NAME

Perl::Critic::Policy::ValuesAndExpressions::ProhibitUnknownBackslash - don't use undefined backslash forms

DESCRIPTION

This policy is part of the "Perl::Critic::Pulp" add-on. It checks for unknown backslash escapes like

    print "\*.c";      # bad

This is harmless, assuming the intention is a literal "*" (which it gives), but unnecessary, and on that basis this policy is under the "cosmetic" theme (see "POLICY THEMES" in Perl::Critic). Sometimes it can be a misunderstanding or a typo though, for instance a backslashed newline is a newline, but perhaps you thought it meant a continuation.

    print "this\       # bad
    is a newline";

Perl already warns about unknown escaped alphanumerics like "\v" under "perl -w" or "use warnings" (see "Unrecognized escape \%c passed through" in perldiag).

    print "\v";        # bad, and provokes Perl warning

This policy extends to report on any unknown escape, with options below to vary the strictness and to check single-quote strings too if desired.

Control Characters \c

Control characters "\cX" are checked and only the conventional A-Z a-z @ [ \ ] ^ _ ? are considered known.

    print "\c*";       # bad

Perl accepts any "\c" and does an upcase xor 0x40, so "\c*" is letter "j", at least on an ASCII system. But that's obscure and likely to be a typo or error.

For reference, "\c\" is the ASCII FS "file separator" and the second backslash is not an escape, except for a closing quote character, which it does escape (basically because Perl scans for a closing quote before considering interpolations). Thus,

    print " \c\  ";     # ok, control-\ FS
    print " \c\" ";     # bad, control-" is unknown
    print qq[ \c\]  ];  # ok, control-] GS

Ending Interpolation

A backslashed colon, bracket, brace or dash is allowed after an interpolated variable or element, since this stops interpolation at that point.

    print "$foo\::bar";    # ok, $foo
    print "@foo\::";       # ok, @foo

    print "$foo[0]\[1]";   # ok, is $foo[0]
    print "$esc\[1m";      # ok

    print "$foo\{k}";      # ok
    print "$foo\{k}";      # ok
    print "$foo{k}\[0]";   # ok, is $foo{k}
    print "@foo\{1,2}";    # ok, is @foo

    print "$foo\->[0]";    # ok, is $foo
    print "$foo\->{zz}";   # ok

A single backslash like "\::" is enough for the colon case, but backslashing the second too as "\:\:" is quite common and is allowed.

    print "$#foo\:\:bar";  # ok

Only an array or hash "->[]" or "->{}" need "\-" to stop interpolation. Other cases such as an apparent method call or arrowed coderef call don't interpolate and the backslash is treated as unknown since unnecessary.

    print "$coderef\->(123)";        # bad, unnecessary
    print "Usage: $class\->foo()";   # bad, unnecessary

For reference, the alternative in all the above is to write "{}" braces around the variable or element to delimit from anything following. Doing so may be clearer than backslashing,

    print "${foo}::bar";    # alternatives
    print "@{foo}::bar";
    print "$#{foo}th";
    print "${foo[0]}[1]";   # array element $foo[0]

The full horror story of backslashing interpolations can be found in "Gory details of parsing quoted constructs" in perlop.

Octal Wide Chars

Octal escapes "\400" to "\777" for wide chars 256 to 511 are new in Perl 5.6. They're considered unknown in 5.005 and earlier (where they end up chopped to 8-bits 0 to 255). Currently if there's no "use" etc Perl version then it's presumed a high octal is intentional and is allowed.

    print "\400";    # ok

    use 5.006;
    print "\777";    # ok

    use 5.005;
    print "\777";    # bad in 5.005 and earlier

Named Chars

Named chars "\N{SOME THING}" are added by charnames, new in Perl 5.6. In Perl 5.16 up, that module is automatically loaded when "\N" is used. "\N" is considered known when "use 5.016" or higher,

    use 5.016;
    print "\N{EQUALS SIGN}";   # ok with 5.16 automatic charnames

or if "use charnames" is in the lexical scope,

    { use charnames ':full';
      print "\N{APOSTROPHE}";  # ok
    }
    print "\N{COLON}";         # bad, no charnames in lexical scope

In Perl 5.6 through 5.14, a "\N" without "charnames" is a compile error so would be seen in those versions immediately anyway. There's no check of the character name appearing in the "\N". "charnames" gives an error for unknown names.

The "charnames" option ("CONFIGURATION" below) can be allow to always allow named characters. This can be used for instance if you always have Perl 5.16 up but without declaring that in a "use" statement.

The "charnames" option can be disallow to always disallow named characters. This is a blanket prohibition rather than an UnknownBackslash as such, but matches the allow option. Disallowing can be matter of personal preference or perhaps aim to save a little memory or startup time.

Other Notes

In the violation messages, a non-ascii or non-graphical escaped char is shown as hex like "\{0x263A}", to ensure the message is printable and unambiguous.

Interpolated $foo or "@{expr}" variables and expressions are parsed like Perl does, so backslashes for refs within are ok, in particular tricks like "${\scalar ...}" are fine (see "How do I expand function calls in a string?" in perlfaq4).

    print "this ${\(some()+thing())}";   # ok

Disabling

As always, if you're not interested in any of this then you can disable "ProhibitUnknownBackslash" from your .perlcriticrc in the usual way (see "CONFIGURATION" in Perl::Critic),

    [-ValuesAndExpressions::ProhibitUnknownBackslash]

CONFIGURATION

"double" (string, default "all")

"heredoc" (string, default "all")

"double" applies to double-quote strings "", "qq{}", "qx{}", etc. "heredoc" applies to interpolated here-documents "<<HERE" etc. The possible values are

    none       don't report anything
    alnum      report unknown alphanumerics, like Perl's warning
    quotemeta  report anything quotemeta() doesn't escape
    all        report all unknowns

"alnum" does no more than compiling with "perl -w", but might be good for checking code you don't want to run.

"quotemeta" reports escapes not produced by "quotemeta()". For example "quotemeta" escapes a "*", so "\*" is not reported, but it doesn't escape an underscore "_", so "\_" is reported. The effect is to prohibit a few more escapes than "alnum". One use is to check code generated by other code where you've used "quotemeta" to produce double-quoted strings and thus may have escaping which is unnecessary but works fine.

"single" (string, default "none")

"single" applies to single-quote strings '', "q{}", "qx''", etc. The possible values are as above, though only "all" or "none" make much sense.

    none       don't report anything
    all        report all unknowns

The default is "none" because literal backslashes in single-quotes are usually both what you want and quite convenient. Setting "all" effectively means you must write backslashes as "\\".

    print 'c:\my\msdos\filename';     # bad under "single=all"
    print 'c:\\my\\msdos\\filename';  # ok

Doubled backslashing like this is correct, and can emphasize that you really did want a backslash, but it's tedious and not easy on the eye and so left only as an option.

For reference, single-quote here-documents "<<'HERE'" don't have any backslash escapes and so are not considered by this policy. "qx{}" command backticks are double-quote but "qx''" is single-quote. They are treated per the corresponding "single" or "double" option.

"charnames" (string, default "version")

Whether to treat named characters "\N{}" in double-quote strings as known or unknown,

    version    known if use charnames or use 5.016
    allow      always allow
    disallow   always disallow

BUGS

Interpolations in double-quote strings are found by some code here in "ProhibitUnknownBackslash" (re-parse the string content as Perl code starting from the "$" or "@"). If this fails for some reason then a warning is given and the rest of the string is unchecked. In the future would like PPI to parse interpolations, for the benefit of string chopping like here or checking of code in an interpolation.

HOME PAGE

http://user42.tuxfamily.org/perl-critic-pulp/index.html

COPYRIGHT

Perl-Critic-Pulp is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version.

Perl-Critic-Pulp is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with Perl-Critic-Pulp. If not, see <http://www.gnu.org/licenses>.