|
|
|
|
|
|
by meitham
998 days ago
|
|
|
We have a restful server that accepts odata query. We translate that to sql using sqlalchemy (it’s a python stack). The application is a financial risk system with billions of rows. The query usually fetches data from sqlserver, does some manipulation using pandas dataframes, then serve it as either json or csv. We added duckdb as a cache distributed across many files (a request cannot return data from more than one file) then that very same odata query goes into duckdb. Applies the standard select, filter, group by or pivot and return a dataframe. In most cases duckdb was twice faster than sqlserver. Apologies about any bad grammar/spelling errors, typing from tiny phone in bed |
|
|