- Instead of modifying cdb_service_quota_info to return max_batch_size,
a new type (service_quota_info_batch) and a new function
(cdb_service_quota_info_batch) are created. That makes deployment safe.
- Fixes geocoding with forced batch size 1.
- Improves namespacing for count_estimate (->
cdb_dataservices_client.cdb_count_estimate).
- Improves namespacing for jsonb_array_casttext (->
cdb_dataservices_client.cdb_jsonb_array_casttext).
New functions to perform a precheck for the user data and metadata
parameters that are going to be used to execute a OBS_GetData operation.
Doing this before the OBD_GetData we are able to check if we're going to
have problems with the current request. Is useful because right now
we're using the *_Exception_Safe functions which return NULL in case we
have an exception, this leads to execute the whole request even if we
know beforehand that it's going to fail.
I was creating the wrappers before the wrapped function and postgres
extensions do not support that.
I did not notice in my local environment cause the old functions were
already there.
As shown in the tests, that type has the flexibility of being able to be
cast to from any other numeric type. This gives us flexibility without
bloating the API with lots of type-specific functions.
- Fixed fallback becuase we have to use spiexceptions in this case
because we retrieve the configuration usig plpy.execute fuctions that
wrap any exception into a spiexception.
- In case we don't want to use Mapzen, we could leave the api_key empty
becuase we arise a ConfigException if the Mapzen api_key is empty so
we are able to make fallback. Right now we can't remove the mapzen
configuration because it'll fail when the InternalGeocoderConfig try to
load the ServiceDBConfig
This is a dirty hack, we should improve how the DB config is loaded. See
