Deprecate setting metadata and original_metadata attributes

[ericpre]

in favour of using set_item and add_dictionary methods or specifying metadata when creating signals - fix #2874 (comment).

Progress of the PR

• Add deprecation warning when setting metadata and original_metadata attributes
• [n/a] update docstring (if appropriate),
• [n/a] update user guide (if appropriate),
• add an changelog entry in the upcoming_changes folder (see upcoming_changes/README.rst),
• Check formatting changelog entry in the readthedocs doc build of this PR (link in github checks)
• add tests,
• ready for review.

Minimal example of the bug fix or the new feature

import hyperspy.api as hs
import numpy as np
s = hs.signals.Signal1D(np.arange(10))
s.original_metadata.add_dictionary({'a':'A', 'b':'B'})
print(s.original_metadata)

[codecov]

Codecov Report

Merging #2913 (ab87615) into RELEASE_next_minor (c7dc06c) will increase coverage by 0.25%.
The diff coverage is 93.18%.

❗ Current head ab87615 differs from pull request most recent head f06b534. Consider uploading reports for the commit f06b534 to get more accurate results

@@                  Coverage Diff                   @@
##           RELEASE_next_minor    #2913      +/-   ##
======================================================
+ Coverage               79.58%   79.84%   +0.25%     
======================================================
  Files                     209      209              
  Lines                   32262    32558     +296     
  Branches                 7239     7328      +89     
======================================================
+ Hits                    25675    25995     +320     
+ Misses                   4836     4806      -30     
- Partials                 1751     1757       +6     

Impacted Files	Coverage Δ
hyperspy/io_plugins/nexus.py	93.83% <ø> (-0.06%)	⬇️
hyperspy/misc/date_time_tools.py	86.95% <ø> (ø)
hyperspy/signal.py	76.20% <90.90%> (+0.10%)	⬆️
hyperspy/samfire.py	65.44% <91.66%> (+0.77%)	⬆️
hyperspy/_signals/complex_signal.py	93.75% <100.00%> (-0.04%)	⬇️
hyperspy/_signals/eds_sem.py	60.60% <100.00%> (ø)
hyperspy/_signals/eds_tem.py	83.50% <100.00%> (ø)
hyperspy/io.py	86.45% <100.00%> (ø)
hyperspy/misc/eels/eelsdb.py	65.21% <100.00%> (ø)
hyperspy/misc/utils.py	92.19% <100.00%> (ø)
... and 20 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c7dc06c...f06b534. Read the comment docs.

[jlaehne]

So the main idea is to prevent that people accidentally put a wrong data structure (dict instead of DictionaryTreeBrowser) into the s.metadata or s.original_metadata attributes of a signal?

I have not checked in detail yet, but I have the feeling that some tests or examples might still be using the direct access to the metadata tree.

[jat255]

I haven't tested this manually, but it overall LGTM. I'm not familiar with the extension tests, so I'm not sure why those are failing.

[jat255]

I have not checked in detail yet, but I have the feeling that some tests or examples might still be using the direct access to the metadata tree.

Yes, there are a number of these; here's one example: https://hyperspy.org/hyperspy-doc/current/user_guide/signal.html#:~:text=To%20change%20the%20default%20value%3A

We could either take care of this when the feature is actually removed (2.0), or now. I'm okay with either, although it's not a great user experience if a user follows an example in the docs and gets a warning as a result.

[hakonanes]

I'm not familiar with the extension tests, so I'm not sure why those are failing.

pyxem developers are aware of tests failing (pyxem/pyxem#788 (comment)), and the failing kikuchipy tests are unrelated (pyxem/kikuchipy#509). Cannot speak for the single failing LumiSpy test.

[ericpre]

So the main idea is to prevent that people accidentally put a wrong data structure (dict instead of DictionaryTreeBrowser) into the s.metadata or s.original_metadata attributes of a signal?

Yes, this is the idea, because setting the metadata to anything than an instance of DTB breaks many things, starting from the documentation!

I have not checked in detail yet, but I have the feeling that some tests or examples might still be using the direct access to the metadata tree.

I couldn't find any! :)

[ericpre]

Yes, there are a number of these; here's one example: https://hyperspy.org/hyperspy-doc/current/user_guide/signal.html#:~:text=To%20change%20the%20default%20value%3A

@jat255, this is different: it is only setting a "leaf" of the DTB, not the metadata attribute itself.

[jlaehne]

I'm not familiar with the extension tests, so I'm not sure why those are failing.

pyxem developers are aware of tests failing (pyxem/pyxem#788 (comment)), and the failing kikuchipy tests are unrelated (pyxem/kikuchipy#509). Cannot speak for the single failing LumiSpy test.

The failing LumiSpy test is OK for the dev-version. So we just need to do our next release in time with v1.7.

[jlaehne]

Yes, there are a number of these; here's one example: https://hyperspy.org/hyperspy-doc/current/user_guide/signal.html#:~:text=To%20change%20the%20default%20value%3A

@jat255, this is different: it is only setting a "leaf" of the DTB, not the metadata attribute itself.

OK, setting a leaf will still be OK as direct assignment. I misunderstood that as well. Then I'm also happy with this PR as far as I can see.