(This post was in progress before your questions about Find All… I came back to this post later, so that future readers of the thread will know which character matches which.)
@Ramanand-Jhingade wrote:
I finally found a solution here: how-to-find-non-ascii-unprintable-characters-using-notepad-plus-plus
We have to select the Regular expression mode and search/find with this code: [\x00-\x08\x0B\x0C\x0E-\x1F\x7F-\x9F]
Good job! I applaud your persistence.
However, the solution surprises me.
It didn’t make sense to me that \x93 wouldn’t find “ but that [\x00-\x08\x0B\x0C\x0E-\x1F\x7F-\x9F] would. In an ANSI-encoded file, that search makes perfect sense, because those bytes aren’t in ASCII or extended-ASCII. But in a UTF-encoded file, those bytes aren’t valid unless they are part of a multi-byte UTF-8 encoding sequence.
So I started cutting parts out of that expression until I narrowed it down to what’s matching: \x13 matched the x93 byte.
With a UTF-8 encoded file that has bad bytes with the MSB set that don’t get interpreted as part of a multi-byte UTF-8 character sequence, the following appears true:
\x13 will match the
\x14 will match the
For example:
So I built up a test file , and started doing searches:
searching \x00 matches many, including x80, xA0, xB0, xB8, xBC, xBE, xC0, xE0, xF0, xF8, xFC, xFE
searching \x01 matches eight, including x81, xA1, xB1, xB9, xBD, xC1, xE1, xF1, xF9, xFD
searching \x02 matches eight, including x82, xA2, xB2, xBA, xC2, xE2, xF2, xFA
searching \x03 matches eight, including x83, xA3, xB3, xBB, xC3, xE3, xF3, xFB
searching \x04 matches six, including x84, xA4, xB4, xC4, xE4, xF4
searching \x05, \x06, and \x07 each match the seven in the same pattern as \x04
searching \x08 only matches four, including x88, xA8, xC8, xF8 . \x09 works analogously
searching \x0A matches that same pattern of four, but also matches all the line-feed character in the EOL sequences
searching \x0B and \x0C each match four, including x8B, xAB, xCB, xFB, x8C, xAC, xCC, xFC,
searching \x0D matches the pattern of four, and all the carriage returns in the EOL sequences
searching \x0E matches the pattern of four: x8E, xAE, xCE, xEE
searching \x0F matches three of the four (but my file didn’t have x8F to match against)
searching \x10 matches two: x90 and xD0. \x11, \x12, \x13, \x14, \x15, \x16, \x17, \x18, \x19, \x1A, \x1B, \x1C, \x1D, \x1E, \x1F all match the two entries from the 90 and D0 rows.
And no, @Alan-Kilborn , to answer your chat question, it does not change behavior inside of the [] character class.
As near as I can tell, in this poorly-encoded situation, the search engine seems to be applying certain bit-masks when searching for invalid bytes, where it masks out some of the upper bits while looking for matches. (My guess is that the multiple matches for a given search escape has to do with the fact that in valid UTF-8, 8-bit sequences that are 10xxxxxx can only be the second, third, or fourth byte of the multibyute sequence (so that’s 0x80-0xBF); 110xxxxx is the start of a 2-byte sequence (0xC0-0xDF); 1110xxxx is the start of a 3-byte sequence (0xE0-0xEF); and 1111xxxx is the start of a 4-byte sequence (0xF0-0xFF). But that’s a wild guess.)
Remember, all of these were for my experiments, with a file that Notepad++ is interpreting as UTF-8, but have these badly-encoded single bytes with spaces in between. This isn’t a general search technique in a well-formed file with the right encoding selected. None of this is best-practice, but is only here to help correct mal-formed UTF-8 files.
