Embedding Python Design

Question

There are lots of tutorials/instructions on how to embed python in an application, but nothing (that I've seen) on overall design for how the embedded interpreter should be used and interact with the application.

The only idea I could think of would be to simply give the user a method (menu option, etc) of executing scripts in the program. So certain classes, functions, objects, etc. would be exported to python, some script would do something, then said script could be run from the program.

Would such a design be "safe?" Meaning is it feasible for a malicious/poorly-written script to "damage" the program and/or computer? I assume its possible depending on the functions available to the script (e.g: it could try to overwrite some important files, etc.) How might one prevent such from happening? (e.g: script certification, program design, etc.)

This is implementation specific, but is it possible/feasible to have the effects of the script stay after its done running? Meaning if a script computes something, will the result be available to the program after execution of the script has finished? I think it is possible to do if the program were setup to interact with a specific script, but the program will be released before most scripts are written; and such a setup seems like a misuse of embedding a scripting language. Is there actually cases where you would want the result of a scripts execution to be available, or is this a contrived situation that doesn't really occur?

Are there any other designs for embedding python?
What about using python in a way similar to a plugin architecture?

Thanks,
Matthew A. Todd

Answer 1

The only idea I could think of would be to simply give the user a method (menu option, etc) of executing scripts in the program.

Correct.

So certain classes, functions, objects, etc. would be exported to python, some script would do something, then said script could be run from the program.

Correct.

Would such a design be "safe?"

Yes. Unless your users are malicious, psychotic sociopaths. They want to make your program do useful things. They bought/downloaded the software in the first place. They think it has value.

They trusted your software. Why not trust them?

Meaning if a script computes something, will the result be available to the program after execution of the script has finished?

Programs like Apache do this all the time. You screw up the configuration ("script"), it crashes. Lesson learned? Don't screw up the configuration.