Bound the work a match may do — re

PCRE2 matches by backtracking, so a pattern such as (a+)+b can take time exponential in the length of the subject. zuregex makes no linear-time promise; for that, use the re2 package. What it promises is that limits bound the work per element, and that hitting one is either a classed error or an NA you asked for.

Usage

re_limits(match = NULL, depth = NULL, heap = NULL, on_limit = c("error", "na"))

Arguments

match: The most times PCRE2's internal match function may be called for one search; PCRE2's default is 10,000,000. The JIT compiler honours it too.
depth: The deepest backtracking nesting for one search; PCRE2's default equals match. The JIT compiler ignores it.
heap: The most heap memory, in kibibytes, the interpreter may use for backtracking in one search; PCRE2's default is 20,000,000. The JIT compiler ignores it.
on_limit: What an element that hits a limit gives: "error" raises a zuregex_error_limit naming the limit and the element; "na" gives NA for that element and one warning of class zuregex_warning_limit for the call, counting how many did.

Value

An object of class re_limits.

Details

Pass re_limits() as limits = to any matching function, or set it for the session with options(zuregex.limits = re_limits(...)). The argument wins over the option. NULL keeps PCRE2's default for that limit.

Examples

x <- strrep("a", 30)
try(re_detect(x, "(a+)+b", limits = re_limits(match = 1000)))
#> [1] FALSE
re_detect(c(x, "ab"), "(a+)+b", limits = re_limits(match = 1000, on_limit = "na"))
#> [1] FALSE  TRUE

old <- options(zuregex.limits = re_limits(match = 1e5))
options(old)