Turn off cache logging when --log-cache-events is not set

[drewbanin]

Issue

Issue description

dbt has a --log-cache-events flag, but it appears that some invocations of dbt log the cache events even when this is flag not provided. Cache events should only be logged when --log-cache-events is provided.

Results

--log-cache-events was not provided, but cache events were logged to logs/dbt.log

This happened for both dbt run and dbt archive, though it may apply to other types of invocations as well.

System information

The output of dbt --version:

0.12.1

The operating system you're running on:
Amazon Linux

The python version you're using (probably the output of python --version)
3.5

Steps to reproduce

TBD. In the logs I have which demonstrates this, I see:

running dbt with arguments Namespace(cls=, debug=False, log_cache_events=False, profile='user', profiles_dir='/tmp/jobs/508650/.dbt', strict=False, target='prod', threads=None, use_cache=True, vars='{}', which='archive')

followed by parsing, and then:

On master: select
table_name as name, table_schema as schema, table_type as type
from information_schema.tables
where table_schema ilike 'analytics'
SQL status: SUCCESS 70 in 1.15 seconds
Adding relation: _CachedRelation(schema=ANALYTICS, identifier=XXX, inner="ANALYTICS"."XXX")
before adding: {'ANALYTICS.XXX': [],
'ANALYTICS.YYY': [],
'ANALYTICS.ZZZ': [],

[beckjake]

This is another issue around calling handle_and_check twice (see #1138) / bypassing dbt's logging machinery. If you never call dbt.logging.initialize_logger, the cache logger will propagate by default due to how python works.

The safest solution here is to flip that flag during handle_and_check based on the given arguments, instead of during one-time logger initialization.

[beckjake]

Fixed in #1144