tests/tests.r

In stringfish: Alt String Implementation

suppressMessages(library(stringfish, quietly = T))
suppressMessages(library(qs, quietly = T))
suppressMessages(library(dplyr, quietly = T))
suppressMessages(library(Rcpp, quietly = T))
suppressMessages(library(stringr, quietly = T))
suppressMessages(library(rlang, quietly = T))

encode_source <- function(file, width = 160) {
  n <- file.info(file)$size
  x <- readChar(con = file, nchars=n, useBytes = T)
  x <- qserialize(x, preset = "custom", algorithm = "zstd", compress_level = 22)
  x <- base91_encode(x)
  starts <- seq(1,nchar(x), by=width)
  x <- sapply(starts, function(i) {
    substr(x, i, i+width-1)
  })
  x <- gsub('\\"', "\'", x)
  dput(x)
}

decode_source <- function(x) {
  x <- paste0(x, collapse = "")
  x <- gsub("\\'", '\\"', x)
  x <- base91_decode(x)
  qdeserialize(x)
}

altrep_support <- function() {
  getRversion() >= "3.5"
}

is_solaris <- function() {
  grepl('SunOS',Sys.info()['sysname'])
}

myfile <- tempfile()
print(myfile)

# i500_utf8 <- readRDS("/tmp/temp.rds")
# i500_latin1 <- iconv(i500_utf8, "UTF-8", "latin1")

# For a test set, we are using the 500 most common Icelandic words
# This is a pretty good test set because all Icelandic words can be encoded as either UTF-8 or latin1. It also contains a mix of ASCII and non-ASCII strings
# https://www.101languages.net/french/most-common-french-words/
# https://en.wikipedia.org/wiki/ISO/IEC_8859-1
# qserialize(x[1:500], preset = "custom", compress_level = 22, check_hash=F) %>% base91_encode %>% catquo
i500_utf8 <- decode_source('un]"BAAA@QBtHACAAAAAAAY4TBAABdk1kure+B3AwoFYdc"lM?hVjfxq5F^5OXd&}iic*u6aS*TsW0{Ur6@MdueBh+g1us}b/1]FhtGOG"/1xj=x/iL}2,4f#H8v+r_qDxuL3U2jD!9.@z%wUvNyLRYr4SugnxE`=hw2L+NI<X8w<q*MV&?(?V:FE!?+CimfUy<&u/Bk7xer7/u;syo9Cu:qx7J57?*_/6I@)xs[J;fXOK0B_l"[4EKyF#:{7z_>g.MhpqTM[&1_P.E@i&,j(3{?d|qu!g,fl:gUcJ;U#KwdN,_?zyzU~K@dwS}Evq&qWYNl(?xqk$j/L#s#{3^H)BXxo`t["IA+j+JLI*yEkS0<p4a3PV6<~N!f3v7HyE$j@/I@W0d[sy)p&z)?(k>qJoTLq#eE)}t8G:F`@bQJHYvHND{eS"@QH)m6Rgv,|a)@nV=!*%=<jElUsK5k}4,axCwZ~NB(+e|XA1OKDJ^@koTm4YAx#I8oCQOq}fyfqL>?loMXRva2k*4.V&`n?1j~MzQ1J[9t?IH1jV})PLT^,GA0J#8|VB=n9=cy91B.#J$m{SgJ=mk(5x*ml:j%u.(+_Nk8#/@w]C~PyEVk+tlU.iicD2hk^dYc2am&6_@ges.x$lbUZcTidf.IoKPnSUJp!gab|bS61!4U_u&XTPnEo%eUP&0J.OisuVyZ:|}b|Nl1[pZ:.7wOOS+*I`ojDvw6E#rq"){8i7Pt{.auL+`GVyZ&w!NKtp3k09J+K?;<:}#O/9+N#5DYF<{|cy7P_*9GzY(f4<qR$a&EaGeQ.fIo7_o#X?"_M$d[S]3;vJzM8Eg!;x(]~Nzn<d9`rCeJwn"5zd(npX_UA)yk=F!ue51ac;#>j6wnSlb,/9|fc|Um|^rb7+nr$w0P^x^nKZvO{Sg%@8{?VwNQgF<+cb@)Tf$3Zu(Oou,fN=EpfnqLI{s3$sp62;L!~x_*]5}`cIwJKDXWgY,/09Ft*;xn2,0KDrz/F.S[:)YJ:vsn{?EgMh`)#ss)1OiU9uAdZ6|bH+m6+x]?y;pq7ralfX;>Y?_h<YRdh7we_WP0{8|ft_oAbTodGh.cT9j&~5EgOVX.njZtoMu>f?0f~>9W0Y?K4+t#nNS(qOKox0"x>Tv|WP4vw|~XTlA]=SNJXdDpQdJbv*PL&8yGGEI%2$kqO1S)1i6?BmI,fTJ:=it_P,2.Mb/DeHFKMrTyB&oUjmpJ/SM^/A<{>g1HTVevQ7MeXcr)st{>AD;OQ)5Vwst^1xwdMa?g>?|yFSh9za`XtRat<xL@Pe7boGp6TaNQbu:bPFO7^S}CuzeR~YN5/_r04C:DOQ;KV?"xbkeU%OE814d~wg;FngvPgD$MCo"Oy|V)?6UXQJ!jPK0VqLjoCo9o[nxCOw"S3?;x$(24kGK,q+IkGAgAuW3L{C+WvL;^7tTH9C|2ID"V*3"az8.U&4{RW)2E5(NcJ#q=`[a7XzsUi;FJMv;G]Z@(mj4"_^5)il{eTT"/`r8)QuwW]Wlzm.XMI|$W~$OZ<L6>hV*2M.F(!o/e?&CEF>PK)z:prsTfugRCp(=%K!%hu.BL;*5tkY60XaMxD7rTc^rGIM#0,c*OD$^?,*2vlB%p11+;_gq]CPyXF4P:F#|WW}6nH,_z~l7bz[1n:m|]Nh[>k)A6Y).!U>w"AC}z1EB@AQ)qsB)oX2+N<a&e+@uCjenbJDZ~kqY7J)i7*+y/4B:pR/MbS{|Pgoi<4yzrymz2u(Hqg(`GNdLS{674]i^NIIHz$MVzw/&iJbn^4r0$)4G_<"A:yq<41d^+l/D^11}vs#LY|)1ghwz.6@&qE0!E[GCK[!"HWZxw1{N%IM2[W7X.A&L)u=E;GjSKo;l&}7W/*g<;E,tv[ZFzXCzd47&afxl]96H/Z?[i&YvmbxiMJeCG^Aaa|?j%`2*,@i#1#W!xfvG@/h+Fi{oGa;|c51I8C8wd:HDBosoF)n$IHioJa`RX.b{/KMM{y~N>}w]SjR<!KXER}h`e]WqHRV23+V;}SIx=N<#*xddZ;%wo?L|HnnOppxwc_[Lh.DG+`X4Nc1v{4;0]5$ufCd$vk}b+eLW,PsPc8E$GKSG$R:VznhU$t,kTg+?f"Z<R~:M=B9igR+DVzULgg8$!2^~T"RcU9:ln<8Jd4i$50@^s1{R9lVwJ^Pll*t>TN!=Lee,!77)G){k&p"C)_}`H&B|^B')
i500_latin1 <- iconv(i500_utf8, "UTF-8", "latin1")

utf8_chars <- i500_utf8 %>% strsplit("") %>% unlist %>% unique
latin1_chars <- iconv(utf8_chars, from="UTF-8", to="latin1")

catn <- function(...) {
  cat(..., "\n")
}
ntests <- 50

use_tbb <- stringfish:::is_tbb()
cat("using thread building blocks?", use_tbb, "\n")
if(use_tbb) {
  nthreads <- c(1,4)
} else {
  nthreads <- 1
}

print(sessionInfo())
print(utils::localeToCharset())

if(!altrep_support()) {
  cat("ALT-REP not supported (R < 3.5)")
} else {
  if(!is_solaris()) {
    R_TESTS <- Sys.getenv("R_TESTS") # startup.Rs
    if (nzchar(R_TESTS)) {
      R_TESTS_absolute <- normalizePath(R_TESTS)
      Sys.setenv(R_TESTS = R_TESTS_absolute)
    }
    # qserialize(readLines("~/GoogleDrive/stringfish/tests/tests.cpp"), preset = "custom", compress_level = 22) %>% base91_encode %>% catquo
src <- '
// [[Rcpp::plugins(cpp11)]]
// [[Rcpp::depends(stringfish)]]
#include <Rcpp.h>
#include "sf_external.h"
using namespace Rcpp;

// [[Rcpp::export]]
SEXP sf_alternate_case(SEXP x) {
  RStringIndexer r(x);
  size_t len = r.size();
  SEXP output = PROTECT(sf_vector(len));
  sf_vec_data & output_data = sf_vec_data_ref(output);
  size_t i = 0;
  for(auto e : r) {
    if(e.ptr == nullptr) {
      i++; // increment output index
      continue;
    }
    std::string temp(e.len, \'\\0\');
    bool case_switch = false;
    for(int j=0; j<e.len; j++) {
      if((e.ptr[j] >= 65) & (e.ptr[j] <= 90)) {
        if((case_switch = !case_switch)) {
          temp[j] = e.ptr[j] + 32;
          continue;
        }
      } else if((e.ptr[j] >= 97) & (e.ptr[j] <= 122)) {
        if(!(case_switch = !case_switch)) {
          temp[j] = e.ptr[j] - 32;
          continue;
        }
      } else if(e.ptr[j] == 32) {
        case_switch = false;
      }
      temp[j] = e.ptr[j];
    }
    output_data[i] = sfstring(temp, e.enc);
    i++;
  }
  UNPROTECT(1);
  return output;
}
'
    sourceCpp(code = src)
    if (nzchar(R_TESTS)) Sys.setenv(R_TESTS = R_TESTS)
  }
  for(.j in 1:4) {
    cat("iteration", .j, "\n")
    if(.j %% 2 == 0) {
      stringfish:::set_is_utf8_locale()
    } else {
      stringfish:::unset_is_utf8_locale()
    }
    for(nt in nthreads) {
      cat("number of threads", nt, "\n")
      catn("sf_assign")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        x <- sf_vector(10)
        y <- character(10)
        for(i in 1:10) {
          new_str <- sample(c(i500_utf8,i500_latin1),1)
          if(.j %% 2 == 1) materialize(x)
          stringfish:::sf_assign(x, i, new_str)
          y[i] <- new_str
        }
        stopifnot(string_identical(x,y))
      }
      
      catn("sf_iconv")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        x <- sf_iconv(i500_latin1, "latin1", "UTF-8")
        if(.j %% 2 == 1) materialize(x)
        y <- sf_iconv(i500_utf8, "UTF-8", "latin1")
        stopifnot(string_identical(x, i500_utf8))
        stopifnot(string_identical(y, i500_latin1))
        x <- sf_iconv(convert_to_sf(i500_latin1), "latin1", "UTF-8")
        y <- sf_iconv(convert_to_sf(i500_utf8), "UTF-8", "latin1")
        stopifnot(string_identical(x, i500_utf8))
        stopifnot(string_identical(y, i500_latin1))
      }
      
      catn("sf_nchar")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        x <- convert_to_sf(i500_latin1)
        if(.j %% 2 == 1) materialize(x)
        y <- convert_to_sf(i500_utf8)
        stopifnot( identical(sf_nchar(x, nthreads = nt), nchar(x)) )
        stopifnot( identical(sf_nchar(y, nthreads = nt), nchar(y)) )
        stopifnot( identical(sf_nchar(x, nthreads = nt), nchar(i500_latin1)) )
        stopifnot( identical(sf_nchar(y, nthreads = nt), nchar(i500_utf8)) )
        stopifnot( identical(sf_nchar(x, type = "bytes", nthreads = nt), nchar(i500_latin1, type = "bytes")) )
        stopifnot( identical(sf_nchar(y, type = "bytes", nthreads = nt), nchar(i500_utf8, type = "bytes")) )
      }
      
      catn("sf_substr")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        start <- sample(-10:10, size=1)
        if(start < 0) {
          rstart <- sf_nchar(i500_latin1, nthreads = nt) + start + 1
        } else {
          rstart <- start
        }
        stop <- sample(-10:10, size=1)
        if(stop < 0) {
          rstop <- sf_nchar(i500_latin1, nthreads = nt) + stop + 1
        } else {
          rstop <- stop
        }
        x <- sf_substr(i500_latin1, start, stop, nthreads = nt)
        if(.j %% 2 == 1) materialize(x)
        y <- substr(i500_latin1, rstart, rstop)
        x2 <- sf_substr(i500_utf8, start, stop, nthreads = nt)
        y2 <- substr(i500_utf8, rstart, rstop)
        stopifnot(string_identical(x, y))
        stopifnot(string_identical(x2, y2))
      }
      
      catn("sf_collapse")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        x <- sf_collapse(i500_latin1, collapse = ":::")
        if(.j %% 2 == 1) materialize(x)
        y <- paste0(i500_latin1, collapse = ":::")
        # stopifnot(string_identical(x, y)) # paste0 converts to UTF-8 -- doesn't respect encoding
        stopifnot(x == y)
        x <- sf_collapse(i500_latin1, collapse = ",")
        if(.j %% 2 == 1) materialize(x)
        y <- paste0(i500_latin1, collapse = ",")
        stopifnot(x == y)
        x <- sf_collapse(i500_utf8, collapse = ":::")
        if(.j %% 2 == 1) materialize(x)
        y <- paste0(i500_utf8, collapse = ":::")
        stopifnot(x == y)
        x <- sf_collapse(i500_utf8, collapse = ",")
        if(.j %% 2 == 1) materialize(x)
        y <- paste0(i500_utf8, collapse = ",")
        stopifnot(x == y)
      }
      
      catn("sf_paste")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        x <- do.call(paste, c(as.list(i500_latin1), sep=":::"))
        y <- do.call(sf_paste, c(as.list(i500_latin1), sep=":::", nthreads = nt))
        if(.j %% 2 == 1) materialize(y)
        stopifnot(x == y)
        x <- do.call(paste, c(as.list(i500_latin1), sep=":::"))
        y <- do.call(sf_paste, c(as.list(i500_latin1), sep=":::", nthreads = nt))
        if(.j %% 2 == 1) materialize(y)
        stopifnot(x == y)
        x <- do.call(paste, c(as.list(i500_utf8), sep=","))
        y <- do.call(sf_paste, c(as.list(i500_utf8), sep=",", nthreads = nt))
        if(.j %% 2 == 1) materialize(y)
        stopifnot(x == y)
        x <- do.call(paste, c(as.list(i500_utf8), sep=","))
        y <- do.call(sf_paste, c(as.list(i500_utf8), sep=",", nthreads = nt))
        if(.j %% 2 == 1) materialize(y)
        stopifnot(x == y)
      }
      
      catn("sf_readLines")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        writeLines(i500_utf8, con=myfile, useBytes=T)
        x <- sf_readLines(myfile, encoding = "UTF-8")
        if(.j %% 2 == 1) materialize(x)
        y <- readLines(myfile); Encoding(y) <- "UTF-8"
        stopifnot(string_identical(x, y))
        writeLines(i500_latin1, con=myfile)
        x <- sf_readLines(myfile, encoding = "latin1")
        if(.j %% 2 == 1) materialize(x)
        y <- readLines(myfile); Encoding(y) <- "latin1"
        stopifnot(string_identical(x, y))
      }
      
      catn("sf_grepl")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        p <- rawToChar(as.raw(c(0x5e, 0xc3, 0xb6, 0x2e, 0x2b)))
        Encoding(p) <- "UTF-8"
        p2 <- rawToChar(as.raw(c(0x5e, 0xf6, 0x2e, 0x2b)))
        Encoding(p2) <- "latin1"
        stopifnot(all(sf_grepl(i500_utf8, p, nthreads = nt) == grepl(p, i500_utf8)))
        stopifnot(all(sf_grepl(i500_latin1, p2, nthreads = nt) == grepl(p2, i500_latin1)))
        
        stopifnot(sf_grepl(i500_utf8, "[a-f]", nthreads = nt) == grepl("[a-f]", i500_utf8))
        stopifnot(sf_grepl(i500_latin1, "[a-f]", nthreads = nt) == grepl("[a-f]", i500_latin1))
      }
      
      catn("sf_grepl fixed")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        p <- rawToChar(as.raw(c(0xc3, 0xb6)))
        Encoding(p) <- "UTF-8"
        p2 <- rawToChar(as.raw(c(0xf6)))
        Encoding(p2) <- "latin1"
        stopifnot(all(sf_grepl(i500_utf8, p, fixed = T, nthreads = nt) == grepl(p, i500_utf8)))
        stopifnot(all(sf_grepl(i500_latin1, p2, fixed = T,  nthreads = nt) == grepl(p2, i500_latin1)))
        
        stopifnot(sf_grepl(i500_utf8, "[a-f]", nthreads = nt) == grepl("[a-f]", i500_utf8))
        stopifnot(sf_grepl(i500_latin1, "[a-f]", nthreads = nt) == grepl("[a-f]", i500_latin1))
      }
      
      catn("sf_gsub")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        p <- rawToChar(as.raw(c(0x5e, 0xc3, 0xb6, 0x2e, 0x2b, 0x28, 0x2e, 0x29, 0x24)))
        Encoding(p) <- "UTF-8"
        p2 <- rawToChar(as.raw(c(0x5e, 0xf6, 0x2e, 0x2b, 0x28, 0x2e, 0x29, 0x24)))
        Encoding(p2) <- "latin1"
        stopifnot(all(sf_gsub(i500_utf8, p, "$1", nthreads = nt) == gsub(p, "\\1", i500_utf8)))
        stopifnot(all(sf_gsub(i500_latin1, p2, "$1", nthreads = nt) == gsub(p2, "\\1", i500_latin1)))
        
        p <- "^h.+(.)$"
        stopifnot(all(sf_gsub(i500_utf8, p, "$1", nthreads = nt) == gsub(p, "\\1", i500_utf8)))
        stopifnot(all(sf_gsub(i500_latin1, p, "$1", nthreads = nt) == gsub(p, "\\1", i500_latin1)))
      }
      
      catn("sf_split")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        # catn("n = ", .)
        # print("sf_split_1")
        # empty split is a special case
        split <- ""
        x <- sf_split(i500_utf8, split, nthreads = nt)
        y <- stringr::str_split(i500_utf8, split)
        r <- sapply(1:length(y), function(i) {
          string_identical(x[[i]], y[[i]])
        })
        stopifnot(all(r))
        
        # print("sf_split_2")
        # empty subject
        x <- sf_split(rep("", 1e3), "a", nthreads=nt)
        stopifnot(all(x == ""))
        
        # print("sf_split_3")
        # empty subject, empty split
        x <- sf_split(rep("", 1e3), "", nthreads=nt)
        stopifnot(all(x == ""))
        
        # print("sf_split_4")
        # split not in subject
        x <- sf_split(rep("abcde", 1e3), "f", nthreads=nt)
        stopifnot(all(x == "abcde"))
        
        # print("sf_split_5")
        # single character split, including UTF-8
        split <- sf_paste(sample(utf8_chars,1))
        x <- sf_split(i500_utf8, split, nthreads = nt)
        y <- stringr::str_split(i500_utf8, split)
        r <- sapply(1:length(y), function(i) {
          string_identical(x[[i]], y[[i]])
        })
        stopifnot(all(r))
        
        # print("sf_split_6")
        # split with regex
        split <- sf_paste(sample(utf8_chars,1), ".")
        x <- sf_split(i500_utf8, split, nthreads = nt)
        y <- stringr::str_split(i500_utf8, split)
        r <- sapply(1:length(y), function(i) {
          string_identical(x[[i]], y[[i]])
        })
        stopifnot(all(r))
        
        # print("sf_split_7")
        split <- sf_paste(sample(utf8_chars,1), ".")
        split_latin1 <- sf_iconv(split, from = "UTF-8", to = "latin1")
        x <- sf_split(i500_latin1, split_latin1, nthreads = nt)
        y <- stringr::str_split(i500_latin1, split_latin1)
        x <- lapply(x, sf_iconv, from = "UTF-8", to = "latin1")
        y <- lapply(y, iconv, from = "UTF-8", to = "latin1")
        r <- sapply(1:length(y), function(i) {
          string_identical(x[[i]], y[[i]])
        })
        stopifnot(all(r))
        
        # print("sf_split_8")
        split_latin1 <- sf_iconv(split, from = "UTF-8", to = "latin1")
        x <- sf_split(i500_latin1, split_latin1, encode_mode = "byte", nthreads = nt)
        y <- stringr::str_split(i500_latin1, split_latin1)
        y <- lapply(y, iconv, from = "UTF-8", to = "latin1")
        r <- sapply(1:length(y), function(i) {
          string_identical(x[[i]], y[[i]])
        })
        stopifnot(all(r))
      }
      
      catn("sf_toupper and sf_tolower")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        x1 <- sf_toupper(i500_latin1)
        x2 <- sf_toupper(i500_utf8)
        y1 <- sf_tolower(i500_latin1)
        if(.j %% 2 == 1) materialize(y1)
        y2 <- sf_tolower(i500_utf8)
        if(.j %% 2 == 1) materialize(y2)
        z1 <- sf_tolower(x1)
        z2 <- sf_tolower(x2)
        stopifnot(string_identical(z1, i500_latin1))
        stopifnot(string_identical(z2, i500_utf8))
        stopifnot(string_identical(y1, i500_latin1))
        stopifnot(string_identical(y2, i500_utf8))
        # base R functions also convert Unicode characters to upper
        # stopifnot(string_identical(x1, iconv(toupper(i500_latin1),"UTF-8", "latin1")))
        # stopifnot(string_identical(x2, toupper(i500_utf8)))
      }
      
      if(!is_solaris()) {
        catn("Rcpp test with sf_alternate_case")
        for(. in 1:ntests) {
          x <- c("hello world", "HELLO WORLD")
          string_identical(sf_alternate_case(x), c("hElLo wOrLd", "hElLo wOrLd"))
        }
        
        catn("sf_trim")
        for(. in 1:ntests) {
          # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
          x <- sf_trim(sf_paste("\t", i500_utf8, " \n"))
          if(.j %% 2 == 1) materialize(x)
          stopifnot(string_identical(x, i500_utf8))
          
          x <- sf_trim(sf_paste("\t", i500_latin1, " \n"), encode_mode = "byte")
          if(.j %% 2 == 1) materialize(x)
          stopifnot(string_identical(x, i500_latin1))
        }
      }
      
      # Disable check due to https://bugs.r-project.org/show_bug.cgi?id=18211
      catn("sf_match")
      # gctorture(TRUE)
      for(. in 1:ntests) {
        i500_utf8_shuffled <- c(NA_character_, i500_utf8[sample(length(i500_utf8))][-1])
        temp <- c(i500_utf8, NA_character_)
        x <- sf_match(temp, i500_utf8_shuffled)
        # y <- match(temp, i500_utf8_shuffled)
        # stopifnot(identical(x,y))
        i500_latin1_shuffled <- c(NA_character_, i500_latin1[sample(length(i500_latin1))][-1])
        temp <- c(i500_latin1, NA_character_)
        x <- sf_match(c(i500_latin1, NA_character_), i500_latin1_shuffled)
        # y <- match(temp, i500_latin1_shuffled)
        # stopifnot(identical(x,y))
      }
      # gctorture(FALSE)
      
      catn("sf_compare")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        i500_utf8_shuffled <- i500_utf8
        i500_utf8_shuffled[sample(length(i500_utf8), size = 100)] <- ""
        x <- sf_compare(c(i500_utf8, NA_character_), c(i500_utf8_shuffled, NA_character_))
        y <- c(i500_utf8, NA_character_) == c(i500_utf8_shuffled, NA_character_)
        stopifnot(identical(x,y))
        
        i500_latin1_shuffled <- i500_latin1
        i500_latin1_shuffled[sample(length(i500_latin1), size = 100)] <- ""
        x <- sf_compare(c(i500_latin1, NA_character_), c(i500_latin1_shuffled, NA_character_))
        y <- c(i500_latin1, NA_character_) == c(i500_latin1_shuffled, NA_character_)
        stopifnot(identical(x,y))
      }
      
      catn("sf_concat")
      for(. in 1:ntests) {
        # if(. == 1) {gctorture(TRUE)} else {gctorture(FALSE)}
        i500_utf8_shuffled <- i500_utf8
        i500_utf8_shuffled[sample(length(i500_utf8), size = 100)] <- ""
        x <- sfc(sfc(i500_utf8, NA_character_), sfc(i500_utf8_shuffled, NA_character_), character(0))
        if(.j %% 2 == 1) materialize(x)
        y <- sfc(sfc(sf_convert(i500_utf8), NA_character_), sfc(sf_convert(i500_utf8_shuffled), NA_character_), character(0))
        z <- c(c(sf_convert(i500_utf8), NA_character_), c(sf_convert(i500_utf8_shuffled), NA_character_), character(0))
        z2 <- c(c(i500_utf8, NA_character_), c(i500_utf8_shuffled, NA_character_), character(0))
        stopifnot(string_identical(x,y))
        stopifnot(string_identical(x,z))
        stopifnot(string_identical(x,z2))
        stopifnot(string_identical(x,y))
        stopifnot(identical(sfc(character(0)), character(0)))
      }
      print(gc())
    }
  }
}
print("end")