I want to write a class with singleton pattern to provide some persistent data storage using pickle/dict:
@singleton
class Pdb:
def __init__(self):
self.cache = None
self.dirty = False
try:
with open("data.pck","rb") as fp:
self.cache = pickle.load(fp)
except FileNotFoundError:
pass
except pickle.PickleError:
pass
if self.cache is None:
self.cache = {}
def flush(self):
if self.dirty:
try:
with open("data.pck","wb") as fp:
pickle.dump(self.cache,fp,protocol=4)
except pickle.PickleError:
pass
else:
self.dirty = False
def __del__(self): # PROBLEM HERE
self.flush()
When I was using python 2, I can do it by overriding __del__. But it does not appear to be correct in python 3. How can I do it?
If I do it by "with" statement, I will need to pass the instance to each function that I call:
def func1(db):
db.set(...)
func3(db,x1,x2,...)
with Pdb() as db:
func1(db)
func2(db)
It is complicated. Is there a pythonic way to do a global scope "with" statement?
If I do it by "with" statement, I will need to pass the instance to each function that I call:
No, you don't. Just use your singleton:
# global
db = Pdb()
# any other context
with db:
All that is required is that the expression produces a context manager. Referencing a singleton object with __enter__ and __exit__ methods would satisfy that requirement. You can even ignore the __enter__ return value, as I did above. The global will still be available to all your functions, the only thing that changes is that __enter__ and __exit__ will be called at the appropriate locations.
Note that even in Python 2, you should not rely on __del__ being called. And in the CPython implementation, outside circular references, the rules for when __del__ are called have not changed between Python 2 and 3.