Multiconf is a framework for describing a complex configuration for multiple environments using Python.
It started from a simple need of deployment automation for Java EE projects, Apache and more. Having worked on different projects with nested levels of plain text property files or XML configuration files,I thought something better was needed. With plain text property files, the number of property files increases as environments and technologies are added to a project. It becomes hard to get an overview of properties describing similar configurations. Has a property value been defined for every environment? And it is getting even harder to describe proper settings: what depends on what and what can be used and what can't. With XML on the other hand, you can create a strict validated model, but you keep having to extend the schema and the tools processing it. And maybe you don't like the verbosity. So why use XML or property files when you can have your configuration directly in python? So, out of this Multiconf was born.
E.g:
- All configured ports follow one convention
- All servers names follow one convention
- Some configuration objects must have mandatory parameters (for example: Database name or URL required for Datasource object)
- Some configuration objects must have mandatory children (for example: WebLogic Cluster doesn't make sense w/o Managed Servers)
- Default settings are propagated through all environments and can be overridden for specific environments
- No duplicated settings
Multiconf provides a set of classes, where attributes may have different values for different environments, while enforcing that a value is defined for all defined environments. Multiconf allows you to implement your own DOM like object model and get early warning that something within your definition is wrong. Other tools use YAML or JSON to define settings of the components, but then you need something to validate those settings. Multiconf is both - definition and validation. Multiconf allows you to define environment groups, so that you can easily create new environments by adding them to a group and only override the values that differ from the group values.
You have to define your configuration data model as classes derived from
Multiconf base classes, one of which is ConfigItem.
E.g, in your config data model (your framework) you define:
class Host(ConfigItem):
def __init__(name=MC_REQUIRED, mem=MC_REQUIRED):
self.name = name
self.mem = mem
@property
def fqd(self):
return "{name}.{env}.my.organisation".format(
self.name, self.env.name)In you project configuration file you can then declare a configuration object with different attribute values for different environments:
...
with Host("web1") as host:
host.setattr('mem', dev="1G", tst="2G", preprod="4G", prod="4G")Above uses the Multiconf setattr method to assign different values to different
envs. Note that the envs dev, tst, preprod and prod must have been declared
beforehand and Multiconf will ensure that all of them get a value.
After instantiating your config for the prod env you can then access properties on the host object:
cfg.host.name -> web1 cfg.host.mem -> 4G cfg.host.fqd -> web1.prod.my.organisation
Note that classes derived from the Multiconf classes (e.g: ConfigItem) do not
allow on the fly creation of attributes. Configuration items are not meant for
general programming, but for strictly validated configurations.
See the documentation and the demo project for details about nested objects, repeatable objects, instantiation, environment definitions, environment groups, default values and other details.
- Multiconf is not tied to configuration of any particular product or technology.
- Multiconf doesn't know how to create any of the environment's components, i.e. Multiconf has no 'playbooks' or 'recipes' to execute.
Execute ./demo/demo.py --env (or 'python demo/demo.py ...'), e.g:
./demo/demo.py --env prod
If run without any arguments it will print a usage message The valid environments are those specified at the top of demo/config.py
Execute: make, py.test or tox Running 'make' will execute the test suite, the demo and build the documentation.
Multiconf: Python 3.6.1+ Test Suite: pytest, pytest-cov (for older Python versions use multiconf 8.x) demjson (optional) - pip install -U pytest pytest-cov demjson
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent