Metric Builder
Utility for building templated metric extraction queries that can be traversed through time.
Prerequisites
You will need the following to run this code:
Installation
To be determined...
Usage
In order to extract a given metric, a Metric
object needs to be instantiated:
metric = Metric(
query="""
SELECT count(*) AS total
FROM `project.dataset.table`
WHERE DATETIME_TRUNC(created_datetime, DAY) = '{{ reference_time | format_date('%Y-%m-%d') }}'
""",
reader = BigQueryReader(json_credentials_path='/path/to/creds.json')
)
The query
parameter is a templated query where you can format the reference_time
datetime
object to the required format using template filters.
The reader
parameter is the object that is actually going to connect to the desired database and perform the queries.
The metric
object can now be used to fetch metrics for a given point in time as follows:
result = metric.fetch(reference_time=datetime.date(2019, 10, 21))
The result is returned as a list of dictionaries.
Template filters
Jinja2 is used as the templating engine. All built in Jinja filters are thus available. Relevant custom template filters have been added though for convenience:
format_date
Specify format of datetime:
'{{ reference_time | format_date('%Y-%m-%d') }}'
day_delta
Change a given datetime object by a specified number of days:
'{{ reference_time | day_delta(-7) | format_date('%Y-%m-%d') }}'
Readers
Any reader will implement the following method that is used to execute queries:
def execute(self, query) -> List[Dict[str, Any]]:
...
BigQueryReader
The underlying client is required to be authenticated with the necessary priviledges to read from the requested BigQuery tables.
If you authenticate with:
gcloud auth login
or
export GOOGLE_APPLICATION_CREDENTIALS="/path/to/keyfile.json"
then you can just instantiate your Reader
like this:
reader = BigQueryReader()
The other option is to explicitly authenticate with a service account key file:
reader = BigQueryReader(json_credentials_path='/path/to/creds.json')
HiveReader
Coming soon...