Skip to content

AutoReload.md

Latest commit

 

History

History
46 lines (24 loc) · 2.99 KB

AutoReload.md

File metadata and controls

46 lines (24 loc) · 2.99 KB

Auto Reload

Other things being equal, the rate of progress in application development is directly related to the time it takes to make and test one small change. By default, the NearlyPurePythonWebAppDemo (NPPWAD hereafter) skeleton will detect changes in any of the Python source files and reload both the server and the client side -- including a rebuild of client.js from client.py if needed.

Note: The mechanism described here is also incorporated in the new single-file skeleton, allinone.py.

How it works

Bottle already provides an auto-reloader and a debug mode. The Bottle Tutorial explains both of those here. NPPWAD takes advantage of them and extends them by

  1. Incorporating a simple rebuild facility in server.py, and
  2. Running a client-side check on server start time in client.py.

The result is that changing any of the .py source files triggers the following chain of events:

  • Bottle notices that a file has changed and reloads server.py.
  • As server.py re-initializes it checks the modification times on the sources and regenerates __html__/index.html and client.js as needed.
  • If a prior version of client.js is running in a browser, the next state update will detect a change in _state['server_start_time'] and reload the page using location.reload().

Usage

Reload and debug are enabled by default when you launch from the command line.

To disable these features in a production version, launch server.py with --no-reloader --no-debug at the command line. If launching by direct call to server.serve(), note that reload and debug are disabled by default in the keyword args to serve().

Limitations and caveats

The reloader in Bottle inspects sys.modules to determine what files to monitor -- meaning that it only tracks modules that are in the tree of modules imported into server.py. To monitor changes in client-side Python sources, e.g. client.py, we must import them into server.py -- which means that such modules must not cause errors or have undesired side effects when imported into Python code. Notice, for example, at the bottom of client.py the lines

try:
    document.addEventListener('DOMContentLoaded', start)
except NameError:
    pass

This prevents a failed import when document, a global object in JS, is not defined in Python. Fortunately, the event-driven nature of client-side code makes it easy to confine module-level execution to a single statement, as above, to avoid littering the code with many similar guard constructs.

Another obvious limitation is that the reloader only monitors Python sources. For a large project with sources in other languages, you would probably want to set up a build system triggered from a monitor program such as entr. To connect it with the server.py reloader, you could arrange for the build system to update a dummy python module, e.g. reloadme.py imported by server.py for the sole purpose of detecting the need for a reload.