hupper is monitor for your Python process. When files change, the process
will be restarted. It can be extended to watch arbitrary files. Reloads can
also be triggered manually from code.
Builtin file monitors (in order of preference):
To install hupper, run this command in your terminal:
$ pip install hupper
The sources for hupper can be downloaded from the Github repo.
$ git clone https://github.com/Pylons/hupper.git
Once you have a copy of the source, you can install it with:
$ pip install -e .
Builtin File Monitors¶
If the watchman daemon is running, it is the preferred mechanism for monitoring files.
On MacOS it can be installed via:
$ brew install watchman
If watchdog is installed, it will be used to more efficiently watch for changes to files.
$ pip install watchdog
This is an optional dependency and if it’s not installed, then
fallback to less efficient polling of the filesystem.
The least efficient but most portable approach is to use basic file polling.
reload_interval parameter controls how often the filesystem is scanned
and defaults to once per second.
Hupper can load any Python code similar to
python -m <module> by using the
hupper -m <module> program.
$ hupper -m myapp
Starting monitor for PID 23982.
The reloading mechanism is implemented by forking worker processes from a
parent monitor. Start by defining an entry point for your process. This must
be an importable path in string format. For example,
def wsgi_app(environ, start_response):
start_response('200 OK', [('Content-Type', 'text/plain'])
if '--reload' in args:
# start_reloader will only return in a monitored subprocess
reloader = hupper.start_reloader('myapp.scripts.serve.main')
# monitor an extra file
Many applications will tend to re-use the same startup code for both the
monitor and the worker. As a convenience to support this use case, the
hupper.start_reloader() function can be invoked both from the parent
process as well as the worker. When called initially from the parent process,
it will fork a new worker, then start the monitor and never return. When
called from the worker process it will return a proxy object that can be used
to communicate back to the monitor.
Checking if the reloader is active¶
hupper.is_active() will return
True if the reloader is active and
the current process may be reloaded.
Controlling the monitor¶
The worker processes may communicate back to the monitor and notify it of
new files to watch. This can be done by acquiring a reference to the
hupper.interfaces.IReloaderProxy instance living in the worker
hupper.start_reloader() function will return the instance
hupper.get_reloader() can be used as well.
Overriding the default file monitor¶
New in version 1.2.
hupper will auto-select the best file monitor based on what
is available. The preferred order is
watchdog is installed but you do not want to use it for any reason, you
may override the default by specifying the monitor you wish to use instead in
HUPPER_DEFAULT_MONITOR environment variable. For example:
$ HUPPER_DEFAULT_MONITOR=hupper.polling.PollingFileMonitor hupper -m foo