Framequery can both be used on its own and in combination with sqlalchemy. Framequery also supports both pure pandas and dask dataframes.
To execute queries against dataframes in the current scope, use the
framequery.execute function:
import framequery as fq
df = pd.read_csv('test.csv')
result_df = fq.execute('select * from df')To switch between pandas and dask modes, pass either model='pandas' or
model='dask'. The dask model can query both pandas and dask dataframes, but
requires extra dependencies and introduces additional overhead for simple
workloads. Per default framequery uses the pandas only model, to enable the
dask model use:
result_df = fq.execute('select * from df', model='dask')While framequery queries the surrounding scope per default, the scope can also be passed explicitly as a dict mapping table names to dataframes. For example:
scope = {'table': df}
result_df = fq.execute('select * from table', scope=scope)It is possible to further customize execution by creating a dedicated
framequery.Executor object:
executor = fq.Executor()
result_df = executor.execute('select * from table')framequery ships with its own sqlalchemy dialect. To create a framequery engine
simply use the framequery:/// url:
from sqlalchemy import create_engine
engine = create_engine('framequery:///')Additional parameters, such as basepath and model can be passed in the
query string. For example, use framequery:///?model=dask to create a
framequery engine with dask support:
engine = create_engine('framequery:///?model=dask')TODO describe spec files, once stable
To access the executor of an engine, use the
framequery.alchemy.get_executor
function. The returned executor can then be used to add custom functions:
from framequery.alchemy import get_executor
executor = get_executor(engine)
executor.add_function('foo', lambda: 'foo')Or, to add dataframes manually to the scope:
executor = get_executor(engine)
executor.update(new_table=df)See the framquery.Executor docs for a detailed
description of availabel functionality.
The constructed engine supports both being queries via text and sqlalchemy objects:
engine.execute('select sum(c1) from table').fetchall()
engine.execute(select([func.count(table.c.c1)]).fetchall()TODO