 |  |
| |
A Unix program should use its name as the first word in error messages it prints. That's important when the program is running in the background or as part of a pipeline -- you need to know which program has the problem:
someprog: quitting: can't read file xxxxxx
It's tempting to use just the program name in the echo commands:
echo "someprog: quitting: can't read file $file" 1>&2
If you ever change the program name, however, it's easy to forget to fix the messages. A better way is to store the program name in a shell variable at the top of the script file and use the variable in all messages:
myname=someprog ... echo "$myname: quitting: can't read file $file" 1>&2
Even better, use the $0 parameter. The shell automatically puts the script's name there. But $0 can have the absolute pathname of the script, such as /xxx/yyy/bin/someprog. The basename (Section 36.13) program fixes this: basename strips off the head of a pathname -- everything but the filename.
For example, if $0 is /u/ehuser/bin/sendit:
myname="`basename $0`"
would put sendit into the myname shell variable.
Just as you can make links (Section 10.3) to give Unix files several names, you can use links to give your program several names (Section 36.8). For instance, see the script named ll, lf, lg (...and so on). Use $0 to get the current name of the program.
-- JP
|  | |
| 35.27. Picking a Name for a New Command |  | 35.29. Reading Files with the . and source Commands |
Copyright © 2003 O'Reilly & Associates. All rights reserved.