Re: What should /\08/ match and do?

On Tue, Jun 01, 2010 at 11:52:38AM -0600, karl williamson wrote:
>
> What about \008 ?  The pods say that a two or three digit octal number  
                                                      ^^^^^^^^^^^  Hmmm ;-)
> follows a backslash, but it turns out that a single digit "works" in  
> character classes, eg. [\7]

The reason backslash followed by a 'single' octals (except for \0)
isn't a character escape is that they are backreferences; regardless
of the number of capturing parenthesis. A backslash followed by two or
three octals is only a character espace if it either starts with a 0,
or cannot be a backreference (due to lack of capturing parenthesis).
(Interestingly enough, the numbers are interpreted as digits when it's
a backreference; and as octals when being a numeric escape).  But inside
a bracketed character class, you cannot have a backreference - so \7 is
an octal escape.

IMO, this potential confusion of \OOO being either an octal escape, or a
backreference is a more serious problem than \008. For instance,

    /(.)(.)(.)(.)(.)(.)(.)(.)(.)\10/

will suddenly match something else if another set of parens is introduced:

    /((.)(.)(.)(.)(.)(.)(.)(.)(.))\10/

This one is sneaky as well:

    $a = '(.)\1';
    "aa" =~ /${a}/;      # True
    "aa0" =~ /${a}0/;    # False!
    "aa\x8" =~ /${a}0/;  # True!

I try to always use \g{} when using backreferences, and avoid octal character
escapes, but it's hard to kill a habit. (And $WORK still uses 5.8.5...)

If I had a time machine, I'd go back in time, and, using the power of
hindsight, explained to Larry it would be better to mandate a leading
0 for octal escapes. Or better yet, to use \g{} for backrefences from
the start.

> What about \x{} ?  It currently yields \000 silently.  I originally
> proposed deprecating it, but this met objections, but I do think it  
> should warn, even if not deprecated.

I'm not easily convinced about introducing warnings about constructs that
have been around of a long time. Do you have any evidence there are many
people using "\x{}", and expecting it to match something other than \000?
If not, I don't see any benefit for a warning; it won't help people, and
it will annoy people who have been using "\x{}" to match "\000".

> \N{} yields two errors, one correct, one misleading:
> ./perl -Mcharnames=:full -Ilib -E '"\N{}"'
> Unknown charname '' at lib/unicore/Name.pl line 1
> Deprecated character in \N{...}; marked by <-- HERE  in \N{}<-- HERE  at  
> -e line 1.
>
> And there is:
>  ./perl -Mcharnames=:full -Ilib -E '"\N{U+}"'
> Invalid hexadecimal number in \N{U+...} at -e line 1, within string
> Execution of -e aborted due to compilation errors.


I don't have a problem with them being errors. I don't have a problem
with them being funky ways of writing "\000" either (but then, it should
be documented, and not warn).



Abigail