Developer Notes¶
This section includes discussion and helpful tips for community members seeking
to contribute to the itk_dreg ecosystem either directly or via extension modules.
Debugging¶
The following tools and approaches may be useful for debugging during development:
Local testing: When developing new registration or reduction techniques it is often useful to first develop against low-resolution data on one PC before moving to a distributed cluster environment.
Dask can be configured to run workers in sequence for easier debugging and for relaxed resource contraints:
dask.config.set(scheduler='single-threaded')
The breakpoint() function or pdb Python module can be useful to set breakpoints and examine state in a local
Python environment.
... # your code
breakpoint() # halts execution for interactive debugging
...
Remote debugging: Dask debugging documentation discusses approaches for parallel and remote debugging when running on a cluster, including printing with
dask.distributed.printand exception handling.
dask.distributed supports the Python logging library as the preferred method for
collecting log information from workers. Logging behavior can be configured
in the Dask Client instance. See Dask
logging documentation
for more information on event-based logging.
We strongly encourage developers to write unit and integration tests during development to help identify failure points and to maintain quality code practices.
itk_dregusespytestfor its testing infrastructure.Use a
dask.distributed.LocalClusterfor debugging issues with graph serialization, resource consumption, and worker behavior.