All sector_classification datasets have no duplicate values of code column

[jdhoffa]

Closes #229

[jdhoffa]

Looking into this now, and there are some issues:
For the cnb_classification file, there is a code that is mapped to a whole bunch of different sectors. My understanding is that this should not be allowed in the Banks code, as it is an indication that the "code level" is not granular enough to identify the correct sector.

[jdhoffa]

A similar issue presents itself in the gics_classification file:

[jdhoffa]

Similarly for isic_classification, there are many sectors that are erroneously tied to two different sectors. We need to define which sectors are correct here:

[jdhoffa]

nace_classification is the most complicated here.

What happens is that original_code gets mapped to code, which removes the decimal mark. This is a bad idea as it effectively converts different codes (e.g. 6.2 and 62) to the same "formatted" code 620.

This is almost certainly a bug, and should be reverted, however, we do risk affecting the downstream user as it will change the expected sector classification code. I leave it to @jacobvjk and George to determine how risky this change will end up being:

[jdhoffa]

• cnb_classification - @george votes to drop this dataset entirely
• gics_classification - @george and @jacobvjk both agree to remove "coal" entirely from the 151040 code
• isic_classification - @george and @jacobvjk think this dataset is messed up, so they want to drop it. Separately, we should have an automated warning message in r2dii.match that suggests users to e-mail us/ or go to a link to GitHub issues where they can request what they want. Tracked in breaking change: deprecate cnb_classification dataset #329 and feat: ensure informative warning message when a user inputs a sector classification that doesn't exist in r2dii.data r2dii.match#447
• nace_classification - Use the original_code AND use the letter prepended. (e.g. "A1.11"). They think that original_code should map to code, and that we should drop the code_level column entirely.

[jdhoffa]

This is ready for review.

Progress on the deprecation of cnb_classification and isic_classification will be tracked in #329

[jacobvjk]

There seems to be some stray code that I could not see used anywhere else in the repo. See comment. Maybe remove?

[jacobvjk]

this does not seem to be used anywhere

[jdhoffa]

Nice catch!

[jacobvjk]

lgtm