Hi Julien, I ran some quick tests. At first I thought they failed. But it turned out that even though I replaced the regex.dll copied in the new pcreposix.dll file (after they were inadvertantly downgraded when I loaded a new PowerPro beta), the old versions were still in memory. :D
I used replaceg to replace in a file.readall that had crlfs: .+ to *\0* ^.+ to *\0* (?m)^.+ to *\0* I reviewed the results in a hex editor, so I could see that the \r's were not eaten by dot. Then I reran the above tests, first replacing all the \r with "" These looked fine. Back to test data with crlfs: \d. to *\0* where the last character b4 \r's was a digit (they didn't match = good) Then .\n to *\0* really thought this might match the \r\n but nope=good Also, this worked in 6.7, still works in 7 (?m)^.+$ to *\0* So my preliminary opinion is that the "any" file would be good for crlf text files and unix text files. I can't speak for utf-8. I wonder what is the impact on pcrecpp? I assume it is or will be updated if necessary, since I know it does some crlf processing of its own. I guess the option value is different in this "any" file. Regards, Sheri --- In [email protected], Julien Pierrehumbert <[EMAIL PROTECTED]> wrote: >> > > Hi regex geeks, > > PCRE7 is out. > > Here's a couple of untested builds for use with the > regex plugins: a CRLF version which should hopefully > included the planned changes WRT partial skipping of > CRLF and an "any" version which should treat any > newline character(s) (including Unicode line endings) > as a newline. Do these build work? Is the "any" build > superior or inferior to the CRLF build? If it works > well, is there anyone who has strong feelings regarding > the use of the "any" build as the default PCRE library? >
