tespent opened a new issue, #1103: URL: https://github.com/apache/datafusion-python/issues/1103
**Is your feature request related to a problem or challenge? Please describe what you are trying to do.** <!--A clear and concise description of what the problem is. Ex. I'm always frustrated when [...] (This section helps Arrow developers understand the context and *why* for this feature, in addition to the *what*)--> I am trying to integrate a SQL engine (and I chose DataFusion in the end!) into our system (based on Ray Data), to enable our users to retrieve and select data using SQL. We already have production-ready Python-interfaced catalog providers and file readers. I am attempting to use the datafusion-python library to work with these custom catalogs, but I’ve run into an issue: the Python API does not currently provide a way to implement or write a custom CatalogProvider. This is preventing me from extending DataFusion to interact with existing catalog sources in Python. Without this functionality, integrating existing custom catalogs becomes difficult, especially in a dynamic or production-level environment where flexibility in data source management is key. **Describe the solution you'd like** <!-- A clear and concise description of what you want to happen. --> As of now, I've had to write custom rust bindings for python objects. These bindings are generic enough to fit into the datafusion-python project as it simply wraps python objects to implement `CatalogProvider`, `SchemaProvider`, `TableProvider`, `ExecutionPlan`, `RecordBatchStream` and more, and extended many existed bindings to expose more information from rust to python. I would like to contribute my solution and code to datafusion-python to implement this feature! **Describe alternatives you've considered** <!-- A clear and concise description of any alternative solutions or features you've considered. --> Based on the existing API, we can first fetch the data, then run a sql against it using something like `from_arrow` or `create_dataframe` (similar to [smallpond](https://github.com/deepseek-ai/smallpond)'s `partial_sql` but they are using DuckDB). However, this solution has many drawbacks: 1. Users are required to separate data fetching logic and the sql query, which negatively impacts user experience. 2. Pushdowns will not work. So end users will need to properly write filters and column selections for our datasource. 3. The existing conversion APIs require all the data to be ready before processing, which breaks data pipelines and leads to high memory usage, as we have to buffer all the data. **Additional context** <!-- Add any other context or screenshots about the feature request here. --> This feature would be beneficial for anyone looking to use datafusion-python in environments with existing python data sources, or non-standard data sources based on python, such as custom databases, external APIs, or even customized cloud-based storage systems. This feature can help DataFusion better embrace the python ecosystem, and benefit the DataFusion community by providing a easier way to prototype customized catalogs (data sources) using python, which is a great glue language for integrating things! By offering this feature, framework developers can easily integrate DataFusion into their python-based systems (like Ray Data and other data processing frameworks), empowering these systems with extensible and advanced SQL functionalities! This feature may also be related to #24, as it involves `SessionContext.register_catalog`, which has not yet been implemented. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: github-unsubscr...@datafusion.apache.org.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: github-unsubscr...@datafusion.apache.org For additional commands, e-mail: github-h...@datafusion.apache.org
