Return

2.3  Matching Multiple Files Using Wildcards

Problem

You want to interactively select groups of files to be processed based on file naming conventions.

Solution

Use the shell's File Name Substitution characters, which are similar to those used in DCL.

Discussion

The 'wildcard' pattern matching characters in Unix are very similar to those used in VMS, but there are distinct and sometimes subtle differences. This can be a problem because you many not match the files you intended or match other files that you did not want to include.

First, you need to be aware of some important differences between VMS filenaming and Unix:

• Unlike VMS any character can be used in a filename. This includes characters that are special to the shell, like the asterisk (* and even the carriage return, tab, or line feed characters! The manual page for the Bourne Shell advises that the characters *, ?, [, or ] not be used in file or directory names, but the operating system does nothing to inforce these guidelines, so such files can be created accidentally or intentially.

• The period (.) is a special character in filenames only if it is the first character. Unlike VMS, the file extension in the filename:

      myprog.f

  is not a separate and distinct field in the file's name. The period is just another character. Thus in Unix filenames like:

      this.is.my.kewl.file

  are perfectly legal.

• When the first character of a filename is a period, it indicates to the shell that this is a 'hidden' file. By default such files will not shown in directory listings, nor matched by wildcard patterns unless the leading period is explicitly specified.

With these differences in mind, consider the following pattern matching characters:

Do not confuse the shell's special pattern matching characters with 'regular expression' pattern matching in programs like grep, sed, and perl. For example, in a regular expression, * matches zero or more of the preceeding character. It does not by itself match characters the way as it would when used in a shell expression.

Again, because the period is not a special character (except when it is first) it will be matched by the ? and * wildcards. Given the following existing filenames in the current directory:

    .foo-bar
    foo-bar
    foo.bar
    foo.foo
    foosbar
    very.foo.bar

unix> ls foo?bar
foo-bar  foo.bar  foosbar
unix> ls foo*
foo-bar  foo.bar  foo.foo  foosbar
unix> ls *foo*
foo-bar       foo.bar       foo.foo       foosbar       very.foo.bar
unix> ls .*foo*
.foo-bar
unix>

Here is an example of using the [...] pattern matching construct:

unix> ls file.~[0-9]~
file.~1~  file.~3~  file.~5~  file.~7~  file.~9~
file.~2~  file.~4~  file.~6~  file.~8~
unix> ls file.~[0-9][0-9]~
file.~10~
unix> ls file.~[0-9]*~
file.~101~  file.~1~    file.~3~    file.~5~    file.~7~    file.~9~
file.~10~   file.~2~    file.~4~    file.~6~    file.~8~
unix>

As an exercise for the reader consider the following given the files:

    myfile
    myfile.1
    myfile.2
    myfile.x

unix> ls myfile*
myfile    myfile.1  myfile.2  myfile.x
unix> ls myfile*[!x]
myfile.1  myfile.2
unix>

One last important note -- always remember that it is the shell that expands the wildcard characters before starting up the program. Use the echo command to easily illustrate this.

unix> echo myfile*
myfile    myfile.1  myfile.2  myfile.x
unix>

unix> echo no*such*file
no*such*file
unix>

See Also

2.4 - Finding Files;
Chapter 2 of Unix for OpenVMS Users ;
Chapter 1 of Unix Power Tools .

Return

Colophon:
This page maintained by:     Bill.Costa@unh.edu     of the Enterprise Computing Group     in the dept of Computing & Information Sevices     at the University of New Hampshire	Typographical Conventions Created:  31-Jan-2001 BC Revised:  27-Mar-2001 BC