re-think hillslopeProbability, surfaceShapeProbability, geomPosHillProbability, geomPosMountainProbability

[dylanbeaudette]

These functions are currently broken, due to a minor SQL syntax error.

Error in SDA_query(q) : 
Invalid query: The multi-part identifier "component.cokey" could not be bound.
The multi-part identifier "component.cokey" could not be bound.

SoilWeb has pre-cached versions of these data available for hillslope, geomorphic component, and mountain slope. A couple of options:

• get these data via fetchOSD() ?
• compute on-the-fly via SDA after fixing errors
• tighter integration with some of Stephen's summary based on soilDB::get_component_from_SDA.
• update these functions to use either the live data, or the pre-cached data

Major differences in implementation between SoilWeb cached proportions and SDA/R on-the-fly:

• SoilWeb normalizes component names, which will bring in family-level mapping
• the SoilWeb version contains number of records and shannon entropy (easily computed in R)
• these kind of queries to SDA aren't very efficient: WHERE UPPER(compname) IN ", in.statement, " doesn't properly utilize existing indexes

More importantly: should proportions be based on acres-mapped or number of occurrences? Incorporating acres-mapped is much more efficiently done in a big batch (e.g. SoilWeb version), but may be possible with SDA if done one series at a time.

Once this is sorted out, we can start making figures like this:

[dylanbeaudette]

SDA/SQL syntax error corrected in 885e14ac10fc1838b09d43a5bec49620b4b80900. Still need to figure out long-term interface to these data.

[dylanbeaudette]

Related issue in soilDB: https://github.com/ncss-tech/soilDB/issues/60

[dylanbeaudette]

Note that soilDB::fetchOSD() returns two of the four summaries.

[dylanbeaudette]

Deprecated and slated for removal in the next major release. Need to add the surface shape morph to fetchOSD.