Skip to main content

A recursive subclass of ChainMap

Project description

DeepChainMap

PyPI codecov pre-commit.ci status Code style: black

A recursive subclass of collections.ChainMap.

Installation

pip install deep-chainmap

Usage

The canonical use case for collections.ChainMap is to aggregate configuration data from layered mapping (basically dictionaries) sources. However, it is not suited for non-flat (nested) mappings, since the lookup mechanism only works for the top level of a mapping.

deep_chainmap.DeepChainMap provides a simple solution to this problem by making recurive lookups in arbitrarily deeply nested mappings. Let's illustrate this with a simple example. We will simulate 3 layers of mapping, and pretend they were obtained from different sources (a default configuration, a configuration file and parameters configured at runtime).

from deep_chainmap import DeepChainMap

default_layer = {
    "architecture": "gpu",
    "logging_level": "warning",
    "solver": "RK4",
    "database": {
        "url": "unset",
        "keep_in_sync": False,
    },
    "mesh": {
        "type": "rectangular",
        "resolution": {
            "x": {
                "npoints": 100,
                "spacing": "linear",
            },
            "y": {
                "npoints": 100,
                "spacing": "linear",
            },
            "z": {
                "npoints": 100,
                "spacing": "linear",
            },
        },
    },
}

config_file_layer = {
    "architecture": "cpu",
    "mesh": {
        "resolution": {
            "x": {
                "spacing": "log",
            },
            "z": {
                "npoints": 1,
            },
        },
    },
}

runtime_layer = {
    "logging_level": "debug",
    "database": {
        "url": "https://my.database.api",
        "keep_in_sync": True
    },
}

# now building a DeepChainMap
cm = DeepChainMap(runtime_layer, config_file_layer, default_layer)

Now when a single parameter is requested, it is looked up in each layer until a value is found, by order of insertion. Here the runtime_layer takes priority over the config_file_layer, which in turns takes priority over the default_layer.

>>> cm["logging_level"]
'debug'
>>> cm["mesh"]["resolution"]["x"]["spacing"]
'log'
>>> cm["mesh"]["resolution"]["x"]["npoints"]
100

Note that submappings at any level can be retrieved as new DeepChainMap instances

>>> cm["mesh"]
DeepChainMap({'resolution': {'x': {'spacing': 'log'}, 'z': {'npoints': 1}}},
             {'resolution': {'x': {'npoints': 100, 'spacing': 'linear'},
                             'y': {'npoints': 100, 'spacing': 'linear'},
                             'z': {'npoints': 100, 'spacing': 'linear'}},
              'type': 'rectangular'})

The other important feature is the to_dict method, which constructs a builtin dict from a DeepChainMap

>>> cm.to_dict()
{
    'architecture': 'cpu',
    'logging_level': 'debug',
    'solver': 'RK4',
    'database': {
        'url': 'https://my.database.api',
        'keep_in_sync': True
    },
    'mesh': {
        'type': 'rectangular',
        'resolution': {
            'x': {'npoints': 100, 'spacing': 'log'},
            'y': {'npoints': 100, 'spacing': 'linear'},
            'z': {'npoints': 1, 'spacing': 'linear'}
        }
    }
}

An important implication is that the DeepChainMap class enables a very simple, functional implementation of a depth-first dict-merge algorithm as

from deep_chainmap import DeepChainMap

def depth_first_merge(*mappings) -> dict:
    return DeepChainMap(*mappings).to_dict()

Limitations

As the standard collections.ChainMap class, DeepChainMap does not, by design, perform any kind of data validation. Rather, it is assumed that the input mappings are similar in structure, meaning that a key which maps to a dict in one of the input mappings is assumed to map to dict instances as well in every other input mapping. Use the excellent schema library or similar projects for this task.

:warning: An important difference with collections.ChainMap is that, when setting a (key, value) pair in a DeepChainMap instance, the new value is stored in the first mapping which already contains the parent map. For example if we run

>>> cm["mesh"]["resolution"]["x"]["spacing"] = "exp"

The affected layer is config_file_layer rather than runtime_layer, as one can see

>>> config_file_layer
{
    'architecture': 'cpu',
    'mesh': {
        'resolution': {
            'x': {'spacing': 'exp'},
            'z': {'npoints': 1}
        }
    }
}
>>> runtime_layer
{
    'logging_level': 'debug',
    'database': {
        'url': 'https://my.database.api',
        'keep_in_sync': True
    }
}

This behaviour is a side effect on an implementation detail and subject to change in a future version. Please do not rely on it.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

deep_chainmap-0.1.1.tar.gz (5.1 kB view hashes)

Uploaded Source

Built Distribution

deep_chainmap-0.1.1-py3-none-any.whl (4.6 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page