pattern matching question

Shell Programming and Scripting
1

Hi guys
I have the following case statement in my script:

case $pn.$db in
   *?.fcp?(db))  set f ${pn} cp   ;;
   *?.oxa?(oxa) )  set oxa $pn  ;;
esac

Can somebody help me to understand how to interpret *?.fcp?(db)) or *?.oxa?(oxa) ?

I cannot figure out how in this case pattern maching works.
Thanks a lot for any advice .

2

You seem to be using the Korn shell (or a shell that supports the ksh extended globbing - bash (with shopt -s extglob ) or zsh ( setopt kshglob ).

This is the meaning of these patterns:

*?.fcp?(db)

* - matches any string, including the null string
? - matches any character
. - matches a literal dot
fcp - matches the literal string fcp
?(db) - matches 0 or 1 occurrences of the subpattern, the string db in this case.

Consider the following:

$ cat s
case $pn.$db in
   *?.fcp?(db)  ) printf '%s in the first case\n' "$pn.$db matched" ;;
   *?.oxa?(oxa) ) printf '%s in the second case\n' "$pn.$db matched" ;;
esac

$ pn=any_ db=fcp
$ . s
any_.fcp matched in the first case
$ pn=any_ db=fcpdb
$ . s
any_.fcpdb matched in the first case
$ pn=any_ db=fcpdbdb
$ . s
$
$ pn=any_ db=oxa
$ . s
any_.oxa matched in the second case

I hope this helps.

3

First, lets take this as an example of why comments are necessary!

Ok in the case statement, the patterns to match follow the shell's file name generation patterns (*=any string, ?=any character, ?(patternlist)). So:

*?.fcp?(db))

(the *? means the first part must have at least 1 character)
means: if the pattern we are analyzing in the case ($pn.$db)
starts with at least 1 character and
ends in "fcp" or "fcpdb" then the pattern is matched.

If I am correct, this could be written in an easier to
follow way. Here's an example script using ksh that accepts 2 args which are then analyzed
by the case statement:

#!/bin/ksh

pn=$1
db=$2

# original code
case $pn.$db in
   *?.fcp?(db))  print 1
                 set f ${pn} cp   ;;
   *?.oxa?(oxa) )  print 2
                   set oxa $pn  ;;
esac

# Hopefully clearer code that shows the author's intentions.
case $pn.$db in
     *?.fcp) ;&  #  Fall through
   *?.fcpdb) print "first case"
            set f ${pn} cp
            ;;
     *?.oxa) ;&  #  Fall through
  *?.oxaoxa) print "Second case"
            set oxa $pn
            ;;
         *) # Always allow for the unexpected!
            print "Unknown value [$pn.$db]"
            exit 1
            ;;
esac

exit 0

Note the

     *.oxa) ;&  #  Fall through
  *.oxaoxa) print "Second case"

could be written as

*.oxa|*.oxaoxa)

Using a logical OR, but I am in the habit of splitting them up on their own lines in case in the future the developer maintaining this needs to take a different action on each case. The framework is already set up then.

Also, always allow for the default case which will catch unexpected values passed in!

4

Depends on what needs to be matched. The * matches the null string too,
so the two patterns * and *? are different (the latter requires at least one character):

$ cat s
case $pn.$db in
   *?.fcp?(db)  ) printf '%s in the first case\n' "$pn.$db matched" ;;
   *.fcp?(db)  ) printf '%s in the second case\n' "$pn.$db matched" ;;
esac

$ pn= db=fcp
$ . s
.fcp matched in the second case
$ pn=a db=fcp
$ . s
a.fcp matched in the first case
1 Like
5

Yes, it appears the case pattern means its pattern must start with at least 1 character. My example should have the ? too as in the original.
I will update my example. Thanks!

6

Thanks a lot guys. It really helped me a lot.