#!/usr/bin/perl -p
[download]

You do know how the kernel deals with #!, don't you?

Abigail

I pondered this on the way home. I know you can use -p on the shebang line, but in that case, the question to ask would be what happens when you run a script that contains

     #! /usr/bin/perl -p
[download]

... and nothing else. Curiously enough, this does do the same thing as cat(1). Which only begs the question why does it work when -p is placed on the shebang line of a script, as opposed to needing -e 1 when used on a command line? The command line behaviour looks like a bug to me.


print@_{sort keys %_},$/if%_=split//,'= & *a?b:e\f/h^h!j+n,o@o;r$s-t%t#u'

Which only begs the question why does it work when -p is placed on the shebang line of a script, as opposed to needing -e 1 when used on a command line?

For the same reason you *never* need -e if you run the program from a file - either by using the shebang line, or by doing perl program.

Perl uses the following strategy to find the program:

1. If there are one or more -e arguments, that's where the program was found.
2. If the command was invoked as perl [options] file, the program text is found in the file.
3. Otherwise, the program text is found on STDIN.

You might say, what about the shebang line? The shebang line is handled by the kernel - the interpreter found on the shebang line is started, and the content of the file (including the shebang line) is passed on STDIN. (This also explains why such interpreters always use # as the start of the comment).

Abigail

..to which some informative and historical references would be good:

Some details about the shebang mechanism on various Unix flavours

shebang from FOLDOC

"env won't accept arguments in shebang"

..of which brings up an interesting point:

"..the #!interpreter syntax may only take one. Everything after the interpreter is passed as a single argument..."

_____________________________________________________
Jeff[japhy]Pinyan: Perl, regex, and perl hacker, who'd like a job (NYC-area)
s++=END;++y(;-P)}y js++=;shajsj<++y(p-q)}?print:??;

A shebang with just -p would then be cat because the remainder of the file is valid perl (type perl on the command line, you're interactive mode. type C-d or F6 etc. (platform dependent) et voila. fini. no complaints)

--
perl -pew "s/\b;([mnst])/'$1/g"

At first, I was thinking I could support "perl -p" (not in the #! line) giving a helpful message since it is expecting both its script and its input from STDIN. But I immediately thought of using a "TTY" as STDIN such that you can have two "ends of file" on the one STDIN:

shell> stty eof ^Z
shell> perl -p >>notes
BEGIN { print ''.localtime(),$/ } $_= "$.: $_"
CTRL-Z
Here are my notes for today.
I should really make this a real script.
CTRL-Z
[download]

But then I realized you can use __END__ to do this trick anywhere:

shell> cat script
$_= "$.: $_";
__END__
Testing
this.
shell> perl -p <script
1: Testing
2: this.
[download]

Granted, I haven't come up with a great use of this feature, but I'm reluctant to promote disabling it merely based on my lack of imagination. (:

I often add cases of test data after __END__ or __DATA__ in scripts that use while(<>) loops --- and I used to always switch between <> and <DATA> for testing purposes (and usually forgot to switch back when done). I recall being quite happy a few years back when I realized ("Doh!") I could just leave the bare diamond op and run the script on the test data via  perl -w < script.pl. Actually, I mostly test within the editor by sending the buffer to a filter of  perl -w with output directed to another buffer, but its the same thing.

-- stefp -- check out TeXmacs wiki

Actually, I've used -p on the shebang to good effect. It can make simple scripts a lot simpler.

buckaduck