On Nov 26, 2:04 pm, Bruno Desthuilliers <[EMAIL PROTECTED]> wrote: > Donn Ingle a écrit : > > >>I see someone already showed you eval. Eval is evil. Don't use it. > >>Especially if the functions are coming to you from a public URL! > > > Yes, I suggested to him (by email) this: > > > thisinstance = SomeObject.__class__.__dict__ > > <Then you have a list of strings that may be function names, so:> > > This will only get attributes defined in SomeObject.__class__ - not the > one defined in the parent classes. Nor methods dynamically bound to a > specific instance. > > > > > for f in yourlist: > > if f in thisinstance: eval(f)(params) > > > Which would vet the functions too. > > You *still* don't need eval here. > > target = <module or any other objet here> > for funcname in funclist: > func = getattr(target, funcname, None) > if callable(func): > func(*args, **kwargs) > > I've almost never had a real use case for eval (or exec) in 7 years.
That's not to say there aren't any. In one project I'm using exec to parse user provided function bodies from a web interface into real python functions (the function signatures are auto generated). Of course it's a secure environment, with trusted users, etc. Although Python's current support for code generation (source <-> ast <-> bytecode roundtrip) is far from perfect, it's much easier and more powerful than building from scratch a domain specific toy language, which might well turn out to be less expressive than necessary for the given problem domain. I guess Lisp/Scheme would be even more suited for this task but then again there's a Web framework (TurboGears), an ORM (SqlAlchemy) an RPC middleware (Pyro) and a dozen more batteries, both standard and 3rd party. Can't think of anything better than Python for this project. George -- http://mail.python.org/mailman/listinfo/python-list
