[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: command line arguments'
- To: firstname.lastname@example.org
- Subject: Re: command line arguments'
- From: Joachim Schrod <email@example.com>
- Date: Thu, 7 Jul 1994 18:57:26 +0100 (MESZ)
- In-reply-to: <9407071427.AA26881@ma2s2.mathematik.uni-karlsruhe.de> from "Bruno Haible" at Jul 7, 94 04:29:50 pm
> Joachim Schrod <firstname.lastname@example.org> asks:
> > Are the command line arguments available in CLISP?
> No, because command line arguments are not reasonably portable across
> all platforms supported by CLISP. (DOS and Atari TOS limit the command
> line length to 128 and 80 characters, respectively, making it unusable
> for "real" applications.) Furthermore, passing quotes and double quotes
> within command line arguments is always a pain and depends on the command
> line interpreter.
Hmm, let me try to convince you... (Get me right: I don't ask you to
do it -- I'll schedule Gabor for that. ;-) But I would like to see
these changes folded back into the main development thread.)
CLISP is implemented in C. C has a defined way to pass command line
arguments, namely via the first two arguments of main(), canonically
called argc and argv. (Actually, my question was simply ``is argv
available as the binding of a CL symbol'' -- obviously, it's not.)
I cannot follow your arguments against providing these values.
(1) That the command line length in certain OSes is restricted is
something that holds for _all_ programs there. This is an
restriction that users of these deficient environments got used
to. (They didn't have any other chance.) And I wouldn't call most
applications I use "unreal applications" -- but they don't need
command line lengths > 80.
Actually, _I_ would call applications with such long command
lines `unusable'. (A synonym to `real'? :-) :-)
(2) How to pass quotes and double quotes, is IMO not a problem of
CLISP. It's -- as you write -- a problem of the command line
interpreter and the C run time library.
Those who use a certain CLI must know how to specify quotes in
it; just like they have to do it in all other programs.
I hope my background wish gots clearer: I want to use Lisp
applications just like any other applications. I want to tell people
`Here is a CLISP image, just use it.' Without having to resort to
shell scripts (written for each OS anew!) where I pass some arguments
via setq etc. Besides here comes exactly the point where your length
argument bites: I don't want to write long -x option arguments in a
DOS .BAT file when I just want to pass a document name!
C'mon, let's go out of the Lisp universe and meet other systems...
(Dick Gabriel has written a nice memo about this problem. ;-)
As an example, my current pet: An SGML style sheet language, realized
as an extension of CLOS; that may be used to transform or query SGML
documents. The cmd line argument is the document(s).
Since I'm at it: I would also like some hook that is evaluated after
a memory image is loaded. (Like \everyjob, for the TeX freaks...
Actually, I want it exactly for this -- I'm working on an executable
formal specification of TeX.) But one can imagine more uses for it,
e.g., reestablishing a link to some display server, etc.
(I want to have this evaluation _before_ the first form is read,
therefore I cannot use *evalhook*.)
> > Actually I would like to have some symbol (say, in package SYSTEM)
> > that is bound to a list of strings; the arguments that are not already
> > processed by CLISP itself.
> This would be dangerous since we don't know which command line options
> will be added to CLISP in the future.
How about two symbols: system::*raw-argv* and system::*argv* ?
Or a stop token for CLISP options? `--' comes immediately to mind...
Joachim Schrod Email: email@example.com
Computer Science Department
Technical University of Darmstadt, Germany