In order to avoid the overhead associated with the shelve module’s writeback option I’m interested in putting together a shelf class that only accepts hashable values, with hashability being a proxy for immutability. So I’d like to subclass shelve.Shelf and override the __set__ method. The catch is that shelve.open can return one of a number of different classes (e.g., Dbfilenameshelf), and I’d like my code to allow shelve this flexibility.
Ideally a solution would have the following properties:
- No need to wrap every function that
Shelfprovides, as in:def keys(self): return self._shelf.keys() - Will not break if Python adds new methods to
Shelfor new subclasses ofShelf - Avoiding fragility, by which I mean doing something complicated that could easily trip up someone (conceivably me) who is modifying or using the code months or years down the road. An example in this context would be reassigning
Shelf.__set__, as inShelf.__set__ = my_func. To me defining a new__getattr__as @Winston hesitantly offers seems moderately fragile–I take it that’s why he suggests it only with hesitation
Less important but still desirable would be:
- Consistent with object-oriented design principles. It seems to me that the new class here is a shelf, not that it has a shelf. And probably not both (it’s not a shelf which contains another shelf)
2
Don’t subclass shelve.Self, instead define a new class with a constructor that that calls shelve.open. Implement the method of interest by calling the actual methods on the returned shelf object.
2