This sounds like a security acident waiting to happen. I wouldn't want to be the one to have to ensure there were no security holes in the surrounding code that might allow the script to get rewritten. Adrian... -- The difficulty of tactical maneuvering consists in turning the devious into the direct, and misfortune into gain. - Sun Tzu ----- Original Message ----- From: "Jerome Alet" <alet@unice.fr> To: <zope-dev@zope.org> Cc: <alet@unice.fr> Sent: Friday, March 08, 2002 7:27 PM Subject: [Zope-dev] RFC : AnyScripts
Hi,
I'm finally toying, after having stalled a too long time, with the creation of ZShellScripts, which would be for ZShell commands what PythonScripts are for Python...
The current, not yet available, incarnation is very highly based on PythonScripts, however I wondered if something better couldn't be done, because it seems to me to be an unneeded repetition of most of the PythonScripts' code :
AnyScripts
AnyScripts would have a mandatory first line similar to the one of any Unix shell script. By unix shell script I mean any script which can be launched directly under Unix, not only bash or csh scripts, but also perl, tcl, etc...
So an AnyScript would contain a first line like :
#! /scripts/interpreter
And the execution of this script in the context on which it is applied would be delegated to another object present in the path which was set on this very first line, i.e. /scripts/interpreter
So you could put any interpreter object of your choice in for example the /scripts folder (no path would mean search the interpreter object within the acquisition path)
For example PythonScripts would contain something like :
#! /builtins/python
or something like that.
Out of my mind I can already think of a ZShell interpreter object, but why not other types of interpreter objects :
perl tcl C (the language ! Why not after all) etc...
Each interpreter object would have at least two methods :
compile() => returns a code object from source code, if meaningful (otherwise returns the source unchanged), to be stored in the AnyScript object which calls the compile() method.
execute() => executes the above code object. This method of the Interpreter object would be called by the AnyScript object at run time, passing it the code object obtained with compile()
Of course all of this come straight out of my mind, and there would probably be some problems I'm not able to understand yet...
Bound variables would be like for PythonScripts, with the exception of their content would be the best format for the Interpreter in question, so each Interpreter object would probably also have a bind() method which would bind vars in the preferred format for the AnyScript object which calls this method, and would be accessible using the preferred syntax of the langage in question, e.g. :
for PythonScripts, vars like context would be Zope Objects, like they currently are.
for ZShell AnyScripts, vars like context, would be paths to Zope Objects, and accessible with names like $context for example.
for C AnyScripts, vars would probably be something like PyObjects pointers.
etc...
So AnyScripts objects would interact with their interpreter object of choice to compile their source code, bind their vars, then finally execute the code and get the result back.
What do you all think about this idea ?
Are any of you interested ?
I admit I don't know anything about the feasability of such a project with other languages than ZShell, however if a single implementation could be done, then anyone could add his own interpreter later...
PS : I'll do ZShellScripts anyway...
bye,
Jerome Alet
_______________________________________________ Zope-Dev maillist - Zope-Dev@zope.org http://lists.zope.org/mailman/listinfo/zope-dev ** No cross posts or HTML encoding! ** (Related lists - http://lists.zope.org/mailman/listinfo/zope-announce http://lists.zope.org/mailman/listinfo/zope )