.. default-role:: literal =============================== Configuration handling basics =============================== `Config` and `Option` objects ================================ Accessing the configuration `Option`'s ----------------------------------------- The `Config` object attribute access notation stands for the value of the configuration's `Option`. That is, the `Config`'s object attribute is the name of the `Option`, and the value is the value accessed by the `__getattr__` attribute access mechanism. If the attribute of the `Config` called by `__getattr__` has not been set before (by the classic `__setattr__` mechanism), the default value of the `Option` object is returned, and if no `Option` has been declared in the `OptionDescription` (that is the schema of the configuration), an `AttributeError` is raised. :: >>> gcdummy = BoolOption('dummy', 'dummy', default=False) >>> gcdummy._name 'dummy' >>> gcdummy.getdefault() False >>> descr = OptionDescription('tiramisu', '', [gcdummy]) >>> cfg = Config(descr) >>> cfg.dummy False >>> cfg.dummy = True >>> cfg.dummy True >>> cfg.idontexist AttributeError: 'OptionDescription' object has no attribute 'idontexist' The configuration `Option` objects (in this case the `BoolOption`), are organized into a tree into nested `OptionDescription` objects. Every option has a name, as does every option group. The parts of the full name of the option are separated by dots: e.g. ``config.optgroup.optname``. Let's make the protocol of accessing a config's attribute explicit (because explicit is better than implicit): 1. If the option has not been declared, an `AttributeError` is raised, 2. If an option is declared, but neither a value nor a default value has been set, the returned value is `None`, 3. If an option is declared and a default value has been set, but no value has been set, the returned value is the default value of the option, 4. If an option is declared, and a value has been set, the returned value is the value of the option. What if a value has been set and `None` is to be returned again ? Don't worry, an option value can be "reseted" with the help of the `option.Option.reset()` method. If you do not want to use the pythonic way, that is the attribute access way to obtain the value of the configuration option, you can also search for it recursively in the whole config namespaces with the ``get()`` method : :: >>> config.get('bool') True To find the right option, `get()` searches recursively into the whole tree. For example, to find an option which is in the `gc` namespace there are two possibilites. If you know the path: :: >>> config.gc.dummy False If you don't remember the path: :: >>> config.get('dummy') False Setting the values of the options ---------------------------------------- An important part of the setting of the configuration consists of setting the values of the configuration options. There are different ways of setting values, the first one is of course the `__setattr__` method :: cfg.name = value wich has the same effect that the "global" `set()` method (it expects that the value owner is the default) :: cfg.set(name=value) .. module:: tiramisu.config The handling of options is split into two parts: the description of which options are available, what their possible values and defaults are and how they are organized into a tree. A specific choice of options is bundled into a configuration object which has a reference to its option description (and therefore makes sure that the configuration values adhere to the option description). The configuration object important methods --------------------------------------------- `config.Config()` object that lives in `config.py` hold the choosen values for the options (or the default value for the `option.Option()` object, if no choice was made). A `Config` object is informed by an `option.OptionDescription` instance. The attributes of the ``Config`` objects are the names of the children of the ``OptionDescription``. Here are the (useful) methods on ``Config``: .. currentmodule:: tiramisu.config .. autoclass:: Config .. automethod:: __init__ .. rubric:: Methods .. autosummary:: ~Config.__init__ ~Config.set .. automethod:: set We can see that values can also be config objects, it's the sub-namespaces that are stored in the values as `Config()` objects. convenience utilities (iteration, exports...) ----------------------------------------------- With this `config.Config()` configuration management entry point, it is possible to - `iter` on config, notice that there is an iteration order wich is the order of the :ref:`optdescr` specification entries, - compare two configs (equality), - export the whole config into a `dict` with `config.make_dict()`, `option.Option()` objects in a config are iterable in the pythonic way, that is something like `[(name, value) for name, value in config]`. To iter on groups in the same manner, use the `config.Config.iter_groups()` method wich yields generators too. **iteration utilities** .. autoclass:: Config .. automethod:: __init__ .. rubric:: Methods .. autosummary:: ~Config.get ~Config.find ~Config.find_first ~Config.getpaths ~Config.iter_groups ~Config.__iter__ .. automethod:: get .. automethod:: find .. automethod:: find_first .. automethod:: getpaths .. automethod:: iter_groups .. automethod:: __iter__