dbt is adding incorrectly cased values to the cache

[beckjake]

Issue

On snowflake, dbt is incorrectly adding lowercase and uppercase versions of model names to the cache somehow. In particular, models added via add after a materialization is successfully executed are potentially the wrong case.

Issue description

dbt run twice on a snowflake project with casing set defaults (I think?) and lowercase model filenames. If you dump the cache after, you'll see both sets of entries in the cache - correctly uppercased from the initial population and incorrectly uppercased from the add calls.

Results

The cache should only think there's one set of names. On snowflake this is benign because we don't have any relations to drop anyway, but on postgres with odd casing choices (uppercase models + quoting=False) one can imagine problems

System information

Any dbt with caching (0.12+), any OS, any python.