Ensure that numeric precision is included only if not None

[mjumbewu]

This pull request ensures that the numeric_size of the Column is not None before using including it in the column type string representation.

---

While unioning tables together with the dbt-utils union_table macro, we were running into syntax errors like the following from SQL generated by DBT:

invalid input syntax for integer: none
LINE 54:            "total_tax_due"::numeric(None) as ...`

We were initially inclined to go and add a precision to the source table, but the models where this was happening were themselves based on derivative tables. For example, in the above case, total_tax_due is defined as an aggregate:

...
sum(tax_due) AS total_tax_due
...

So identifying the precision beforehand is nontrivial.

[drewbanin]

don't try to get too clever
🙄 😞

Thanks @mjumbewu - this looks good to me. I think this is a Postgres-only quirk. On Redshift, numeric will expand to numeric(18, 0). There's a similar issue with text types -- dbt interprets them as varchar(255) whereas it should be unlimited in length on pg.

I think it would be really great to add unit tests for these methods. Is that something you're interested in contributing? If not, I can tack some on here early next week!

Thanks for contributing :)

[mjumbewu]

@drewbanin Yes, definitely makes sense. I started down that path, but didn't immediately see a test module to add them to (and was in a bit of a rush 😊). I went ahead and added a test module.

Also thought it'd be nice to see an integration test for this on Postgres. I added a couple numeric fields for the 008_schema_tests_test.py module. I wasn't sure what I needed to run the integration tests locally, so we'll see whether those break anything.

[drewbanin]

perfect! Looks great - I'll merge when appveyor passes :)

Thanks again!