Convert glob notation to regular expression

Here is a function, globtore (Glob to Regular Expression), that takes the "wildcard" syntax used by the 4DOS command shell and converts it to a Perl regular expression string.

Basically, * matches anything and ? matches one character. But there are a few special cases, and 4DOS has more, such as ranges and OR's. I figure for anything that's not really basic I'll use a RegExp anyway, so I don't want to make too fancy of a globber. But the docs for 4DOS was fairly complete, so I used that as my design spec. The test data below includes the examples from the 4DOS help page.

The last_paren_match_ordinal function is another story.

Edit: chipmunk 2001-06-27

use strict;
use warnings;

sub last_paren_match_ordinal()
 {
 my $n= $#-;
 return $n;
 
 # if the above stops working, comment out the return and let the foll
+owing code
 # run.  It seems that @- is populated only up to the actual number of
+ captures,
 # while @+ always contains the max number present in the pattern.  Th
+is is not
 # documented, and may be an artifact.
 
 while ($n) {
    # take $n initially as a maximum, and find the highest that's actu
+ally present.
    no strict 'refs';
    last if defined $$n;
    --$n;
    }
 return $n;    
 }
 
sub globtore ($)
# initial version, 27-June-2001 by JMD
 {
 my $s= shift;

 my $f= sub {
    my $n= last_paren_match_ordinal;
    return "($n)";
    };
 my @f= (
    undef,  # index 0 never used, uses 1..6
    sub {  # brackets
       my $s= $+;
       return '.'  if $s eq '[?]';  # special case meaning.
       return '(?=\.|;|\z)'  if $s eq '[]' || $s eq '[!?]';  # also sp
+ecial: match "no character present".
       # shell uses !, change to ^.  Don't worry about anything else. 
+ Could deal with original ^ in that spot.
       $s =~ s/^\[!/[^/;
       return $s;
       },
    sub {   # ?'s before dot or end
       my $count= length $+;
       return '.?'  if $count == 1;
       return ".{0,$count}";
       },
    sub {   # ?'s normal case
       my $count= length $+;
       return '.'  if $count == 1;
       return ".{$count}";
       },
    sub {   # *'s
       return '.*'
       },
    sub {   # ;
       return '$|^';
       },
    sub {   # other funny chars that aren't otherwise matched.
       my $s= $+;
       return "\\$s";
       }
    );
    
 # smash out consecutive stars
 $s =~ s/\*+/*/g;
 # do the main replacements
 $s =~ s/
       (\[.*?\]) | # brackets in $1
       (\?+(?=\.|;|\z)) |  # ?'s before dot or end in $2
       (\?+) |   # ?'s normally in $3
       (\*+) |   # *'s in $4
       (;) |  # ; seperator in $5
       ([\$\^\\\.\{\}\[\]])  # special otherwise unmatched chars in $6
    /$f[last_paren_match_ordinal]->()  # see above functions for massi
+ve replacement block.
    /gex;
 # fix up begin/end marks
 $s= "^$s\$";
 $s =~ s/\Q^.*\E|\Q.*\E\$//;
 return $s;
 }

while (<DATA>) {
   chomp;
   s/\s+#.*$//;  # allow comments in input data
   print "$_ ==> ";
   my $result= globtore ($_);
   print "$result\n";
   }

__DATA__
LETTER?.DOC  # comments allowed in test data, separated by at least on
+e space.
funny#file#name.txt  # the prev # are not comments because no space.
*.DO?
file{braces}.^^x
*.DO[?]
foobar.exe
xxy
*.exe
l?tter?.d??
*.*
st*.d*
*am*.txt
letter[0-9].doc
?[aeiouy]*.*
[a-dt-v]ip
letter[?].doc
letter[].doc
letter[!?].doc
test[!0-9].doc   # anything except digits.
??[abc]*[def]*.[pq]*
letter1;v2
letter1[;]v2
[download]

Comment on Convert glob notation to regular expression Select or Download Code

Replies are listed 'Best First'.
(tye)Re: Convert glob notation to regular expression by tye (Sage) on Jun 28, 2001 at 11:03 UTC
See also my File::KGlob from years ago which includes such functionality. (: - tye (but my friends call me "Tye")	[reply]
My specifications by John M. Dlugosz (Monsignor) on Jun 28, 2001 at 02:54 UTC
From the JP Software documentation for 4NT 3.02, text by Hardin Brothers, Tom Rawson, and Rex Conn, Wildcards Wildcards let you specify a file or group of files by typing a partial + filename. The appropriate directory is scanned to find all of the f +iles that match the partial name you have specified. Wildcards are usually used to specify which files should be processed +by a command. If you need to specify which files should not be proce +ssed see File Exclusion Ranges (for internal commands), or EXCEPT (fo +r external commands). Most internal commands accept filenames with wildcards anywhere that a + full filename can be used. There are two wildcard characters, the a +sterisk [] and the question mark [?], plus a special method of speci +fying a range of permissible characters. An asterisk [] in a filename means "any zero or more characters in th +is position." For example, this command will display a list of all f +iles in the current directory: [c:\] dir . If you want to see all of the files with a .TXT extension, you could t +ype this: [c:\] dir .txt If you know that the file you are looking for has a base name that beg +ins with ST and an extension that begins with .D , you can find it this way. Filenames such as STATE.DAT, STEVEN.DOC, +and ST.D will all be displayed: [c:\] dir st.d* With 4NT, you can also use the asterisk to match filenames with specif +ic letters somewhere inside the name. The following example will dis +play any file with a .TXT extension that has the letters AM together +anywhere inside its base name. It will, for example, display AMPLE.T +XT, STAMP.TXT, CLAM.TXT, and AM.TXT: [c:\] dir am.txt A question mark [?] matches any single filename character. You can pu +t the question mark anywhere in a filename and use as many question m +arks as you need. The following example will display files with name +s like LETTER.DOC and LATTER.DAT, and LITTER.DU: [c:\] dir l?tter.d?? The use of an asterisk wildcard before other characters, and of the ch +aracter ranges discussed below, are enhancements to the standard wild +card syntax, and may not work properly with software other than 4DOS, + 4OS2, 4NT, and Take Command. "Extra" question marks in your wildcard specification are ignored if t +he file name is shorter than the wildcard specification. For example +, if you have files called LETTER.DOC, LETTER1.DOC, and LETTERA.DOC, +this command will display all three names: [c:\] dir letter?.doc The file LETTER.DOC is included in the display because the "extra" que +stion mark at the end of "LETTER? " is ignored when matching the shor +ter name LETTER. In some cases, the question mark wildcard may be too general. You can + also specify what characters you want to accept (or exclude) in a pa +rticular position in the filename by using square brackets. Inside th +e brackets, you can put the individual acceptable characters or range +s of characters. For example, if you wanted to match LETTER0.DOC thr +ough LETTER9.DOC, you could use this command: [c:\] dir letter[0-9].doc You could find all files that have a vowel as the second letter in the +ir name this way. This example also demonstrates how to mix the wild +card characters: [c:\] dir ?[aeiouy]. You can exclude a group of characters or a range of characters by usin +g an exclamation mark [!] as the first character inside the brackets. + This example displays all filenames that are at least 2 characters +long except those which have a vowel as the second letter in their na +mes: [c:\] dir ?[!aeiouy]. The next example, which selects files such as AIP, BIP, and TIP but no +t NIP, demonstrates how you can use multiple ranges inside the bracke +ts. It will accept a file that begins with an A, B, C, D, T, U, or V +: [c:\] dir [a-dt-v]ip You may use a question mark character inside the brackets, but its mea +ning is slightly different than a normal (unbracketed) question mark +wildcard. A normal question mark wildcard matches any character, but + will be ignored when matching a name shorter than the wildcard speci +fication, as described above. A question mark inside brackets will m +atch any character, but will not be discarded when matching shorter f +ilenames. For example: [c:\] dir letter[?].doc will display LETTER1.DOC and LETTERA.DOC, but not LETTER.DOC. A pair of brackets with no characters between them [], or an exclamati +on point and question mark together [!?],will match only if there is +no character in that position. For example, [c:\] dir letter[].doc will not display LETTER1.DOC or LETTERA.DOC, but will display LETTER.D +OC. This is most useful for commands like [c:\] dir /I"[]" .btm which will display a list of all .BTM files which don't have a descrip +tion, because the empty brackets match only an empty description stri +ng (DIR /I selects files to display based on their descriptions). You can repeat any of the wildcard characters in any combination you d +esire within a single file name. For example, the following command +lists all files which have an A, B, or C as the third character, foll +owed by zero or more additional characters, followed by a D, E, or F, + followed optionally by some additional characters, and with an exten +sion beginning with P or Q. You probably won't need to do anything t +his complex, but we've included it to show you the flexibility of ext +ended wildcards: [c:\] dir ??[abc][def].[pq] You can also use the square bracket wildcard syntax to work around a c +onflict between long filenames containing semicolons [;], and the use + of a semicolon to indicate an include list. For example, if you have + a file named C:\DATA\LETTER1;V2 and you enter this command: [c:\] del \data\letter1;v2 you will not get the results you expect. Instead of deleting the name +d file, 4NT will attempt to delete LETTER1 and then V2, because the s +emicolon indicates an include list. However if you use square bracke +ts around the semicolon it will be interpreted as a filename characte +r, and not as an include list separator. For example, this command w +ould delete the file named above: [c:\] del \data\letter1[;]v2 Extra caution should be taken using wildcards on long file names becau +se operations using wildcards will be performed on both long and shor +t filenames. See LFN File Searches for additional details. [download]	[reply] [d/l]