| .ptime <- proc.time() | |
| ### tests of recursion in PCRE matching | |
| ### Based on PR16757 | |
| ## This is expected to throw a warning at some point if PCRE uses a stack, | |
| ## depending on the system and stack size. | |
| ## Typical stack 8-10M, some people use 40M. | |
| pcre_config()["stack"] | |
| op <- options(warn = 1) | |
| for (n in c(seq(5000L, 10000L, 1000L), 20000L, 50000L, 100000L)) { | |
| print(n) | |
| x <- paste0(rep("a", n), collapse="") | |
| print(grepl("(a|b)+", x, perl = TRUE)) | |
| } | |
| options(op) | |
| ### tests of PCRE's JIT. | |
| if(!pcre_config()["JIT"]) { | |
| message("These tests are pointless without JIT support") | |
| q("no") | |
| } | |
| ## Test from example(grep) | |
| txt2 <- c("The", "licenses", "for", "most", "software", "are", | |
| "designed", "to", "take", "away", "your", "freedom", | |
| "to", "share", "and", "change", "it.", | |
| "", "By", "contrast,", "the", "GNU", "General", "Public", "License", | |
| "is", "intended", "to", "guarantee", "your", "freedom", "to", | |
| "share", "and", "change", "free", "software", "--", | |
| "to", "make", "sure", "the", "software", "is", | |
| "free", "for", "all", "its", "users") | |
| grep("[gu]", txt2, perl = TRUE) | |
| st <- function(expr) sum(system.time(expr)[1:2]) | |
| ## here JIT is slightly slower | |
| options(PCRE_study = FALSE) | |
| st(for(i in 1:1e4) grep("[gu]", txt2, perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = FALSE) | |
| st(for(i in 1:1e4) grep("[gu]", txt2, perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = TRUE) | |
| st(for(i in 1:1e4) grep("[gu]", txt2, perl = TRUE)) | |
| ## and for more inputs, study starts to pay off | |
| txt3 <- rep(txt2, 10) | |
| options(PCRE_study = FALSE) | |
| st(for(i in 1:1e3) grep("[gu]", txt3, perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = FALSE) | |
| st(for(i in 1:1e3) grep("[gu]", txt3, perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = TRUE) | |
| st(for(i in 1:1e3) grep("[gu]", txt3, perl = TRUE)) | |
| ## An example where JIT really pays off (e.g. 10x) | |
| pat <- "([^[:alpha:]]|a|b)+" | |
| long_string <- paste0(rep("a", 1023), collapse="") | |
| N <- 10 | |
| options(PCRE_study = FALSE, PCRE_use_JIT = FALSE) | |
| st(for(i in 1:1e3) grep(pat, rep(long_string, N), perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = FALSE) | |
| st(for(i in 1:1e3) grep(pat, rep(long_string, N), perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = TRUE) | |
| st(for(i in 1:1e3) grep(pat, rep(long_string, N), perl = TRUE)) | |
| ## This needs to test 50 strings to see much gain from study | |
| txt <- rep("a test of capitalizing", 50) | |
| options(PCRE_study = FALSE, PCRE_use_JIT = FALSE) | |
| st(for(i in 1:1e4) gsub("(\\w)(\\w*)", "\\U\\1\\L\\2", txt, perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = FALSE) | |
| st(for(i in 1:1e4) gsub("(\\w)(\\w*)", "\\U\\1\\L\\2", txt, perl = TRUE)) | |
| options(PCRE_study = TRUE, PCRE_use_JIT = TRUE) | |
| st(for(i in 1:1e4) gsub("(\\w)(\\w*)", "\\U\\1\\L\\2", txt, perl = TRUE)) | |
| cat("Time elapsed: ", proc.time() - .ptime,"\n") |