Yappd is a persistent dictionary like object in Python based on jsonpickle and sqlite3. You store values in it like a normal dict and they are written to disk automatically. When you load it up again later, everything is like you left it.

It can store any object that's supported by jsonpickle ("almost any Python object"), including nested structures.

Since Yappd maintains a regular Python dict under the hood, it's only suitable for data sets that can be kept in RAM (whatever you'd normally use a dict for).

Just install with pip (into a venv of course):

pip install yappd

Then import and use PDict:

from yappd import PDict
p = PDict('mydata')
p['hello'] = 'world'

p = PDict('mydata') # Load it from disk again
p['hello'] == 'world' # True

Each PDict needs a file to write its data into, and the path to the file is the first argument to the constructor. An optional second argument is the dictionary name, which allows multiple dictionaries to be stored in the same file. By default the name default used:

p1 = PDict('mydata', 'default') # Equivalent to example above
p2 = PDict('mydata', 'p2') # This one uses p2 as the name so there's no conflict

Most dict functions are implemented and are usually just proxies to the underlying dict object.

There's a major caveat to keep in mind, which is mutable data types. Take this usage for example:

p = PDict('mydata')

p['mylist'] = []
p['mylist'].append(1)
p['mylist'] == [1] # True

p = PDict('mydata') # Reload the dict from disk
['mylist'] == [] # True. Change not stored to disk

At this point the value stored on disk for mylist is an empty list, but the value stored in the program memory is [1]. Since yappd can't detect if stored values are mutated, you must explicitly store any mutated values again to update them on disk.

The save function is provided for this purpose:

p = PDict('mydata')

p['mylist'] = []
p['mylist'].append(1)
p.save('mylist')
p['mylist'] == [1] # True

p = PDict('mydata') # Reload the dict from disk
p['mylist'] == [1] # True. This time it's okay

You can also call save with no arguments to save all values. It's more efficient though to specify which values to save.

Pickling objects in Python has some major downsides. With jsonpickle we get human readable output that can always be loaded, even if the original classes aren't around in the current scope.

But why not just store the resulting JSON to disk in a text file? Well, it turns out that reliably updating files on disk in an atomic way is not so easy. SQLite has this covered and is a no brainer choice for storing app data on disk.

That said, if you don't need SQL features, then adopting a database brings a lot of extra hassle (even with an ORM). The implementation of yappd means we can just shove arbitrary objects into the disk and get them back later. There's no need to perform a database migration if you decide later that every Pet object should have a favorite_food field.