Tutorial for cli2: Dynamic CLI for Python 3 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Architecture ============ Overview -------- cli2 is built on 3 moving parts which you can swap with your own or inherit from with ease: - :py:class:`~cli2.command.Command`: Represents a target callback, in charge of CLI args parsing and execution, can serve as entry point. - :py:class:`~cli2.group.Group`: Same as above, except that it routes multiple Commands, can serve as entry point as well. - :py:class:`~cli2.argument.Argument`: Represents a target callback argument, in charge of deciding if it wants to take an argument as well as casting it into a Python value. Also, despite this package name which certainly reminds some female genitalia parts (maybe intentionally), we're going to be kind of gay with CLIs and require very pretty outputs, so `cli2` also provides a couple of helper modules: - :py:mod:`~cli2.colors`, available in the `cli2.c` namespace, provides a bunch of arbitrary ANSI color codes which I like. - :py:mod:`cli2.display` provides a :py:func:`~cli2.display.print` function you're supposed to be able to throw anything at, it will try to prettify any command result with colored YAML output. **To disable colors in output, set `NO_COLORS=1` environment variable!** Python's standard `difflib` module's `unified_diff` method. - :py:mod:`~cli2.configuration`, available in ``cli2.cfg`` variable, provides a lazy environment variable configuration wrapper - :doc:`cli2.client`, turned out to be a framework to build ORMs on top of REST APIs Command ======= Create a command from any callable: .. code-block:: python import cli2 def yourcmd(somearg: str): """ Your own command. :param somearg: It's some string argument that this function will return """ return somearg # this if is not necessary, but prevents CLI execution when the module is # only imported if __name__ == '__main__': cli2.Command(yourcmd).entry_point() Running the script without argument will show the generated help: .. image:: example.png Also, passing `--help` will display help for a command. Complex signature ----------------- The same function with a rich signature like this: .. code-block:: python def yourcmd(somearg, x=None, verbose : bool = False, *args, foo=None, **kwargs): Will display help as such: .. image:: example2.png Posix style ----------- You might prefer to have dashes in front of argument names in the typical style of command lines, you just need to enable the posix attribute: .. code-block:: python cli2.Command(yourcmd, posix=True).entry_point() In this case, help will look like this: .. image:: example2posix.png .. warning:: I still don't use the POSIX mode, it's far from perfect, but I'll gladly try to fix bugs! Testing ------- The :py:meth:`~cli2.command.Command.parse` method will provision the :py:attr:`~cli2.command.Command.bound` attribute which is a Python 3 BoundArguments instance, so you could test parsing as such: .. code-block:: python cmd = cli2.Command(yourcmd) cmd.parse('a', 'b', 'c=d') assert cmd.bound.arguments == dict(somearg='a', x='b', kwargs={'c': 'd'}) Same if you want to use the posix style: .. code-block:: python cmd = cli2.Command(yourcmd, posix=True) cmd.parse('a', 'b', '--c=d') assert cmd.bound.arguments == dict(somearg='a', x='b', kwargs={'c': 'd'}) Entry point ----------- Another possibility is to add an entry point to your setup.py as such: .. code-block:: python entry_points={ 'console_scripts': [ 'yourcmd = yourcmd:cli.entry_point', ], }, Then, declare the command in a ``cli`` variable in ``yourcmd.py``: .. code-block:: python # if __name__ == '__main__': if block not required in entry point cli = cli2.Command(yourcmd) Method commands --------------- We've seen how to expose functions to the CLI. If you want to expose methods of an object, constructed based on CLI arguments, see the example in example_obj.py: .. literalinclude:: ../example_obj.py :language: python The rest of the tutorial will describe each part used in there. ``_cli2`` --------- If you need to know if a function is executed from cli2, you can add a ``_cli2`` special keyword argument: .. code-block:: python def your_cmd(some, _cli2=None): if _cli2: print('in cli2') else: return some ``_cli2`` will be the :py:class:`~cli2.command.Command` instance if detected in function signature. Group ===== :py:class:`~cli2.group.Group` can be used in place of :py:class:`~cli2.argument.Command`, and new commands can be added into it. Decorator syntax ---------------- Example using :py:meth:`~cli2.group.Group.cmd`: .. code-block:: python """My command docstring""" import cli2 cli = cli2.Group(doc=__doc__) @cli.cmd def yourcmd(): """Your command""" @cli.cmd(color='red') def dangerzone(something): """A dangerous command""" if __name__ == '__main__': cli.entry_point() As you can see, the decorator may be called with or without arguments, any argument that are passed would override the default attributes from the generated :py:class:`~cli2.argument.Command`. Running this script without argument will show: .. image:: example3.png Group overrides --------------- A group may define default overrides for arguments: .. code-block:: python cli = cli2.Group('foo') class Foo: @classmethod def factory(cls): return cls() @cli.cmd def send(self, something): return something cli['foo'].overrides['self']['factory'] = Foo.factory Will shadow ``self`` from the CLI and instead call Foo.factory. Python API ---------- Equivalent example using :py:meth:`~cli2.group.Group.add`: .. code-block:: python import cli2 def yourcmd(): """Your command""" def dangerzone(something): """A dangerous command""" if __name__ == '__main__': cli = cli2.Group() cli.add(yourcmd) cli.add(dangerzone, color='red') cli.entry_point() Lazy loading: overriding Group ------------------------------ Equivalent example, but built during runtime, having the arguments at disposal: .. code-block:: python import cli2 def yourcmd(): """Your command""" def dangerzone(something): """A dangerous command""" class Cli(cli2.Group): def __call__(self, *argv): # you could use the *argv variable here self.add(yourcmd) self.add(dangerzone, color='red') return super().__call__(*argv) if __name__ == '__main__': Cli().entry_point() This is the same as the other command group examples above, but here the Group is built during runtime. See the source code for the ``cli2`` command, which implements an infitely lazy loaded command tree based on introspection of the passed arguments with extremely little code. Lazy loading: using Group.load ------------------------------ You could also load commands more massively with the :py:meth:`~cli2.group.Group.load` method which will load any callable given as Python object or as dotted python path, all the following work: .. code-block:: python group = cli2.Group() group.load(YourClass) group.load(your_object) group.load('your_module') group.load('your_module.your_object') Injecting CLIs from submodules ------------------------------ You don't need to declare all your commands in a single script because ny script can declare a :py:class:`~cli2.group.Group` that can be imported in the main CLI. For example, in `foo.py`: .. code-block:: python cli = cli2.Group() @cli.cmd def foo(): pass You can load it in another main CLI script, `cli.py`: .. code-block:: python from . import foo cli = cli2.Group() cli['foo'] = foo.cli @cli.cmd def bar(): pass Argument ======== Factory ------- You may want some arguments to have automatically computed values instead of being exposed to the user CLI. This is what argument factories are for: .. code-block:: python class Foo: @cli2.arg('auto', factory=lambda: 'autoval') @cli2.arg('self', factory=lambda cmd, arg: Foo()) def test(self, auto, arg): return auto, arg cli2.Command(Foo.test) This command will only expose the `arg` argument to the user. Both self and auto will have the result of the lambda passed as factory. If the factory callback takes an `arg` argument, then the :py:class:`~cli2.argument.Argument` object will be passed. If the factory callback takes an `cmd` argument, then the :py:class:`~cli2.command.Command` object will be passed. Aliases ------- By default, named arguments are given aliases (CLI argument names) generated from their Python argument names. For example: .. code-block:: python def yourcmd(foo=True): print(foo) cmd = cli2.Command(yourcmd) cmd.help() Will render help as such: .. image:: example_alias.png Posix ----- If posix mode is enabled, then a couple of dashes will prefix the Python argument name, and another one-letter-long alias with a single dash will be generated. .. image:: example_alias_posix.png Overrides --------- You may overrides :py:class:`~cli2.argument.Argument` attributes for a callable argument with the :py:func:`~cli2.decorators.arg` decorator: .. code-block:: python @cli2.arg('foo', alias='bar') def yourcmd(foo): pass This also takes a list of aliases: .. code-block:: python @cli2.arg('foo', alias=['foo', 'f', 'foooo']) def yourcmd(foo): pass This decorator basically sets ``yourcmd.cli2_foo`` to a dict with the alias key. Hide ---- You can also hide an argument from CLI: .. code-block:: python import cli2 @cli2.hide('foo', 'bar') def yourcmd(a, foo=None, bar=None): pass Integers -------- Type hinting is well supported, the following example enforces conversion of an integer argument: .. code-block:: python def yourcmd(i : int): pass cmd = cli2.Command(yourcmd) cmd.parse('1') assert cmd.bound.arguments == dict(i=1) Boolean ------- Declare a boolean type hint for an argument as such: .. code-block:: python def yourcmd(yourbool : bool): You won't have to specify the value of a boolean argument, but if you want to then: - for ``False``: no, 0, false - for ``True``: yes, 1, true, anything else Values don't need to be specified, which means that you don't have to type ``yourbool=true``, just ``yourbool`` or ``--yourbool`` in POSIX mode will set it to True. Since the mere presence of argument aliases suffice to bind a parameter to True, an equivalent is also possible to bind it to False: :py:attr:`~cli2.argument.Argument.negate`. It is by default generated by prefixing ``no-`` to the argument name, as such, passing ``no-yourbool`` on the command line will bind ``yourbool`` to ``False``, or in posix mode by passing ``--no-yourbool``. Note that a single-dash two-letter negate is also generated in posix mode, so ``-ny`` would also work to bind ``yourbool`` to ``False``. False ----- While the negates are set by default on boolean arguments, you may also set it on non-boolean arguments, just like you could override it like you would override aliases: .. code-block:: python @cli2.arg('yourbool', negate='--no-bool') def yourcmd(yourbool): List and Dicts -------------- Arguments annotated as list or dict will have CLI values automatically casted to Python using JSON. .. code-block:: python def yourcmd(foo: list): print(foo) But be careful with spaces on your command line: one sysarg goes to one argument:: yourcmd ["a","b"] # works yourcmd ["a", "b"] # does not because of the space However, space is supported as long as in the same sysarg: .. code-block:: python subprocess.check_call(['yourcmd', '["a", "b"]') Typable lists and dicts ----------------------- So, the above will work great when called by another program, but not really nice to type. So, another syntax for the purpose of typing is available and works as follow. Arguments with the list type annotation are automatically parsed as JSON, if that fails it will try to split by commas which is easier to type than JSON for lists of strings:: yourcmd a,b # calls yourcmd(["a", "b"]) Keep in mind that JSON is tried first for list arguments, so a list of ints is also easy:: yourcmd [1,2] # calls yourcmd([1, 2]) A simple syntax is also supported for dicts by default:: yourcmd a:b,c:d # calls yourcmd({"a": "b", "c": "d"}) The disadvantage is that JSON decode exceptions are swallowed, but by design cli2 is supposed to make Python types more accessible on the CLI, rather than being a JSON validation tool. Generated JSON args should always work though. Custom type casting ------------------- You may also hack how arguments are casted into python values at a per argument level, using decorator syntax or the lower level Python API. For example, you can override the :py:meth:`~cli2.argument.Argument.cast()` method for a given argument as such: .. code-block:: python @cli2.args('ages', cast=lambda v: [int(i) for i in v.split(',')]) def yourcmd(ages): return ages cmd = Command(yourcmd) cmd(['1,2']) == [1, 2] # same as CLI: yourcmd 1,2 You can also easily write an automated test: .. code-block:: python cmd = cli2.Command(yourcmd) cmd.parse('1,2') assert cmd.bound.arguments == dict(ages=[1, 2]) Logging ======= By default, :py:class:`~cli2.entry_point.EntryPoint`: will setup a default logger streaming all python logs to stdout with info level. Use the ``LOG`` environment variable to change it, ie:: LOG=debug yourcommand ... LOG=error yourcommand ... Or, disable this default feature with ``log=False``:: cli = cli2.Group(log=False) Tables ====== cli2 also offers a simple table rendering data that will do it's best to word wrap cell data so that it fits in the terminal. Example: .. code-block:: python cli2.Table(*rows).print() Documentation ============= Add ``'cli2.sphinx'`` to your extensions in ``docs/conf.py``, then use the following directive to generate full documentation for your CLI: ``.. cli2:auto:: cli2-example`` Then, you can refer to your commands and arguments as such: - ``:cli2:grp:`cli2-example``` - ``:cli2:cmd:`cli2-example get``` - ``:cli2:arg:`cli2-example get base_url``` Shown in :doc:`example` Overridding default code ======================== Argument overriding ------------------- Overriding an Argument class can be useful if you want to heavily customize an argument, here's an example with the age argument again: .. code-block:: python class AgesArgument(cli2.Argument): def cast(self, value): # logic to convert the ages argument from the command line to # python goes in this method return [int(i) for i in value.split(',')] @cli2.arg('ages', cls=AgesArgument) def yourcmd(ages): return ages assert yourcmd('1,2') == [1, 2] Command class overriding ------------------------ Overriding the Command class can be useful to override how the target callable will be invoked. Example: .. code-block:: python class YourThingCommand(cli2.Command): def call(self, *args, **kwargs): # do something return self.target(*args, **kwargs) @cli2.cmd(cls=YourThingCommand) def yourthing(): pass cmd = cli2.Command(yourthing) # will be a YourThingCommand You may also override at the group level, basically instanciate your :py:class:`~cli2.group.Group`: with the ``cmdclass`` argument: .. code-block:: python cli = cli2.Group(cmdclass=YourThingCommand) cli.add(your_function) CLI only arguments ------------------ A more useful example combining all the above, suppose you have two functions that take a "schema" argument that is a python object of a "Schema" class of your own. .. code-block:: python class Schema(dict): def __init__(self, filename, syntax): """ parse file with given syntax ...""" @cli.cmd def build(schema): """ build schema """ @cli.cmd def manifest(schema): """ show schema """ In this case, overriding the schema argument with custom casting won't work because the schema argument is built with two arguments: filename in syntax! Solution: .. code-block:: python class YourCommand(cli2.Command): def setargs(self): super().setargs() # hide the schema argument from CLI del self['schema'] # create two arguments programatically self.arg( 'filename', position=0, kind='POSITIONAL_ONLY', doc='File to use', ) self.arg( 'syntax', kind='KEYWORD_ONLY', doc='Syntax to use', ) def call(self, *args, **kwargs): schema = Schema( self['filename'].value, self['syntax'].value, ) return self.target(schema, *args, **kwargs) @cli.cmd def build(schema): """ build schema """ @cli.cmd def manifest(schema): """ show schema """ There you go, you can automate command setup like with the creation of a schema argument and manipulate arguments programatically! Check `cli2/test_inject.py` for edge cases and more fun examples! .. literalinclude:: ../cli2/test_inject.py :language: python Configuration ============= See :py:mod:`cli2.configuration` Edge cases ========== Simple and common use cases were favored over rarer use cases by design. Know the couple of gotchas and you'll be fine. Args containing ``=`` when ``**kwargs`` is present -------------------------------------------------- Simple use cases are favored over rarer ones when a callable has varkwargs. When a callable has ``**kwargs`` as such: .. code-block:: python def foo(x, **kwargs): pass Then, arguments that look like kwargs will be attracted to the kwargs argument, so if you want to call ``foo("a=b")`` then you need to call as such:: foo x=a=b Because the following will call ``foo(a='b')``, and fail because of missing ``x``, which is more often than not what you want on the command line:: foo a=b Now, even more of an edgy case when ``*args, **kwargs`` are used: .. code-block:: python def foo(*args, **kwargs): return (args, kwargs) Call ``foo("a", b="x")`` on the CLI as such:: foo a b=x **BUT**, to call ``foo("a", "b=x")`` on the CLI you will need to use an asterisk with a JSON list as such:: foo '*["a","b=x"]' Admittedly, the second use case should be pretty rare compared to the first one, so that's why the first one is favored. For the sake of consistency, varkwarg can also be specified with a double asterisk and a JSON dict as such:: # call foo("a", b="x") foo a **{"b":"x"} Calling with ``a="b=x"`` in ``(a=None, b=None)`` ------------------------------------------------ The main weakness is that it's difficult to tell the difference between a keyword argument, and a keyword argument passed positionnaly which value starts with the name of another keyword argument. Example: .. code-block:: python def foo(a=None, b=None): return (a, b) Call ``foo(b='x')`` on the CLI like this:: foo b=x **BUT**, to call ``foo(a="b=x")`` on the CLI, you need to name the argument:: foo a=b=x Admitadly, that's a silly edge case. Protect yourself from it by always naming keyword arguments ... ... Because the parser considers token that start with a keyword of a keyword argument prioritary to positional arguments once the positional arguments have all been bound.