lazily loading and overriding configuration for the lazy coder
Project description
lazyConfig
lazily loading and overriding configuration
lazyConfig is an opinionated configuration provider. It can load .yml
, .yaml
,
.json
and .toml
configuration files out of the box but can be extended with
custom loaders. Opinionated since it requires you to
structure your configuration in a certain way in order to work.
Why lazyConfig?
see Motivation
Usage
In order to load your configuration folder into a dict
like structure, simply
point lazyConfig
to the folder
import lazyConfig
config = lazyConfig.from_path(
config='/path/to/default/config',
override=[
'/path/to/override',
'/path/to/another/override/',
...
]
)
# or with environment variables:
os.environ['CONFIG'] = '/path/to/default/config'
os.environ['CONFIG_OVERRIDE'] = f'/path/to/override/{os.pathsep}/path/to/another/override/'
config = lazyConfig.from_env()
You can change the envrionment variable names by passing their names, e.g.
lazyConfig.from_env(config='DEFAULT_CONFIG', override='CONFIG')
Files are overriden left to right (i.e. the last argument has priority)
you could mix and match using
Config.from_path
andos.environ['ENV_VAR']
Add override
If you are using an command line interface parser (e.g. argparse) to generate a dictionary for overriding configuration, you might want to add an additional override. You can use
config.add_override(argparse_results)
where argparse_results
is of type Mapping
use
vars()
to obtain adict
from anarparse.Namespace
which is the return type of an arparse parser
arparse uses None
to indicate missing configuration. So add_override
ignores
None
by default. If you provide the parameter none_can_override=True
, you can
remove configuration with None
values.
Override Restrictions
You can only override keys which exist in the default configuration. This
requirement ensures that the default configuration documents all possible
configuration options. Use the value None
in the default configuration for
settings you do not want to select a default for. You can obtain a dict without
these entries with
config.as_dict(strip_none=True)
strip_none=True
is the default, set it to false if you want a dict including allNone
values
Assumptions about the file structure
Filenames are used as keys in the config
dict, so without any caveats
you would be forced to have a minimum depth of 1
for any configuration.
This might not make sense for some high level, flat configuration. For this
reason there is a special filename __config__
(.yml
,.yaml
, .json
,...),
which allows you to define top level keys and values.
Example
config
database
configuration.yml
__config__.yml
app.yml
__config__.yml
config/__config__.yml
:
name: my-app
author: ME!
version: -1.0
app.yml
:
primary_color: 'blue'
secondary_color: 'green'
config/database/__config__.yml
:
connection:
hosts:
- {host: "myElasticsearchServer", port: 9200}
timeout: 6000
config/database/configuration.yml
:
indices:
index1: {...}
index2: {...}
pipelines:
pipeline1: {...}
would be loaded as the following dictionary (formatted in yml
for readability):
name: my-app
author: ME!
version: -1.0
app:
primary_color: 'blue'
secondary_color: 'green'
database:
connection:
hosts:
- {host: "myElasticsearchServer", port: 9200}
timeout: 6000
configuration:
indices:
index1: {...}
index2: {...}
pipelines:
pipeline1: {...}
Attribute access
You could then access index1
with:
config.database.configuration.indices.index1
Python does not allow attributes to start with a number. So
config.1attribute2breakthings
will not work. In these cases you will have to use
config['1attribute2breakthings']
Duplicate Keys
Duplicate Keys are not allowed. lazyConfig
tries to find keys in
- Keyfiles (
__config__
) - Filenames
- Directories
in that order! It will not keep looking whether or not there is a duplicate and thus ignore a directory if a file with the same name (sans extension) or a key with the same name in the keyfile exists.
There might be a configuration validation function in the future to check for duplicate keys (in debug mode or on manual call)
Lists
Lists are overriden not extended. If the default configuration has the same key for a list as an override, then the default list is ignored and the override is used.
Extending Lists would mean that you can not remove elements with a configuration override. As this is not desirable you will just have to duplicate lists in its entirety if you want to change them (even if only slightly)
Since numbers are discouraged anyway, you can define Directory Lists like this:
config
list
0.yml
1.yml
2.yml
3.yml
which is then converted to
config:
list:
- {content from 0.yml}
- {content from 1.yml}
- {content from 2.yml}
Directory Lists must start with 0
and end with n-1
where n
is the
number of files in the directory.
It is currently not possible to create a list of directories (instead of files). This might become a feature in a future version if requested
Security
Using pyYAML.unsafe_load()
, lazyConfig
is currently not meant for external data.
Pass a custom_extension_loader
to the factory method:
lazyConfig.from_path('path/to/config', custom_extension_loader={
'.yml': yaml.safe_load,
'.yaml': yaml.safe_load
})
This overrides the yaml loader while leaving the other loaders in place. To remove
a default loader, override the extension with a Falsy value (e.g. None
)
Future Features
things which I might get around to do some time:
- Configuration Validation
- Horizontal override (Templates, e.g. for indices)
- Configuration Setter
Since it is not obvious where to set directory/file boundaries inside the dictionary, this is not as trivial as it might seem at first. It might be necessary to provide the grouping level as an argument.
- generate JSON Schema from default configuration
Versions
- 0.5 proper none handling
- 0.4 add_overwrite
- 0.3 as_primitive, as_dict, as_list, laziness modes and custom loader, TOML support
- 0.2.2 fix equality
- 0.2.1 fix broken iterator
- 0.2 Config implements Mapping
- 0.1.1 handle NULL environment variables
- 0.1: basic proof of concept
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.