Stage 2 um2nc Redevelopment - Refactoring & Corrections

[truth-quark]

This step is facilitated with a unit test framework & substantially modularised the codebase for flexibility. Stage #1 development identified several code issues which were deferred to manage project complexity. Some general problems include:

• Occasional bugs or unexpected behaviour
• Some algorithms may:
  • Not work as expected
  • Contain dead code branches
• Dependencies between modules (e.g. um2netcdf is strongly coupled to iris cubes)
• Reliance on test mocks to manage dependencies
• Code doesn't use cleaner Python idioms

Step 2 tasks could include:

• Fix buggy code/algorithms
  • Timing/calendar change functions?
  • No data handling?
• Address code rework issues of high priority
• Refactoring:
  • Reduce I/O dependencies
  • Refactor mule ops to constrain usage to specific, separate section of code
  • Extract iris ops as per mule (if possible)
  • Use newer Python idioms for looping, containers, warnings & exceptions
  • Generally simplify the code base
    • Fix complex logic, e.g. cube renaming & time/cal fixes
  • Reorder code into modules & namespaces

Notes on external dependencies:

As of October 2024, um2netcdf is strongly coupled to iris cubes & 'mule. The dependency means um2nchas to change if either of these dependencies change significantly or are discontinued etc. Additionally, themule/iris` dependencies affect the test suite, both mocking the dependencies & their complex data structures.

Goal: work towards a Functional Core, Imperative Shell software architecture, as per Gary Bernhardt's Boundaries: https://www.youtube.com/watch?v=yTkzNHF6rMs talk.

• Move I/O ops to imperative shell code (the top level code, near the entry points)
  • Avoid I/O ops in low level algorithms, helper functions
  • Pass open file objs
• Exchange information with data primitives & data structures (not files)
  • Depend on the data, not other files, modules etc
• The core should be modification functions that only take & return data
  • Manipulating & saving cubes etc should occur out in the imperative shell