NSLC cleanup

[sboltz]

This PR standardizes and cleans up the way that NSLC information is handled, particularly if it contains numeric values (particularly location codes). This is necessary because pandas can do strange things to dtypes that result in unexpected behavior.

It should have the following behavior:

• If dtypes for any of the NSLC columns are floats, then they will be converted to 2-digit zero-padded strings (1.0 becomes "01")
• If dtypes for any of the NSLC columns are ints, then they will be converted to 2-digit zero-padded strings (1 becomes "01")
• If dtypes for any of the NSLC columns are strs, then they will be left alone ("1" becomes "1" and "01" becomes "01")
• If dtypes for any of the NSLC columns are strs, but they happen to contain decimals, a TypeError will raise because this will cause confusion elsewhere ("1.0" will not be allowed)

This behavior should coerce things to follow the seed convention, but it will also allow for special cases where the NSLC are purely numeric.

[codecov-commenter]

Codecov Report

Merging #199 into master will increase coverage by 0.00%. The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #199   +/-   ##
=======================================
  Coverage   97.36%   97.36%           
=======================================
  Files          37       37           
  Lines        4175     4179    +4     
  Branches      664      665    +1     
=======================================
+ Hits         4065     4069    +4     
  Misses         52       52           
  Partials       58       58           

Flag	Coverage Δ
#unittests	97.36% <100.00%> (+<0.01%)	:arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
obsplus/constants.py	100.00% <ø> (ø)
obsplus/utils/stations.py	99.47% <ø> (-0.01%)	:arrow_down:
obsplus/events/validate.py	94.80% <100.00%> (ø)
obsplus/structures/dfextractor.py	87.35% <100.00%> (+0.60%)	:arrow_up:
obsplus/utils/pd.py	97.59% <100.00%> (+0.02%)	:arrow_up:

---

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 2e3d1f2...357d638. Read the comment docs.

[d-chambers]

This looks well thought out. I especially like how you are using a custom dtype for NSLC codes to replace a multitude of ad-hacks.

It looks like it still needs a changelog entry. Apart from that, is there anything left to do?

[d-chambers]

I do have one question about station codes. These are typically 3 characters, probably for historical reasons. I think the seed spec just requires they are between 1 and 5 chars though. It might be nice if station 1 became station "001" rather than "01" but I suppose it isn't that important. What do you think?

[sboltz]

I do have one question about station codes. These are typically 3 characters, probably for historical reasons. I think the seed spec just requires they are between 1 and 5 chars though. It might be nice if station 1 became station "001" rather than "01" but I suppose it isn't that important. What do you think?

That is an interesting point. For simplicity, I would say that if the user wants to name their station "001", they should use the string convention that we have outlined. But that is largely out of laziness and because it's difficult to figure out what the appropriate thing to do there really is. For instance, for us we would almost certainly want station 1 to become station "1". The reason for having it be "01" is because all of the NSLC components are now using the same data type, and in particular I wanted to make sure it handles location codes properly.

[d-chambers]

Looks good