Layer-loader is a Python library designed to allow easy composition of configuration files.
This project was inspired by hiera, though intends to be somewhat simpler.
The primary interface is a Python API:
import json
import layer_loader
data = layer_loader.load_files(
['dev.json', 'main.json'],
loader=json.load,
)
This will load the data from dev.json
and main.json
, interpreting it as JSON,
combine it into a single object and then any placeholders are expanded.
For convenience you can also run layer_loader
as a module:
$ cat dev.json
{
"url": "http://localhost:8000"
}
$ cat main.json
{
"endpoints": [
"{url}/abc",
"{url}/def",
"{url}/ghi"
],
"url": "https://production.system"
}
$ python -m layer_loader --loader=json.load dev.json main.json
{
"endpoints": [
"http://localhost:8000/abc",
"http://localhost:8000/def",
"http://localhost:8000/ghi"
],
"url": "http://localhost:8000"
}
Layers are specified in order from most important (top) to least important (bottom). When combining layers, layer-loader uses the following rules:
- Mappings can be merged, combining the entries from several layers
- Scalar values and lists replace entries in the lower layers
- Keys in more important layers override those in lower layers
- Where keys are present in multiple layers, it is an error for those
keys to have values of different types. The exception to this is
that keys can explicitly be set to
null
in a higher layer in order to remove a value set in a lower layer.
Placeholders are specified similarly to python format strings and may appear
anywhere inside a string value, for example 'foo {place.holder} bar'
.
Placeholders should not appear in keys -- they will not be found there.
Placeholders are expanded by walking the flattened data structure, using dots in the placeholder to separate names at each level of nesting.
There is no mechanism to specify which layer a placeholder should be looked up within -- all placeholder expansion happens after the layers have been flattened.