Extending zsh capabilities

Extending zsh capabilities

[Mark Hessling]

Hi Guys,

I have been using zsh for several years now and have had no complaints!

What I want to do now is add optional support for Rexx programs into a shell,
and figured that zsh would be the shell to use.

Firstly, would there be any objections to me adding this support and getting
it into the "official" distribution ?

Secondly, what assistance might I be able to get to point me in the right
direction ?  I originally tried adding this support to bash, but couldn't
get it to work; probably because I didn't know enough of the internals of
bash to know where the right place was to hook the Rexx support into.

I've had a look at Zsh modules in 3.1.3, but don't think that this mechanism
will work. Maybe it will, but that's where I need some knowledgeble advice.

Before I go too much further, I'll explain what I want to do and someone
hopefully can provide some pointers.

All Rexx program begin with /* in the first two characters of the first line
of an executable file (much like #!).  When zsh comes across a Rexx program,
I want to execute this program within the shell, but unlike #! programs,
execute within the current process.  So I can't use the #! hook.

Executing a Rexx program within a current process is straighforward, I need
to call the API function RexxStart() with the filename as one of the arguments.
The interface is simple; its knowing where to hook it into zsh that is my
biggest problem.

Any comments, suggestions or pointers greatly appreciated.

Cheers, Mark
------------------------------------------------------------------------
 Mark Hessling                       Email:       M.Hessling@qut.edu.au
 PO Box 203                          http://www.lightlink.com/hessling/
 Bellara                                AUTHOR of  |  MAINTAINER of
 QLD 4507                                 THE      |    PDCurses
 Australia                              Rexx/SQL   |     Regina
                Member of RexxLA: http://www.rexxla.org/
------------------------------------------------------------------------

Re: Extending zsh capabilities

[Bruce Stephens]

Mark Hessling <m.hessling@qut.edu.au> writes:

> Executing a Rexx program within a current process is straighforward,
> I need to call the API function RexxStart() with the filename as one
> of the arguments.  The interface is simple; its knowing where to
> hook it into zsh that is my biggest problem.

[caveat: I don't know much about zsh internals]

A first thought is to look at where the shell executes external
commands, and fiddle there.  But as you comment, that's in the wrong
process.  There're presumably lots of subtleties to do with io
redirection, interruptability and things, so looking at builtins is
probably better.  

How about, as a first cut, writing a builtin "rexx" which would
execute a Rexx script?  That could be done as a module (like stat).

I'd have thought that would be acceptable in usage too: you'd just
create aliases for Rexx scripts you used a lot.  If not, you could
change the unknown function so it would try executing an unknown
command as a Rexx script, or work a bit harder.

Probably Zefram has better ideas; I'm sure he was threatening to do
the same for Perl.

Re: Extending zsh capabilities

[Peter Stephenson]

Bruce Stephens wrote:
> > Executing a Rexx program within a current process is straighforward,
> > I need to call the API function RexxStart() with the filename as one
> > of the arguments.  The interface is simple; its knowing where to
> > hook it into zsh that is my biggest problem.
> 
> How about, as a first cut, writing a builtin "rexx" which would
> execute a Rexx script?  That could be done as a module (like stat).

I would have thought that was a good place to start; adding additional
capabilities in modules is very much preferred from the development point
of view, and makes it much more likely the code would be acceptable in
the main distribution.  This should be very easy:  just take e.g. stat.c,
rename everything, and prune it right down so that you just have one
function bound to the builtin rexx which calls RexxStart on its filename.

What you really then want is for a shell function beginning /* to run Rexx.
You can do this by sleight of hand.

% fpath=(. $fpath)		# just for convenience in this example
% alias '/*'='rexx_load $0'
% cat rexx_load
local file=$1 dir
for dir in $fpath; do
    [[ -f $dir/$file ]] && file=$dir/$file && break
done

print "Executing $file using Rexx"

# rexx $file

# Now force the calling function to return immediately.
trap 'return 0' EXIT
% cat rtest
/* This is to test rexx_load */
Rexx commands here
% autoload rexx_load rtest
% rtest
Executing ./rtest using Rexx


This works because zsh allows you to alias '/*' which presumably isn't
going to appear as a command anywhere else.  The other tricky bit is
the trap in rexx_load which forces its parent (in this case the function
rtest running under zsh, which you don't want to continue) to return
immediately, since the EXIT trap is executed in the surrounding context.

You could extend rexx_load so that it redefines rtest to run rexx on
a series of commands (if e.g. rexx has a -e switch like perl), eliminating
the intermediate step for subsequent calls.

-- 
Peter Stephenson <pws@ifh.de>       Tel: +39 50 844536
WWW:  http://www.ifh.de/~pws/
Gruppo Teorico, Dipartimento di Fisica
Piazza Torricelli 2, 56100 Pisa, Italy

Re: Extending zsh capabilities

[Mark Hessling]

On Mon, 27 Jul 1998, Peter Stephenson wrote:

> Bruce Stephens wrote:
> > > Executing a Rexx program within a current process is straighforward,
> > > I need to call the API function RexxStart() with the filename as one
> > > of the arguments.  The interface is simple; its knowing where to
> > > hook it into zsh that is my biggest problem.
> >
> > How about, as a first cut, writing a builtin "rexx" which would
> > execute a Rexx script?  That could be done as a module (like stat).
>
> I would have thought that was a good place to start; adding additional
> capabilities in modules is very much preferred from the development point
> of view, and makes it much more likely the code would be acceptable in
> the main distribution.  This should be very easy:  just take e.g. stat.c,
> rename everything, and prune it right down so that you just have one
> function bound to the builtin rexx which calls RexxStart on its filename.
>
> What you really then want is for a shell function beginning /* to run Rexx.
> You can do this by sleight of hand.
>
> % fpath=(. $fpath)		# just for convenience in this example
> % alias '/*'='rexx_load $0'
> % cat rexx_load
> local file=$1 dir
> for dir in $fpath; do
>     [[ -f $dir/$file ]] && file=$dir/$file && break
> done
>
> print "Executing $file using Rexx"
>
> # rexx $file
>
> # Now force the calling function to return immediately.
> trap 'return 0' EXIT
> % cat rtest
> /* This is to test rexx_load */
> Rexx commands here
> % autoload rexx_load rtest
> % rtest
> Executing ./rtest using Rexx
>
>
> This works because zsh allows you to alias '/*' which presumably isn't
> going to appear as a command anywhere else.  The other tricky bit is
> the trap in rexx_load which forces its parent (in this case the function
> rtest running under zsh, which you don't want to continue) to return
> immediately, since the EXIT trap is executed in the surrounding context.
>
> You could extend rexx_load so that it redefines rtest to run rexx on
> a series of commands (if e.g. rexx has a -e switch like perl), eliminating
> the intermediate step for subsequent calls.
>
> --
> Peter Stephenson <pws@ifh.de>       Tel: +39 50 844536
> WWW:  http://www.ifh.de/~pws/
> Gruppo Teorico, Dipartimento di Fisica
> Piazza Torricelli 2, 56100 Pisa, Italy
>
Thanks guys. I started to go this way, but got stumped early on.

I downloaded zsh 3.1.3 and am attempting to build it. Unfortunately I
am having problems :-(  I have tried building zsh under HP-UX 9.04 and
Solaris 2.5 and am getting the same errors.  I've even tried replacing
the HP-UX sed with the latest GNU sed, thinking it was a problem with
sed, but that didn't help.

If anyone can help with the following problems, please let me know.
TIA.

................ output from ./configure --enable-dynamic .............

checking whether symbols in the executable are available... (cached) yes
checking whether executables can be stripped... (cached) yes
checking whether libraries can be stripped... (cached) yes
creating ./config.status
creating Makefile
Can't open :Makefile.in
creating Doc/Makefile
Can't open :Doc/Makefile.in
creating Etc/Makefile
Can't open :Etc/Makefile.in
creating Src/Makefile
Can't open :Src/Makefile.in
creating config.h
:config.h.in
cat: cannot open :config.h.in
config.h is unchanged

zsh configuration
-----------------
zsh version               : 3.1.3
...................................................................

With my very limited REGEXP knowledge I made a couple of changes and
got the Makefiles and config.h created.  Then, this is what I got on
the first attempt at make:

$ make
if test D = N; then \
    cat ./xmods.conf > modules-bltin; \
elif test yes != yes; then \
    echo comp1 > modules-bltin; \
else \
    echo > modules-bltin; \
fi
( cd .. && /bin/sh Src/mkmodindex.sh Src ) \
    > modules.index.tmp
Updated `modules.index'.
cd .. && /bin/sh $top_srcdir/Src/mkmakemod.sh Src Makemod
creating Src/Makemod.in
creating Src/Makemod
Can't open !Src/Makemod.in
make: Fatal error: Don't know how to make target `prep'
Current working directory /home/solaris/mark/zsh-3.1.3/Src
make: Fatal error: Can't find `Makemod': No such file or directory
Current working directory /home/solaris/mark/zsh-3.1.3/Src
*** Error code 1
make: Fatal error: Command failed for target `headers'
Current working directory /home/solaris/mark/zsh-3.1.3/Src
*** Error code 1
make: Fatal error: Command failed for target `all'

Cheers, Mark
------------------------------------------------------------------------
 Mark Hessling                       Email:       M.Hessling@qut.edu.au
 PO Box 203                          http://www.lightlink.com/hessling/
 Bellara                                AUTHOR of  |  MAINTAINER of
 QLD 4507                                 THE      |    PDCurses
 Australia                              Rexx/SQL   |     Regina
                Member of RexxLA: http://www.rexxla.org/
------------------------------------------------------------------------

Re: Extending zsh capabilities

[Zefram]

Mark Hessling wrote:
>I downloaded zsh 3.1.3 and am attempting to build it. Unfortunately I
>am having problems

There were some bugs in the build process in 3.1.3.  These are fixed
in 3.1.4.

-zefram

Re: Extending zsh capabilities

[Peter Stephenson]

Mark Hessling wrote:
> I downloaded zsh 3.1.3 and am attempting to build it. Unfortunately I
> am having problems :-(

These are intrinsic to 3.1.3, unless you are running a sh which is
actually zsh in disguise.  3.1.4 should be around on the official archives
and has this fixed.

I realised my trick of aliasing /* probably doesn't work for complex
Rexx scripts, because zsh tries to parse the whole thing first and runs
into incompatible syntax.  I then thought of doing something with
preexec, but I now think that's too limited.  Of course, if you know in
advance that something is a Rexx script, there are all sorts of things you
can do, but I can't offhand think of a really good way when you don't.
(Even my /* way implies you had already sorted Rexx scripts into function
directories, and if you do that, you might as well have a special directory
and use some code in .zshrc to alias foo="rexx $rexxpath/foo"
or whatever.)

-- 
Peter Stephenson <pws@ifh.de>       Tel: +39 50 844536
WWW:  http://www.ifh.de/~pws/
Gruppo Teorico, Dipartimento di Fisica
Piazza Torricelli 2, 56100 Pisa, Italy