Skip to content

ashishrv/yaycl

 
 

Repository files navigation

YAYCL: YAML Configuration loader and cache

All YAML files stored in the given directory are automatically parsed and loaded on request. The parsed files are exposed as object attributes based on the yaml file name.

For example, consider a conf/config.yaml file:

dict_key:
    key1: value1
    key2: value2
    key3: value3
list_key:
    - 1
    - 2
    - 3
string_key: 'string value'

Now, you can interact with that yaml in python with minimal fuss:

import yaycl
conf = yaycl.Config('conf')
# assuming config.yaml is valid yaml, this should work:
assert 'key2' in conf.config.dict_key

Once loaded, the yaml contents are cached. The entire cache of a config object can be cleared, or a single config file's cache can be cleared:

conf.clear()  # clears the entire cache
conf.config.clear() or conf['config'].clear()  # clears the cache only for config.yaml

Note that, as in the example above, yaml files loaded by yaycl (currently) must be a mapping type at the top level. Files containing more than one yaml document are (currently) unsupported.

.yaml vs. .yml

Many projects use the .yml file extension for YAML files. This is supported by passing the extension keyword argument to yaycl.Config. For this example, assume conf/config.yaml has been renamed to conf/config.yml:

import yaycl
conf = yaycl.Config('conf', extension='.yml')
# Now this config will be loaded from conf/config.yml
assert 'key2' in conf.config.dict_key

Module Impersonation

yaycl.Config is indended to manage config files for an entire project. To facilitate that goal, it supports acting as a module, making configurations importable.

The module's name doesn't matter, as long as it can be imported by that name.

In this example, we'll make a module called conf.py, with contents:

import sys

from yaycl import Config

sys.modules[__name__] = Config('/path/to/yaml/config/dir')

Now, the first time conf is imported, it will replace itself in conf with an instance of yaycl.Config, which will be what python imports thereafter. Once done, you can import config files directly. Here's the same example from before, but using the direct import method:

from conf import config
assert 'key2' in conf.config.dict_key

For brevity, following examples will use the module impersonation mechanism.

Shenanigans

Special care has been taken to ensure that all objects are mutated, rather than replaced, so all names will reference the same config object.

All objects representing config files (attributes or items accessed directly from the conf module) will be some type of AttrDict. Attempting to make such a config object be anything other than an AttrDict (see "Inherited methods section below) will probably break everything and should not be attempted, lest shenanigans be called.

# Don't do this...
conf.key = 'not a dict'
# ...or this.
conf['key'] = ['also', 'not', 'a', 'dict']

Generally speaking, with the exception of runtime overrides (see below), a yaycl.Config instance should be considered read-only.

Local Configuration Overrides

In addition to loading YAML files, the yacl.Config loader supports local override files. This feature is useful for maintaining a shared set of config files for a team, while still allowing for local configuration.

Take the following example YAML file, config.local.yaml:

string_key: 'new string value'

When loaded by the conf loader, the string_key will be automatically overridden by the value in the local YAML file::

from conf import config
print config.string_key

This will print: new string value, instead of the value in the base config, string value

The existing keys (dict_key and list_key in this case) will not altered by the local config override.

This allows for configurations to be stored in revision control, while still making it trivial to test new configs, override an existing config, or even create configs that only exist locally.

# .gitignore suggestion; adapt to your SCM of choice
*.local.yaml

Runtime Overrides

Sometimes writing to the config files is an inconvenient way to ensure that runtime changes persist through configuration cache clearing. These "runtime" changes can be stashed in the runtime overrides dict, allowing them to persist through a cache clear.

The runtime overrides dictionary mimics the layout of the conf module itself, where configuration file names are keys in the runtime overrides dictionary. So, for example, to update the base_url in a way that will persist clearing of the cache, the following will work:

import conf
conf.runtime['config']['string_key'] = 'overridden string key'
print conf.config.string_key

That should print overridden string key

runtime.yaml

If you have a config file named 'runtime.yaml' that you'd like to load, or really any config name that interferes with python names ('get.yaml', for example), note that the configs are always available via dictionary lookup; attribute lookup is supported for brevity, but dict item lookup should always work.

conf.runtime['runtime'] = {'shenanigans': True}
assert conf['runtime']['shenanigans']

Inherited methods

Once loaded, all configs are instances of AttrDict, a very helpful class from the layered-yaml-attrdict-config package. As such, all methods normally available to AttrDicts are available here.

For example, Config.save and Config's inheritance abilities rely on AttrDict's dump and rebase methods, respectively.

Of course, since AttrDict is a dict subclass, dictionary methods can also be used to manipulate a yaycl.Config at runtime. The clear method is particularly useful as a means to trigger a reload of all config files by clearing yaycl's cache.

Thread safety

No care whatsoever has been taken to ensure thread-safety, so if you're doing threaded things with the conf module you should manage your own locking when making any conf changes. Since most config are loaded from the filesystem, generally this means that any changes to the runtime overrides should be done under a lock.

Coverage Status Build Status

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 99.2%
  • Makefile 0.8%