Review the Iris user experience

[trexfeathers]

Raised as a 'wish-list' item in an offline meeting. User experience is usually something to think about while making some other change. Approaching UX as its own 'problem' to 'solve' offers more scope to truly improve the experience. Vote if you want to prioritise this 🙂

Related Discussions:

• Remove the need for calling `data` in `cube.data.rechunk` #6592
• Should copy() still offer the option to replace the data? #6265
• Iris User Experience #6594 - A discussion for all things UX!

[ukmo-ccbunney]

Some loose thoughts from a Dragon Taming meeting - these need scoping out a bit more:

• Tidying up the API (but minimizing user impact)
• Sharing of data objects in Iris
• Documentation, both RTD and docstring; probably already quite good but worth reviewing
  • Terminology: Eager (Xarray) vs Real (Iris) data terminology. Is in "package phrasebook", but worth highlighting
  • Update Phrasebook and highlight/publicise more widely.
• Make it easier to load complex data:
  • often easier to (at least) load using xarray
  • make Iris more lenient so we can at least load the data in some form and fix later. (Martin's Capture un-loadable NetCDF objects instead of discarding/crashing #6317 will help this)
• Cube copying with data=None is a bit confusing with dataless cubes. Align the API with other packages where possible.
• Interpolation sometimes silently fails.
  • Example for AVD Surgery: interpolation silently failing due not to mismatched coord bounds
  • Not a lot of questions on the Discussion board from external users??? Last one over a year ago. Just an observation

Going forward - we need to cast a wide net here to capture examples of where UX was suboptimal:

• Trawling viva engage for user issues
• Going through Iris issues
• Reach out to known "power users".