead of .dotargs <- list(...);if (length(.dotargs) > 0), try using if (length(list(...))) like I did in the SO post ;)

[gagolews]

@gsee: I've updated the benchmarks; in the test1 case I would get an err on undefined something anyway, as I do c(opts_regex, ...). I think I'll use length(list(...) for this feature, thanks!

[gsee]

agreed, you'd get an error with either test1 or test2 if you call with something, but if you were doing a lot of processing between when you check the arguments and when you attempt to use them, you might prefer to get the error sooner to avoid unnecessary computation.