immerframe

Intuitively perform deep updates on python objects

• without mutating them
• while being efficient via structural sharing - no deepcopy
• while maintaining type correctness

This is (almost) a Python port of immer.

pip install immerframe

First, let's import some stuff

from dataclasses import dataclass
from immerframe import Proxy

Now, consider the data:

@dataclass
class Ant:
    age: int

ant_10 = Ant(age=10)
ant_20 = Ant(age=20)
nested = {
    "ants": [ant_10, ant_20, None],
}

let's pretend to mutate it

with Proxy(nested) as (p, new_nested):
    p["ants"][0].age += 1
    p["ants"].pop()
    p["foo"] = 99

(note p and new_nested should have the correct types in mypy)

nested will remain the same

assert nested == {
    "ants": [ant_10, ant_20, None],
}

new_nested will be nested, but with the mutations with specified in the with Proxy(...) block

assert new_nested == {
    "ants": [Ant(age=11), ant_20],
    "foo": 99,
}

but it won't be a deep copy

assert new_nested["ants"][1] is ant_20

immerframe supports things that can be copyed

Things to remember

Mutating the same thing in a block twice may not do what you'd expect in the following type of case:

with Proxy([1]) as (l, new_l):
    l[0] += 5
    l[0] += 10

will give new_l == [11]

Use keys rather than references to loop over and mutate nested dicts/lists:

with Proxy(new_nested) as (p, new_nested):
    for k in new_nested:
        p[k] += 1
    for i, n in enumerate(new_nested["ants"]):
        if n.age < 15:
            p["ants"][i].age + 10

TODO

• implement all dunder methods
• finish typing