Enable casting to IamDataFrame multiple times

[danielhuppmann]

Please confirm that this PR has done the following:

• Tests Added
• Documentation Added
• Description in RELEASE_NOTES.md Added

Description of PR

I recently noticed that casting an IamDataFrame again (i.e., IamDataFrame(IamDataFrame(<file>)) raises a non-intuitive error message.

This PR implements a simple approach to circumvent this problem, with all attributes of the first and second casting of the IamDataFrame instance pointing to the same objects. This follows the logic for pandas.DataFrames (as far as I understand it) where performing an operation on one instance has an impact also on the second (see the new test to show how this works in pyam).

An alternative approach may be to do a full copy during the double-casting.

Use case

I wrote a script recently that cast an object to an IamDataFrame (see here) before performing a few checks and operations - this failed when the object was already an IamDataFrame. That function has been extended to first check before casting, but this may be a problem that other users run into.

[coveralls]

Coverage decreased (-0.2%) to 90.888% when pulling 98df565 on danielhuppmann:feature/init_from_iamdf into 808294b on IAMconsortium:master.

[znicholls]

Looks good to me. It might be worth clarifying this in the docs too. It’s the right behaviour I think but this ability to change two things at once is also one of the most surprising features so might be worth discussing a bit more.

[Rlamboll]

I think this does what it initially set out to do, and am happy to approve it. However the one use-case that I would understand for this is 'refreshing' databases. I don't know if you're aware (@znicholls has described this as a bug before but I assumed it was intentional) but changing the data doesn't update the values extracted by calls like "df.scenarios()", so I often call df = DataFrame(df.data) to ensure everything is consistent after modifying data directly. To me it would be most intuitive if these had the same effect regarding consistency.

[danielhuppmann]

However the one use-case that I would understand for this is 'refreshing' databases. I don't know if you're aware (@znicholls has described this as a bug before but I assumed it was intentional) but changing the data doesn't update the values extracted by calls like "df.scenarios()", so I often call df = DataFrame(df.data) to ensure everything is consistent after modifying data directly. To me it would be most intuitive if these had the same effect regarding consistency.

I'm not quite sure what you mean with "refreshing" databases - do you mean an IamDataFrame? Also, "everything is consistent" - this should be the case anyway if you use the proper functions like rename(), convert_unit(), append().

df.scenarios() returns a list of scenarios at the time when you call the function - it is not intended to be updated when df is changed...

Maybe bring this into a new issue if it should be discussed more thoroughly?

About the use case, I added a section to the PR description.

[Rlamboll]

• this should be the case anyway if you use the proper functions like rename(), convert_unit(), append().
  I don't, I directly operate on the .data and then make it consistent afterwards. I don't think it's necessarily a problem, just something to be aware of. I will start a "bug report" for clarity anyway.

[danielhuppmann]

@znicholls added a test and an additional paragraph in the notes on the to-be-expected behavior. Please say if that is sufficiently clear!

[znicholls]

Looks good, couple of minor things I'd tweak

[danielhuppmann]

@gidden, do you mind taking a look?

[codecov]

Codecov Report

Merging #396 into master will increase coverage by 0.92%.
The diff coverage is 95.23%.

@@            Coverage Diff             @@
##           master     #396      +/-   ##
==========================================
+ Coverage   92.49%   93.42%   +0.92%     
==========================================
  Files          35       35              
  Lines        4051     4046       -5     
==========================================
+ Hits         3747     3780      +33     
+ Misses        304      266      -38     

Impacted Files	Coverage Δ
setup.py	0.00% <ø> (ø)
tests/test_tutorials.py	96.42% <0.00%> (+45.81%)	⬆️
pyam/core.py	92.37% <100.00%> (+0.08%)	⬆️
tests/test_core.py	99.65% <100.00%> (+<0.01%)	⬆️

---

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 0a25b29...e135a8c. Read the comment docs.