Tcl Source Code

View Ticket
Login
Ticket UUID: 219233
Title: string match oddity when - is the last char
Type: Bug Created on: 2000-10-26 05:03:56
Submitter: nobody Assigned to: dkf
Subsystem: 18. Commands M-Z Severity: Minor
Priority: 9 Immediate Last modified: 2026-10-19 17:56:34
Status: Open Closed by: nobody
Resolution: Remind Closed on:
Version: 8.6.18, 9.0.5, 9.1.0
Description:
OriginalBugID: 4438 Bug
Version: 8.3
SubmitDate: '2000-03-21'
LastModified: '2000-04-03'
Severity: MED
Status: Assigned
Submitter: techsupp
ChangedBy: hobbs
OS: Windows 95
OSVersion: OSR2
FixedDate: '2000-10-25'
ClosedDate: '2000-10-25'


Name:
Keith Lea

ReproducibleScript:
% string match {[a-z0-9_/-]} \\
1
% string match {[a-z0-9_/]} \\
0



It's accidently interp'ing the "/-]" as "/-]]", taking last
] as ] endrange and ] endblock. 
-- 04/03/2000 hobbs
This needs to be fixed in Tcl_String(Case)Match in tclUtil.c,
but wait until 8.4 just in case someone was counting on the
previous perverse behavior. 
-- 04/03/2000 hobbs
User Comments:
dkf added on 2000-11-24 18:35:30:
Hmm.  The problem seems to be that the first pattern is actually malformed by the rules of [string match], but there is no way to indicate this.  I suppose the correct way of dealing with this is to decide that we were not really matching a range after all, but that's not very good at all.  Either that, or we state that a malformed pattern matches nothing at all.

Hmm.  On successful matching of a range, should we really back up a character at the unexpected end of string , or should we fail at that point?

dkf added on 2000-11-24 20:18:50:
Improved detection of bug:
  % sstring match \[a-] ]
  1
  % string match \[a-]x ]x
  0

dkf added on 2001-02-15 23:32:37:
Apparently, the way [string match] handles syntactically invalid patterns is by failing to match anything at all.  It's not entirely clear to me that this is an optimal strategy...

dkf added on 2001-02-23 17:24:41:
See Patch #103932 https://sourceforge.net/patch/?func=detailpatch&patch_id=103932&group_id=10894

dkf added on 2001-03-17 03:57:56:
Logged In: YES 
user_id=79902

This behaviour won't be changed before 9.0

Any fixes that *are* done must be applied to code in both tclUtil.c and tclUtf.c

lupylucke added on 2003-11-23 15:20:03:
Logged In: YES 
user_id=915599

I encountered a problem with `string match' too. I suppouse
it is, finally, the same bug: it is possible to match a [ in
a character set quoting int with \. Thits should work with ]
too, but it doesn't!

% string match {[\[]} {[}
1
% string match {[\]]} {]}
0

dkf added on 2003-11-24 23:17:10:
Logged In: YES 
user_id=79902

Backslashes aren't special inside the square-bracket term,
so the first match doesn't match what you are expecting and
the second match isn't looking for what you think it is:

% set str \\
\
% string match {[\[]} $str
1
% set str {\]}
\]
% string match {[\]]} $str
1

The glob-matching engine used by [string match] isn't very
smart...  :^(  Consider using regular expressions instead.

matzek added on 2007-06-14 20:56:20:
Logged In: YES 
user_id=330806
Originator: NO

Hi!

I add another example to the list:

% string match {*[.-]*} "2.1-Beta"
0
% string match {*[-.]*} "2.1-Beta"
1
% string match {*[.-]*} "Beta-2.1"
1
% string match {*[-.]*} "Beta-2.1"
1

If this is not going to be fixed soon, I suggest to at least document this behavior. Could be enough to just add the hyphen to the list of characters that need to be escaped...

kind regards
-- Matthias Kraft

matzek added on 2007-06-14 21:00:55:
Logged In: YES 
user_id=330806
Originator: NO

I forgot to mention that the example below is taken from a recent Tcl interpreter (8.4.14)...

dkf added on 2007-06-14 21:50:12:
Logged In: YES 
user_id=79902
Originator: NO

Comments below indicate "wait until 9.0", this being a strategy that was mainly from Jeff.

gneumann added on 2010-11-18 16:52:04:
i was hit by apparently the same problem.

% string match {[-.]} -
1
% string match {[.-]} -
0 
# even worse
% string match {[.-]} A
1 

This is certainly unexpected behavior and should be at least documented.

dgp added on 2018-05-07 16:19:56:

"Wait until 9.0" is no longer a blockage.

This should be fixed at least on the trunk, if not
on earlier branches too.

dgp added on 2018-05-07 16:21:29:
If someone wants to pursue a "refactor first, then fix" strategy,
I can accept that.

sebres added on 2018-05-07 22:03:22:
If the behavior like below is acceptable for you, I can do a back-porting from my own branches (I assume it's relative easy):

  % string match {[.-B} A
  invalid match pattern: brackets [] not balanced

Otherwise, I would like to know how exactly it must be then "fixed" (after the possible refactoring)...

E.g. another variant of the "fix" can be:

  % string match {[.-B} A
  0
  % string match {[.-B} \[.-A
  0
  % string match {[.-B} \[.-B
  1

Quasi by unbalanced brackets the pattern {[.-B} will be equivalent to the {\[.-B}

Anyway, current implementation is definitely wrong:
  % string match {[.-B} A
  1
should result either to an error or to 0.

serhiy.storchaka added on 2026-10-19 17:43:27:

Still present in 8.6, 9.0 and 9.1. Proposed fix for 9.1 in branch glob-bracket-posix, on top of glob-match-linear (the matcher change for [a4a7f2abd3]): in all four matchers, a - right before the closing bracket is a literal, as in POSIX and Python's fnmatch. So {[.-]} matches only "." and "-", and {*[.-]*} matches "2.1-Beta". Test string-11.57; util-5.40, 5.41, 5.43 and 5.44 tested the old behaviour and are changed. Unbalanced brackets ({[.-B}) are not changed.


dgp added on 2026-10-19 17:56:34:
Alert that if this change is accepted it will require bold
labels of

POTENTIAL INCOMPATIBILITY

in the release notes.

I'm unsettled by changing the functioning of RE processing
in order to "fix" a "bug", most especially one that has survived
for 25 years and more.

Would help to have more people examine the impact.