Choixe¶

Beside the XConfig class, Choixe functionalities can also be accessed with the new Choixe class. You can construct a Choixe by passing anything you want to its constructor.

from pipelime.choixe import Choixe

# A dictionary containing stuff
data = {
    "alpha": {
        "a": 10,
        "b": -2,
    },
    "beta": {
        "a": "hello",
        "b": "world",
    },
}

# A Choixe instance
chx = Choixe(data)

# Parse into an AST
ast = chx.parse()

# Decode into a plain python object
obj = chx.decode()

# Walk the AST
nodes = chx.walk()

# Process into something else
result = chx.process()

# Process with branching
results = chx.process_all()

# Inspect, looking for variables and imports
inspection = chx.inspect()

# Get what's inside
data_2 = chx.data
assert data is data_2

# File I/O
chx = Choixe.from_file("path/to/file.json")
chx.save_to("path/to/file.yml")

But why?¶

Don’t we have already XConfig? Yes, but Choixe is a bit different, it does less, but nobody really needed all that XConfig, but it’s also more lightweight and more maintainable.

Here is a comparison table:

Ok, let’s review in detail why Choixe is better than XConfig:

Being a python Mapping¶

The fact that XConfig is a python Mapping is a double-edged sword. On one hand, it’s great because you automatically get all the methods you would expect from a plain python dict. On the other hand, what if you want to use XConfig with something that is not a dict, like a Sequence? You can’t. And what if the result of your processing is something else than a dict? You have to use special methods like unsafe_xxx to achieve that, and that’s not very pretty.

Furthermore, beside being a Mapping, XConfig is also a Box, which carries a lot of extra methods and hidden behaviors that make it extremely hard to extend. What if we want to create an XConfig that is a Sequence? That’s pretty hard to do.

Dot-Notation access¶

There is a big problem with dot-notation access in Box-like classes: you will need to clutter your code with # type: ignores everywhere. That’s because the dot-notation access is not a standard python feature, and it’s not supported by most common type checkers and language servers. So, you reduce the amount of characters by 4 per access (from my_obj.a to my_obj['a']), but you have to manually hint the returned type and/or ignore potential errors. That’s not very useful.

Deep access¶

Deep access is cool, but its very coupled with the fact that XConfig implements a specific data structure. So, we decided to remove it from the new Choixe class. Besides, you can still perform deep access by using the walk method, or by simply using pydash functionalities.

Schema validation¶

Nobody really uses choixe schema validation, because that feature is usually implemented in a more robust way in the application itself, or by means of external json-schema validators. So, we decided to remove it from the new Choixe class.