Fix xtrigger entry point example and improve docs (#6)

* Fix xtrigger entry point example

* Update xtrigger plugins docs

src/conf.py

@@ -111,6 +111,10 @@
 autosummary_generate_overwrite = True
 autosummary_imported_members = False
 
+# Autodoc type hints can look very messy if they are included in the signature,
+# so we only include them in the description instead:
+autodoc_typehints = 'description'
+
 # Mapping to other Sphinx projects we want to import references from.
 # NOTE: To search available references, use:
 # $ python -m sphinx.ext.intersphinx <url>/objects.inv | less

src/plugins/xtriggers/index.rst

@@ -1,54 +1,68 @@
 Xtrigger Plugins
 ======================================
 
-Xtrigger plugins allow you to install and use xtriggers without them being
-in your ``CYLC_PYTHONPATH``.
+.. versionadded:: 8.3
 
+   Xtrigger plugins allow you to install and use
+   :ref:`xtriggers <Section External Triggers>` without them being
+   in ``<workflow-dir>/lib/python/`` or ``$CYLC_PYTHONPATH``.
 
-Built In Plugins
-----------------
+.. seealso::
 
-Cylc Flow provides the following xtriggers.
+   * :ref:`Built-in Clock Triggers`
+   * :ref:`Built-in Workflow State Triggers`
 
-.. autosummary::
-   :toctree: built-in
-   :template: docstring_only.rst
-
-   cylc.flow.xtriggers.echo.echo
-   cylc.flow.xtriggers.workflow_state.workflow_state
-   cylc.flow.xtriggers.xrandom.xrandom
-
-.. Note: Autosummary generates files in this directory, these are cleaned
-         up by `make clean`.
 
 .. _developing.xtrigger.plugins:
 
 Developing ``xtrigger`` plugins
 -------------------------------
 
-Cylc uses entry points registered by setuptools to search for xtrigger
-plugins.
+Cylc uses the ``cylc.xtriggers`` entry point registered by setuptools to search
+for xtrigger plugins. Each xtrigger is registered individually.
 
 Example
 ^^^^^^^
 
-Plugins are registered by registering them with the ``cylc.xtriggers``
-entry points. Each xtrigger is registered individually.
+Consider a package called ``my_package`` with the following structure:
+
+.. code-block:: python
+   :caption: ``my_package/foo.py``
+
+   def foo():
+       ...
+
+   def bar():
+       ...
+
+.. code-block:: python
+   :caption: ``my_package/baz.py``
+
+   def baz():
+       ...
+
+These xtriggers can be registered in the package's ``setup.cfg`` or
+``pyproject.toml`` file.
 
 .. code-block:: ini
    :caption: ``setup.cfg``
 
    [options.entry_points]
-       cylc.xtriggers =
-           foo = my_package.foo:foo
-           bar = my_package.foo:bar
-           baz = my_package.baz:baz
+   cylc.xtriggers =
+       foo = my_package.foo
+       bar = my_package.foo
+       baz = my_package.baz
 
 .. code-block:: toml
    :caption: ``pyproject.toml``
 
    [project.entry-points."cylc.xtriggers"]
-   foo = "my_package.foo:foo"
-   bar = "my_package.foo:bar"
-   baz = "my_package.baz:baz"
+   foo = "my_package.foo"
+   bar = "my_package.foo"
+   baz = "my_package.baz"
+
+.. tip::
 
+   It is recommended to implement only one xtrigger per module. This allows
+   you to write a ``validate`` function for each xtrigger - see
+   :ref:`xrandom.validate` for an example.

src/user-guide/writing-workflows/external-triggers.rst

@@ -34,13 +34,11 @@ broker.
 
 Cylc has several built-in external trigger functions:
 
-- clock triggers - see :ref:`Built-in Clock Triggers`
-- inter-workflow triggers - see :ref:`Built-in Workflow State Triggers`
+- :ref:`Built-in Clock Triggers`
+- :ref:`Built-in Workflow State Triggers`
 
 Trigger functions are normal Python functions, with certain constraints as
-described below in:
-
-- custom trigger functions - see :ref:`Custom Trigger Functions`
+described below in :ref:`Custom Trigger Functions`.
 
 External triggers are configured in the
 :cylc:conf:`flow.cylc[scheduling][xtriggers]` section.
@@ -62,9 +60,7 @@ in :ref:`ClockTriggerTasks`.
 Clock triggers, unlike other trigger functions, are executed synchronously in
 the main process. The clock trigger function signature looks like this:
 
-.. code-block:: python
-
-   wall_clock(offset=None)
+.. autofunction:: cylc.flow.xtriggers.wall_clock.wall_clock
 
 The ``offset`` argument is a datetime duration (``PT1H`` is 1
 hour) relative to the dependent task's cycle point (automatically passed to the
@@ -130,10 +126,7 @@ tasks off of remote task statuses or messages in other workflows.
 
 The workflow state trigger function signature looks like this:
 
-.. code-block:: python
-
-   workflow_state(workflow, task, point, offset=None, status='succeeded',
-                  message=None, cylc_run_dir=None, debug=False)
+.. autofunction:: cylc.flow.xtriggers.workflow_state.workflow_state
 
 The first three arguments are compulsory; they single out the target workflow name
 (``workflow``) task name (``task``) and cycle point
@@ -240,12 +233,7 @@ properties:
   - In ``<workflow-dir>/lib/python/``;
   - Anywhere in your ``$CYLC_PYTHONPATH``;
   - Defined using the ``cylc.xtriggers`` entry point for an installed
-    package.
-
-.. seealso::
-
-   :ref:`developing.xtrigger.plugins` for more information on writing
-   xtrigger plugins.
+    package - see :ref:`developing.xtrigger.plugins`
 
 - They can take arbitrary positional and keyword arguments
 - Workflow and task identity, and cycle point, can be passed to trigger
@@ -255,10 +243,11 @@ properties:
 - If a trigger function depends on files or directories (for example)
   that might not exist when the function is first called, just return
   unsatisfied until everything required does exist.
-- The module containing the xtrigger function may also contain a "validate"
-  function taking arguments ``args``, ``kwargs`` and ``signature``, and
-  raising ``cylc.flow.exceptions import WorkflowConfigError`` if validation
-  fails. See :ref:`xrandom.validate` for an example of a validate function.
+- The module containing the xtrigger function may also contain a ``validate``
+  function taking a single argument, which is a dictionary of all the arguments
+  passed to the xtrigger function in ``flow.cylc``. It should raise an
+  exception if validation fails. See :ref:`xrandom.validate` for an example
+  of a validate function.
 
 .. note::
 
@@ -320,14 +309,9 @@ echo
 The trivial built-in ``echo`` function takes any number of positional
 and keyword arguments (from the workflow configuration) and simply prints
 them to stdout, and then returns False (i.e. trigger condition not
-satisfied). Here it is in its entirety.
+satisfied).
 
-.. code-block:: python
-
-   def echo(*args, **kwargs):
-       print("echo: ARGS:", args)
-       print("echo: KWARGS:", kwargs)
-       return False, {}
+.. autofunction:: cylc.flow.xtriggers.echo.echo
 
 Here's an example echo trigger workflow:
 
@@ -356,9 +340,7 @@ time (useful for testing the effect of a long-running trigger function
 - which should be avoided) and has a configurable random chance of
 success. The function signature is:
 
-.. code-block:: python
-
-   xrandom(percent, secs=0, _=None, debug=False)
+.. autofunction:: cylc.flow.xtriggers.xrandom.xrandom
 
 The ``percent`` argument sets the odds of success in any given call;
 ``secs`` is the number of seconds to sleep before returning; and the
@@ -377,38 +359,40 @@ An example xrandom trigger workflow:
 Validation example using xrandom
 """"""""""""""""""""""""""""""""
 
-The xrandom xtrigger module might also contain a ``validate`` function.
+The ``xrandom`` xtrigger module contains a ``validate`` function.
+
+This will be run on the inputs to the xtrigger when calling
+``cylc validate`` or before the workflow starts, and should raise an exception
+with a meaningful description if a condition is not met.
 
-This will be run on the inputs to an xtrigger, and should raise
-a ``WorkflowConfigError`` with a meaningful description if a condition
-is not met.
+A simplified example looks like this:
 
 .. code-block:: python
 
    from cylc.flow.exceptions import WorkflowConfigError
 
-   def validate(args: List, kwargs: Dict, signature: str) -> None:
-       # Total number of args & kwargs makes sense:
-       totalargs = len(args) + len(kwargs)
-       if totalargs < 1 or totalargs > 4:
-           raise WorkflowConfigError(
-               f'{signature} xtrigger should have between 1 and 4 arguments')
-
-       # Check whether we have a percentage argument:
-       if not args and not 'percent' in kwargs:
-           raise WorkflowConfigError(
-               f'{signature} xtrigger should have a percentage argument')
-
+   def validate(args: Dict[str, Any]) -> None:
        # Check that percentage is a reasonable value:
-       percentage = args[0] if args else kwargs['percentage']
+       percent = args['percent']
        if (
-           not isinstance(percentage, (int, float))
-           or percentage > 100
-           or percentage < 0
+           not isinstance(percent, (float, int))
+           or not (0 <= percent <= 100)
        ):
            raise WorkflowConfigError(
-               f'{signature} xtrigger percentage argument'
-               f' must be a number between 0 and 100, not {percentage}')
+               "'percent' should be a float between 0 and 100"
+           )
+
+See below for a link to the full ``validate`` function:
+
+.. autofunction:: cylc.flow.xtriggers.xrandom.validate
+
+.. tip::
+
+   The arguments you call the xtrigger function with are automatically
+   validated against the function signature, so you don't necessarily need
+   to check for the presence of arguments or their types in the validate
+   function. However, you may want to check that the values are of the correct
+   type or within a certain range.
 
 
 .. _Current Trigger Function Limitations: