-"""A decorator to help with dead simple parallelization. See usage
-below.
-
-This will just work with `Method.THREAD` and `Method.PROCESS` but to
-use `Method.REMOTE` you need to do some setup work. You need to
-configure a pool of workers. Each worker should run the same version
-of Python, ideally in identically configured virtual environments.
-And you need to be able to ssh into each machine using key-based
-authentication (i.e. non-iteractively) and run python. List machines
-in the location set by `--remote_worker_records_file` (see
-:file:executors.h for flag and an example JSON file under examples).
+"""A decorator to help with dead simple parallelization. When decorated
+functions are invoked they execute on a background thread, process or
+remote machine depending on the style of decoration::
+
+ @parallelize # defaults to thread-mode
+ def my_function(a, b, c) -> int:
+ ...do some slow / expensive work, e.g., an http request
+
+ @parallelize(method=Method.PROCESS)
+ def my_other_function(d, e, f) -> str:
+ ...do more really expensive work, e.g., a network read
+
+ @parallelize(method=Method.REMOTE)
+ def my_other_other_function(g, h) -> int:
+ ...this work will be distributed to a remote machine pool
+
+This will just work out of the box with `Method.THREAD` (the default)
+and `Method.PROCESS` but in otder to use `Method.REMOTE` you need to
+do some setup work:
+
+ 1. To use this stuff you need to hook into :mod:`pyutils.config`
+ so that this code can see commandline arguments.
+
+ 2. You need to create and configure a pool of worker machines.
+ All of these machines should run the same version of Python,
+ ideally in a virtual environment (venv) with the same
+ Python dependencies installed. Different versions of code
+ or of the interpreter itself can cause issues with running
+ cloudpicked code.
+
+ 3. You need an account that can ssh into any / all of these
+ machines non-interactively and run Python in the aforementioned
+ virtual environment. This likely means setting up ssh with
+ key-based authentication.
+
+ 4. You need to tell this parallelization framework about the pool
+ of machines where it can dispatch work by creating a JSON based
+ configuration file. The location of this file defaults to
+ :file:`.remote_worker_records` in your home directory but can
+ be overridden via the `--remote_worker_records_file`
+ commandline argument. An example JSON configuration `can be
+ found under examples
+ <https://wannabe.guru.org/gitweb/?p=pyutils.git;a=blob_plain;f=examples/parallelize_config/.remote_worker_records;hb=HEAD>`_.
+
+ 5. Finally, you will also need tell the
+ :class:`executors.RemoteExecutor` how to invoke the
+ :file:`remote_worker.py` on remote machines by passing its path
+ on remote worker machines in your setup via the
+ `--remote_worker_helper_path` commandline flag.