SQLAlchemy

Introduction

sqlcommenter-sqlalchemy provides a factory to create before_cursor_execute, called BeforeExecuteFactory

We provide a BeforeExecuteFactory that takes options such as

CommenterCursorFactory(with_opencensus=<True or False>)

We provide options such as with_opencensus because

Since OpenCensus trace_id and span_id are highly ephemeral, including them in SQL comments will likely break any form of statement-based caching that doesn’t strip out comments.

Requirements

Installation

pip3 install google-cloud-sqlcommenter
git clone https://github.com/google/sqlcommenter.git
cd python/sqlcommenter-python && python3 setup.py install

and then we shall perform the following imports in our source code:

BeforeExecuteFactory

BeforeExecuteFactory is a factory that creates a before_cursor_execute hook to your engine to grab information about your application and augment it as a comment to your SQL statement.

from sqlalchemy import create_engine, event
from sqlcommenter.sqlalchemy.executor import BeforeExecuteFactory

engine = create_engine(...) # Create the engine with your dialect of SQL
event.listen(engine, 'before_cursor_execute', BeforeExecuteFactory(), retval=True)
engine.execute(...) # comment will be appended to SQL before execution

NOTE Please ensure that you set retval=True when listening for events

and this will produce such output on for example a Postgresql database logs:

2019-06-30 18:01:16.315 PDT [96973] LOG:  statement: SELECT * FROM polls_question
/*traceparent='00-ade4c36dc5e43b503a5bba237ea11746-578a74a562044332-01'*/

with_openCensus=True

To enable the comment cursor to also attach information about the current OpenCensus span (if any exists), pass in option with_opencensus=True when invoking BeforeExecuteFactory, so

engine = create_engine("postgresql://:$postgres$@127.0.0.1:5432/quickstart_py")
event.listen(engine, 'before_cursor_execute', BeforeExecuteFactory(with_opencensus=True), retval=True)
engine.execute(...) # comment will be appended to SQL before execution

NOTE Please ensure that you set retval=True when listening for events

Fields

Field Description Included by default
db_driver The underlying database driver e.g. 'psycopg2'
db_framework The version of SQLAlchemy in the form 'sqlalchemy:<sqlalchemy_version>'
traceparent The W3C TraceContext.Traceparent field of the OpenCensus trace – optionally defined with with_opencensus=True
tracestate The W3C TraceContext.Tracestate field of the OpenCensus trace – optionally defined with with_opencensus=True

End to end examples

Source code

#!/usr/bin/env python3

from sqlalchemy import create_engine, event
from google.cloud.sqlcommenter.sqlalchemy.executor import BeforeExecuteFactory

def main():
    tracer = Tracer(exporter=NoopExporter, sampler=AlwaysOnSampler())
    engine = create_engine(DB_URL)

    listener = BeforeExecuteFactory(with_opencensus=True)
    event.listen(engine, 'before_cursor_execute', listener, retval=True)

    with tracer.span():
        result = engine.execute('SELECT * FROM polls_question')
        for row in result:
            print(row)

if __name__ == '__main__':
    main()
#!/usr/bin/env python3

from sqlalchemy import create_engine, event
from google.cloud.sqlcommenter.sqlalchemy.executor import BeforeExecuteFactory

DB_URL = '...'  # DB connection info

def main():
    engine = create_engine(DB_URL)

    listener = BeforeExecuteFactory(with_db_framework=True)
    event.listen(engine, 'before_cursor_execute', listener, retval=True)

    result = engine.execute('SELECT * FROM polls_question')
    for row in result:
        print(row)

if __name__ == '__main__':
    main()
#!/usr/bin/env python3

from sqlalchemy import create_engine, event
from google.cloud.sqlcommenter.sqlalchemy.executor import BeforeExecuteFactory

DB_URL = '...'  # DB connection info

def main():
    engine = create_engine(DB_URL)

    listener = BeforeExecuteFactory(with_db_driver=True)
    event.listen(engine, 'before_cursor_execute', listener, retval=True)

    result = engine.execute('SELECT * FROM polls_question')
    for row in result:
        print(row)

if __name__ == '__main__':
    main()
python3 main.py
(1, 'Wassup?', datetime.datetime(2019, 5, 30, 13, 51, 12, 910545, tzinfo=psycopg2.tz.FixedOffsetTimezone(offset=-420, name=None)))
(2, 'Wassup?', datetime.datetime(2019, 5, 30, 13, 57, 45, 905771, tzinfo=psycopg2.tz.FixedOffsetTimezone(offset=-420, name=None)))
(3, 'Wassup?', datetime.datetime(2019, 5, 30, 13, 57, 46, 908185, tzinfo=psycopg2.tz.FixedOffsetTimezone(offset=-420, name=None)))
(4, 'Wassup?', datetime.datetime(2019, 5, 30, 13, 57, 47, 557196, tzinfo=psycopg2.tz.FixedOffsetTimezone(offset=-420, name=None)))
(5, 'Wassup?', datetime.datetime(2019, 5, 30, 13, 57, 47, 853424, tzinfo=psycopg2.tz.FixedOffsetTimezone(offset=-420, name=None)))

Results

Examining our Postgresql server logs

2019-07-18 14:10:15.228 -03 [30071] LOG:  statement: SELECT * FROM polls_question
/*traceparent='00-bf66750ad4c76f614c0a99d843758cbb-e6b27c3caf35de73-01'*/
2019-07-18 14:11:19.576 -03 [30108] LOG:  statement: SELECT * FROM polls_question
/*db_framework='sqlalchemy%3A1.3.5'*/
2019-07-18 14:03:33.426 -03 [29858] LOG:  statement: SELECT * FROM polls_question
/*db_driver='psycopg2'*/

With flask

When coupled with the web framework flask, we still provide middleware to correlate your web applications with your SQL statements from sqlalchemy. Please see this end-to-end guide below:

References

Resource URL
sqlcommenter-sqlalchemy on PyPi https://pypi.org/project/google-cloud-sqlcommenter
sqlcommenter-sqlalchemy on Github https://github.com/google/sqlcommenter
OpenCensus https://opencensus.io/
OpenCensus SpanID https://opencensus.io/tracing/span/spanid
OpenCensus TraceID https://opencensus.io/tracing/span/traceid