82. config — A general yet simple configuration class.

(C) 2005, 2019 Benedict Verhegghe
Distributed under the GNU GPL version 3 or later
I wrote this simple class because I wanted to use Python expressions in my configuration files. This is so much more fun than using .INI style config files. While there are some other Python config modules available on the web, I couldn’t find one that suited my needs and my taste: either they are intended for more complex configuration needs than mine, or they do not work with the simple Python syntax I expected.
Our Config class is just a normal Python dictionary which can hold anything. Fields can be accessed either as dictionary lookup (config[‘foo’]) or as object attributes (config.foo). The class provides a function for reading the dictionary from a flat text (multiline string or file). I will always use the word ‘file’ hereafter, because that is what you usually will read the configuration from. Your configuration file can have named sections. Sections are stored as other Python dicts inside the top Config dictionary. The current version is limited to one level of sectioning.

82.1. Classes defined in module config

class config.Config(data={}, default=<function Dict.returnNone>)[source]

A configuration class allowing Python expressions in the input.

The Config class is a subclass of the Dict mapping, which provides access to items by either dict-style key lookup config['foo'], attribute syntax config.foo or call syntax config('foo'). Furthermore, the Dict class allows a default_factory function to lookup in another Dict. This allows a chain of Config objects: session_prefs -> user_prefs -> factory_defaults.

The Config class is different from its parent Dict class in two ways:

  • Keys should only be strings that are valid Python variable names, except that they can also contain a ‘/’ character. The ‘/’ splits up the key in two parts: the first part becomes a key in the Config, and its value is a dict where the values are stored with the second part of the key. This allows for creating sections in the configuration. Currently only one level of sectioning is allowed (keys can obnly have a single ‘/’ character.
  • A Config instance can be initialized or updated with a text in Python source style. This provides an easy way to store configurations in files with Python style. The Config class also provides a way to export its contents to such a Python source text: updated configurations can thus be written back to a configuration file. See Notes below for details.
  • data (dict or multiline string, optional) – Data to initialize the Config. If a dict, all keys should follow the rules for valid config keys formulated above. If a multiline string, it should be an executable Python source text, with the limitations and exceptions outlined in the Notes below.
  • default (Config object, optional) – If provided, this object will be used as default lookup for missing keys.


The configuration object can be initialized from a dict or from a multiline string. Using a dict is obvious: one only has to obey the restriction that keys should be valid Python variable names.

The format of the multiline config text is described hereafter. This is also the format in which config files are written and can be loaded.

All config lines should have the format: key = value, where key is a string and value is a Python expression The first ‘=’ character on the line is the delimiter between key and value. Blanks around both the key and the value are stripped. The value is then evaluated as a Python expression and stored in a variable with name specified by the key. This variable is available for use in subsequent configuration lines. It is an error to use a variable before it is defined. The key,value pair is also stored in the Config dictionary, unless the key starts with an underscore (‘_’): this provides for local variables.

Lines starting with ‘#’ are comments and are ignored, as are empty and blank lines. Lines ending with ‘’ are continued on the next line. A line starting with ‘[‘ starts a new section. A section is nothing more than a Python dictionary inside the Config dictionary. The section name is delimited by ‘[‘and ‘]’. All subsequent lines will be stored in the section dictionary instead of the toplevel dictionary.

All other lines are executed as Python statements. This allows e.g. for importing modules.

Whole dictionaries can be inserted at once in the Config with the update() function.

All defined variables while reading config files remain available for use in the config file statements, even over multiple calls to the read() function. Variables inserted with addSection() will not be available as individual variables though, but can be accessed as self['name'].

As far as the resulting Config contents is concerned, the following are equivalent:


There is an important difference though: the second line will make a variable key (with value ‘value’) available in subsequent Config read() method calls.


>>> C = Config('''# A simple config example
...     aa = 'bb'
...     bb = aa
...     [cc]
...     aa = 'aa'    # yes ! comments are allowed)
...     _n = 3       # local: will get stripped
...     rng = list(range(_n))
...     ''')
>>> C
Dict({'aa': 'bb', 'bb': 'bb', 'cc': Dict({'aa': 'aa', 'rng': [0, 1, 2]})})
>>> C['aa']
>>> C['cc']
Dict({'aa': 'aa', 'rng': [0, 1, 2]})
>>> C['cc/aa']

Create a new Config with default lookup in C

>>> D = Config(default=C)
>>> D
>>> D['aa']       # Get from C
>>> D['cc']       # Get from C
Dict({'aa': 'aa', 'rng': [0, 1, 2]})
>>> D['cc/aa']            # Get from C
>>> D.get('cc/aa','zorro')      # but get method does not cascade!

Setting values in D will store them in D while C remains unchanged.

>>> D['aa'] = 'wel'
>>> D['dd'] = 'hoe'
>>> D['cc/aa'] = 'ziedewel'
>>> D
Dict({'aa': 'wel', 'dd': 'hoe', 'cc': Dict({'aa': 'ziedewel'})})
>>> C
Dict({'aa': 'bb', 'bb': 'bb', 'cc': Dict({'aa': 'aa', 'rng': [0, 1, 2]})})
>>> D['cc/aa']
>>> D['cc']
Dict({'aa': 'ziedewel'})
>>> D['cc/rng']
[0, 1, 2]
>>> 'ee' in D
>>> 'cc/ee' in D
>>> D['cc/bb'] = 'ok'
>>> list(D.keys())
['aa', 'dd', 'cc', 'cc/aa', 'cc/bb']
>>> del D['aa']
>>> del D['cc/aa']
>>> list(D.keys())
['dd', 'cc', 'cc/bb']
>>> del D['cc']
>>> list(D.keys())
update(data={}, name=None, removeLocals=False)[source]

Add a dictionary to the Config object.

The data, if specified, should be a valid Python dict. If no name is specified, the data are added to the top dictionary and will become attributes. If a name is specified, the data are added to the named attribute, which should be a dictionary. If the name does not specify a dictionary, an empty one is created, deleting the existing attribute.

If a name is specified, but no data, the effect is to add a new empty dictionary (section) with that name.

If removeLocals is set, keys starting with ‘_’ are removed from the data before updating the dictionary and not included in the config. This behaviour can be changed by setting removeLocals to false.

load(filename, debug=False)[source]

Read a configuration from a file in Config format.

Parameters:filename (path_like) – Path of a text file in Config format.
Returns:Config – Returns the Config self, update with the settings read from the specified file.
read(txt, debug=False)[source]

Read a configuration from a file or text

txt is a sequence of strings. Any type that allows a loop like for line in txt: to iterate over its text lines will do. This could be an open file, or a multiline text after splitting on ‘n’.

The function will try to react intelligently if a string is passed as argument. If the string contains at least one ‘n’, it will be interpreted as a multiline string and be splitted on ‘n’. Else, the string will be considered and a file with that name will be opened. It is an error if the file does not exist or can not be opened.

The function returns self, so that you can write: cfg = Config().

write(filename, header='# Config written by pyFormex -*- PYTHON -*-\n\n', trailer='\n# End of config\n')[source]

Write the config to the given file

The configuration data will be written to the file with the given name in a text format that is both readable by humans and by the Config.read() method.

The header and trailer arguments are strings that will be added at the start and end of the outputfile. Make sure they are valid Python statements (or comments) and that they contain the needed line separators, if you want to be able to read it back.


Return the keys in the config.

By default this descends one level of Dicts.

82.2. Functions defined in module config


Format a dict in Python source representation.

Each (key,value) pair is formatted on a line of the form:

key = value

If all the keys are strings containing only characters that are allowed in Python variable names, the resulting text is a legal Python script to define the items in the dict. It can be stored on a file and executed.

This format is the storage format of the Config class.