Option and Configuration Handling
Command-line options are often also set in configuration files for |Flake8|. While not all options are meant to be parsed from configuration files, many default options are also parsed from configuration files as well as most plugin options.
In |Flake8| 2, plugins received a :class:`optparse.OptionParser` instance and called :meth:`optparse.OptionParser.add_option` to register options. If the plugin author also wanted to have that option parsed from config files they also had to do something like:
parser.config_options.append('my_config_option') parser.config_options.extend(['config_opt1', 'config_opt2'])
This was previously undocumented and led to a lot of confusion about why registered options were not automatically parsed from configuration files.
Since |Flake8| 3 was rewritten from scratch, we decided to take a different
approach to configuration file parsing. Instead of needing to know about an
undocumented attribute that pep8 looks for, |Flake8| 3 now accepts a parameter
parse_from_config which is a boolean
|Flake8| does this by creating its own abstractions on top of :mod:`optparse`. The first abstraction is the :class:`flake8.options.manager.Option` class. The second is the :class:`flake8.options.manager.OptionManager`. In fact, we add three new parameters:
The last two are not specifically for configuration file handling, but they do improve that dramatically. We found that there were options that, when specified in a configuration file, often necessitated being spit multiple lines and those options were almost always comma-separated. For example, let's consider a user's list of ignored error codes for a project:
[flake8] ignore = # Reasoning E111, # Reasoning E711, # Reasoning E712, # Reasoning E121, # Reasoning E122, # Reasoning E123, # Reasoning E131, # Reasoning E251
It makes sense here to allow users to specify the value this way, but, the standard libary's :class:`configparser.RawConfigParser` class does returns a string that looks like
"\nE111, \nE711, \nE712, \nE121, \nE122, \nE123, \nE131, \nE251 "
This means that a typical call to :meth:`str.split` with
',' will not be
sufficient here. Telling |Flake8| that something is a comma-separated list
comma_separated_list=True) will handle this for you. |Flake8| will
["E111", "E711", "E712", "E121", "E122", "E123", "E131", "E251"]
Next let's look at how users might like to specify their
Presently OpenStack's Nova project has this line in their tox.ini:
exclude = .venv,.git,.tox,dist,doc,*openstack/common/*,*lib/python*,*egg,build,tools/xenserver*,releasenotes
We think we can all agree that this would be easier to read like this:
exclude = .venv, .git, .tox, dist, doc, *openstack/common/*, *lib/python*, *egg, build, tools/xenserver*, releasenotes
In this case, since these are actually intended to be paths, we would specify
normalize_paths=True because we
want the paths to be provided to us with some consistency (either all absolute
paths or not).
Now let's look at how this will actually be used. Most plugin developers will receive an instance of :class:`~flake8.options.manager.OptionManager` so to ease the transition we kept the same API as the :class:`optparse.OptionParser` object. The only difference is that :meth:`~flake8.options.manager.OptionManager.add_option` accepts the three extra arguments we highlighted above.
Configuration File Management
In |Flake8| 2, configuration file discovery and management was handled by pep8. In pep8's 1.6 release series, it drastically broke how discovery and merging worked (as a result of trying to improve it). To avoid a dependency breaking |Flake8| again in the future, we have created our own discovery and management. As part of managing this ourselves, we decided to change management/discovery for 3.0.0. We have done the following:
- User files (files stored in a user's home directory or in the XDG directory
inside their home directory) are the first files read. For example, if the
user has a
~/.flake8file, we will read that first.
- Project files (files stored in the current directory) are read next and merged on top of the user file. In other words, configuration in project files takes precedence over configuration in user files.
- New in 3.0.0 The user can specify
--append-config <path-to-file>repeatedly to include extra configuration files that should be read and take precedence over user and project files.
- New in 3.0.0 The user can specify
--config <path-to-file>to so this file is the only configuration file used. This is a change from |Flake8| 2 where pep8 would simply merge this configuration file into the configuration generated by user and project files (where this takes precedence).
- New in 3.0.0 The user can specify
--isolatedto disable configuration via discovered configuration files.
To facilitate the configuration file management, we've taken a different approach to discovery and management of files than pep8. In pep8 1.5, 1.6, and 1.7 configuration discovery and management was centralized in 66 lines of very terse python which was confusing and not very explicit. The terseness of this function (|Flake8|'s authors believe) caused the confusion and problems with pep8's 1.6 series. As such, |Flake8| has separated out discovery, management, and merging into a module to make reasoning about each of these pieces easier and more explicit (as well as easier to test).
Configuration file discovery is managed by the :class:`~flake8.options.config.ConfigFileFinder` object. This object needs to know information about the program's name, any extra arguments passed to it, and any configuration files that should be appended to the list of discovered files. It provides methods for finding the files and similar methods for parsing those fles. For example, it provides :meth:`~flake8.options.config.ConfigFileFinder.local_config_files` to find known local config files (and append the extra configuration files) and it also provides :meth:`~flake8.options.config.ConfigFileFinder.local_configs` to parse those configuration files.
local_config_files also filters out non-existent files.
Configuration file merging and managemnt is controlled by the :class:`~flake8.options.config.MergedConfigParser`. This requires the instance of :class:`~flake8.options.manager.OptionManager` that the program is using, the list of appended config files, and the list of extra arguments. This object is currently the sole user of the :class:`~flake8.options.config.ConfigFileFinder` object. It appropriately initializes the object and uses it in each of
takes the user and local configuration files that are parsed by
main usage of the
MergedConfigParser is in
Aggregating Configuration File and Command Line Arguments
:func:`~flake8.options.aggregator.aggregate_options` accepts an instance of :class:`~flake8.options.maanger.OptionManager` and does the work to parse the command-line arguments passed by the user necessary for creating an instance of :class:`~flake8.options.config.MergedConfigParser`.
After parsing the configuration file, we determine the default ignore list. We use the defaults from the OptionManager and update those with the parsed configuration files. Finally we parse the user-provided options one last time using the option defaults and configuration file values as defaults. The parser merges on the command-line specified arguments for us so we have our final, definitive, aggregated options.
.. autofunction:: flake8.options.aggregator.aggregate_options
.. autoclass:: flake8.options.manager.Option :members: __init__, normalize, to_optparse
.. autoclass:: flake8.options.manager.OptionManager :members: :special-members:
.. autoclass:: flake8.options.config.ConfigFileFinder :members: :special-members:
.. autoclass:: flake8.options.config.MergedConfigParser :members: :special-members: