Make STDIN executable with shebang

Question

I have a program to list database files.

It is called direkly from the shell like

db filename

to list the whole file, or like

db 'filename :: conditions'

to list only selected elements ...

---

Another way is to call it with a file, wich contains all parameters.

db < parameterfile

The content is like (quite the same as the content in '' above):

filename
::
conditions

---

Now I woud like do make such a file executable. So that i can call just ./parameterfile.

To use a shebang #!/usr/bin/env db failed, because # is not a comment sign, I think. I got the error message

db - Line 1 near ""#.//r" - " - syntax error

shell returned 26

Is there a one-liner to do this?

Answer 1

Use awk as a shebang wrapper:

#! /usr/bin/awk BEGIN{getline;cmd="db"}{print|cmd}END{exit(close(cmd))}

... "db" commands here ...

If your script is not supposed to be run directly via execve(2) but only via the shell (or via commands like find -exec) you have a lot more options, including making it executable without prepending it a shebang, and keep it as a valid command script for whatever language that db command understands. It all depends on what that language is. For instance, if comment lines in that language start with // (as in C++), you can use:

// 2>/dev/null; exec db "$0" "$@"

... "db" commands here ...

Answer 2

You could write a shell script that removes the shebang line, if any, and passes the result to db. Place this script in a directory of your PATH. Otherwise you might have to specify the full path of the script in the shebang line. Use this script as the interpreter for your parameterfile.

Example script runparam to remove the first line if it is a shebang line:

#!/bin/sh
awk 'FNR>1 || ! /^#!/' "$@" | db

Example parameterfile:

#!/usr/bin/env runparam
filename
::
conditions

You can run it as

./parameterfile

In this case the script can assume that there will always be a shebang line.

You could also call the script directly with a parameterfile in the same way as you would call db, but this has no advantage.

runparam parameterfile
runparam parameterfile1 parameterfile2 [...]
runparam < parameterfile

If you will never call runparam with a file that doesn't contain a shebang line, you can use tail instead of awk to unconditionally remove the first line. This might be faster.

#!/bin/sh
tail -n+2 | db

Answer 3

Just make it a shell script. Either something like:

exec db << 'EOF'
filename
::
conditions
EOF

(depending on the shell, that either puts the contents of that here document into a tempfile which is made the stdin of the process that will run db and deleted before db is executed, or a pipe with a shell process feeding the data through it).

Or:

tail -n+2 < "$0" | db
filename
::
conditions

Here with a pipe (or possibly socketpair with ksh93) between tail and db.

Or:

exec < "$0"; IFS= read -r ignore; exec db
filename
::
conditions

Here with db's stdin being directly opened on the script, with the first line skipped with read. That would be the most efficient.