Bug report: POSIX regular expression doesn't match for somewhat higher values of upper bound

  <dietmar.schindler at manroland-web.com>
    on Tue, 4 Apr 2017 08:45:30 +0000 writes:

> Dear Sirs,
    > while

    >> regexpr('(.{1,2})\\1', 'foo')
    > [1] 2
    > attr(,"match.length")
    > [1] 2
    > attr(,"useBytes")
    > [1] TRUE

    > yields the correct match, an incremented upper bound in

    >> regexpr('(.{1,3})\\1', 'foo')
    > [1] -1
    > attr(,"match.length")
    > [1] -1
    > attr(,"useBytes")
    > [1] TRUE

    > incorrectly yields no match.

Hmm, yes, I would also say that this is incorrect
(though I'm always cautious: The  ?regex  help page explicitly
 mentions greedy repetitions, and these can "bite you" ..)

The behavior is also different from the  perl=TRUE one which is
correct (according to the above understanding).

Using  grep() instead of regexpr() makes the behavior easier to parse.
The following code 
----------------------------------------------------------------------

tx <- c("ab","abc", paste0("foo", c("", "b", "o", "bar", "oofy")))
setNames(nchar(tx), tx)
## ab     abc     foo    foob    fooo  foobar foooofy
##  2       3       3       4       4       6       7

grep1r <- function(n, txt, ...) {
    pattern <- paste0('(.{1,',n,'})\\1', collapse="") ## can have empty n
    ans <- grep(pattern, txt, value=TRUE, ...)
    cat(sprintf("pattern '%s' : ", pattern)); print(ans, quote=FALSE)
    invisible(ans)
}

grep1r({}, tx)# '.{1,}' : because of _greedy_ matching there is __no__ repetiion!
grep1r(100,tx)# i.e., these both give an empty match :  character(0)

## matching at most once:
grep1r(1, tx)# matches all 5 starting with "foo"
grep1r(2, tx)# ditto    : all have more than 2 chars
grep1r(3, tx)# not "foo": those with more than 3 chars
grep1r(4, tx)# .. those with more than 4 characters
grep1r(5, tx)# .. those with more than 5 characters
grep1r(6, tx)# .. those with more than 6 characters
grep1r(7, tx)# NONE (= those with more than 7 characters)

for(p in c(FALSE,TRUE)) {
    cat("\ngrep(*, perl =", p, ") :\n")
    for(n in c(list(NULL), 1:7))
        grep1r(n, tx, perl = p)
}

----------------------------------------------------------------------

ends with

for(p in c(FALSE,TRUE)) {

+     cat("\ngrep(*, perl =", p, ") :\n")
+     for(n in c(list(NULL), 1:7))
+         grep1r(n, tx, perl = p)
+ }

grep(*, perl = FALSE ) :
pattern '(.{1,})\1' : character(0)
pattern '(.{1,1})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,2})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,3})\1' : [1] foob    fooo    foobar  foooofy
pattern '(.{1,4})\1' : [1] foobar  foooofy
pattern '(.{1,5})\1' : [1] foobar  foooofy
pattern '(.{1,6})\1' : [1] foooofy
pattern '(.{1,7})\1' : character(0)

grep(*, perl = TRUE ) :
pattern '(.{1,})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,1})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,2})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,3})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,4})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,5})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,6})\1' : [1] foo     foob    fooo    foobar  foooofy
pattern '(.{1,7})\1' : [1] foo     foob    fooo    foobar  foooofy

Von: Martin Maechler [mailto:maechler at stat.math.ethz.ch]
Gesendet: Mittwoch, 5. April 2017 11:15

  <dietmar.schindler at manroland-web.com>
    on Tue, 4 Apr 2017 08:45:30 +0000 writes:

    > Dear Sirs,
    > while

    >> regexpr('(.{1,2})\\1', 'foo')

    > [1] 2
    > attr(,"match.length")
    > [1] 2
    > attr(,"useBytes")
    > [1] TRUE

    > yields the correct match, an incremented upper bound in

    >> regexpr('(.{1,3})\\1', 'foo')

    > [1] -1
    > attr(,"match.length")
    > [1] -1
    > attr(,"useBytes")
    > [1] TRUE

    > incorrectly yields no match.

Hmm, yes, I would also say that this is incorrect
(though I'm always cautious: The  ?regex  help page explicitly
 mentions greedy repetitions, and these can "bite you" ..)

The behavior is also different from the  perl=TRUE one which is
correct (according to the above understanding).

...

Shouldn't this be submitted on R's Bugzilla then (which I as a non-member can't)?
--
Best regards,
Dietmar Schindler
________________________________
manroland web systems GmbH -- Managing Director: Alexander Wassermann
Registered Office: Augsburg -- Trade Register: AG Augsburg -- HRB-No.: 26816 -- VAT: DE281389840

Confidentiality note:
This eMail and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you are not the intended recipient, you are hereby notified that any use or dissemination of this communication is strictly prohibited. If you have received this eMail in error, then please delete this eMail.