Implementing "quotemeta" \Q ... \E in Tcl?

Question

The typical use-case is when a regex needs to include user input. Characters with special meaning in regex (i.e. "the dirty dozen" in Perl) need to be escaped. Perl provides the "quotemeta" functionality to do this: simply encapsulate interpolating variables in \Q and \E. But Tcl provides no such functionality (and according to this page, even with ARE).

Is there a good (rigorous) implementation of quotemeta in Tcl out there?

see: http://stackoverflow.com/a/4352893/390913 – perreal Jul 12 '12 at 04:34 — perreal, Jul 12 '12 at 04:34

Keith Thompson · Accepted Answer · 2012-07-12T14:45:47.390

3

Perl's quotemeta function simply replaces every non-word character (i.e., characters other than the 26 lowercase letters, the 26 uppercase letters, the 10 digits, and underscore) with a backslash. This is overkill, since not all non-word characters are regexp metacharacters, but it's simple and safe, since escaping a non-word character that doesn't need escaping is harmless.

I believe this implementation is correct:

proc quotemeta {str} {
    regsub -all -- {[^a-zA-Z0-9_]} $str {\\&} str
    return $str
}

But thanks to glenn's comment, this one is better, at least for modern versions of Tcl (\W matches any non-word character starting some time after Tcl 8.0.5):

proc quotemeta {str} {
    regsub -all -- {\W} $str {\\&} str
    return $str
}

(I'm assuming that Tcl's regular expressions are similar enough to Perl's so that this will do the same job in Tcl that it does in Perl.)

edited Jul 12 '12 at 14:45

answered Jul 11 '12 at 22:06

Keith Thompson

254,901
44
429
631

2

Like Perl, Tcl has `\W` to match a non-word char -- http://www.tcl.tk/man/tcl8.5/TclCmd/re_syntax.htm#M65 – glenn jackman Jul 12 '12 at 03:52
I didn't know Perl implemented it so simply! I guess I thought it possible that some escaped special characters I didn't now about might have special meaning. Thanks. – Andrew Cheong Jul 12 '12 at 14:33
@acheong87: Future versions of Perl could easily add more metacharacters. Keeping `quotemeta` simple means it doesn't have to change to track changes in the regexp syntax. – Keith Thompson Jul 12 '12 at 14:35
Agreed, thanks. And do keep the first version up, because Tcl 8.0.5 doesn't support special character classes like ``\W``. Also, I noticed that whitespace characters will be escaped, but I think that is fine and safer anyway. – Andrew Cheong Jul 12 '12 at 14:40
@acheong87: Thanks for the info about Tcl 8.0.5; I've added it to my answer. – Keith Thompson Jul 12 '12 at 14:46
1

While there are a lot of differences in the detail, backslashing all non-alnum characters (i.e., `\W`) is entirely adequate for the same reason it's adequate in Perl. The RE engine was indeed changed in 8.1; you're not really recommended to use 8.0 any more though, as 8.4 is entirely faster (and packed with many more goodies too). – Donal Fellows Jul 12 '12 at 14:50
@DonalFellows: Alas, it's not always practical to use the latest-and-greatest version. I know that's true for Perl; I'm sure it's equally true for Tcl. (If you *can* count on having a reasonably modern version on *all* the systems where your code needs to run, that's great.) – Keith Thompson Jul 12 '12 at 19:06

Andrew Cheong · Answer 2 · 2012-07-12T14:23:37.487

1

I'll propose a solution, but I'm not confident it's correct.

#
#   notes
#
#   -  "[]" has to appear in the beginning of a character class
#   -  "-" has to come last in a character class
#   -  "#" is not special, but anticipating the x modifier...
#   -  "-" is not special, but anticipating interpolation within "[]"...
#   -  "/" is not special in Tcl
#
proc quotemeta {str} {
    regsub -all -- {[][#$^*()+{}\|.?-]} $str {\\\0} str
    return $str
}

edited Jul 12 '12 at 14:23

answered Jul 11 '12 at 21:57

Andrew Cheong

29,362
15
90
145

1

You have to swap the `[]` to `][` to make it compile (instead of throwing an exception) and the backslash needs to be doubled to make it “really” there. `{[][#$^*()+{}\\|.?-]}` works when I run it against a punctuation sampler. – Donal Fellows Jul 11 '12 at 23:43
Thanks, @DonalFellows! Edited. But for me I don't seem to need the double backslash. tclsh (8.0.5, I'm that guy): ``% quotemeta {$6.00 - Magic Hat #9 \ [Draught Special]}`` outputs ``\$6\.00 \- Magic Hat \#9 \\ \[Draught Special\]`` – Andrew Cheong Jul 12 '12 at 14:30

Implementing "quotemeta" \Q ... \E in Tcl?

2 Answers2