delta-rs VS arrow-datafusion

I work at Databricks, but am pretty must just an OSS nerd, mainly focusing on Delta Rust recently: https://github.com/delta-io/delta-rs

I did some keyword research and wrote this post cause lots of folks are doing searches for Delta Lake vs Parquet. I'm just trying to share a fair summary of the tradeoffs with folks who are doing this search. It's a popular post and that's why I figured I would share it here.

Seeing a lot of great libraries coming out with python bindings in the data world e.g delta-rs Polars. I see it growing in this space as a C++ alternative

If I’m not misunderstanding, you could both decode the gRPC protobuf AND write to delta lake in Rust. Tonic, Delta-rs.

I know someone will be adding this on the python side in the coming weeks. On the rust side you can use delta-rs with polars. Though you would be compiling both arrow2 and arrow-rs, so that's quite heavy.

You don’t need DBX to use Delta Lake. You can use S3 as the backend and just use the Python Delta Lake library. It works great! https://github.com/delta-io/delta-rs

Maybe if I get some free time soon I can formalize into a working example. Been wanting an excuse to try similar concept in delta-rs and polars/duckdb vs databricks/spark vs iceberg/polars.

PyO3 is being used to expose the Python bindings to the delta-rs project: https://github.com/delta-io/delta-rs

It's a great way to expose Python bindings because it "feels" Pythonic. Most users run pip install deltalake and are completely unaware that the backend is implemented in Rust.

This is quite a different user experience than Python bindings for Java backends exposed via py4j. The py4j interfaces have the Java feel and require Java to be installed, which most Python users don't like.

Yea, there are tons of employees from companies that have made massive contributions to the Spark ecosystem. Apple built Delta Lake with Databricks, see this video for more detail. Lots of Spark PMCs are from various companies. delta-rs was initially built by Scribd and is now actively maintained by engineers at Voltron & other companies. It's awesome the community has so many contributors from various sources.

Have a look at this https://delta-io.github.io/delta-rs/python/

Python's Substrait seems like the biggest/most-used competitor-ish out there. I'd love some compare & contrast; my sense is that Substrait has a smaller ambition, and more wants to be a language for talking about execution rather than a full on execution engine. https://github.com/substrait-io/substrait

We can also see from the DataFusion discussion that they too see themselves as a bit of a Velox competitor. https://github.com/apache/arrow-datafusion/discussions/6441

Agree, substrait is a really cool project! Related: if you like substrait you might want to check out datafusion too. The project is a query execution engine built on top of Apache Arrow (with SQL parser, query planner & optimizer, execution engine, extensible user defined functions, among others) and it implements a substrait provider and consumer: https://github.com/apache/arrow-datafusion/tree/main/datafus...

BTW you can see a version of what an industrial strength query optimizer / execution engine looks like in Rust https://arrow.apache.org/datafusion/

(can also use it in your own projects)

It is quite similar to what is described in this post

Specifically: https://github.com/apache/arrow-datafusion/tree/main/datafus...

The draft contains some preliminary benchmark results, comparing it to DuckDB.

https://github.com/apache/arrow-datafusion/issues/6782

Would be curious how the performance compares to [DataFusion](https://github.com/apache/arrow-datafusion) as one of the top contenders to DuckDB on this area (albeit they being different in a lot of parts, I find it one of the closest compared to all others).

ClickBench (from ClickHouse) has some benchmarks[1] where it can be compared, but am not super sure how up to date it is. At least a while back, they were majorly out of date and haven't looked too closely on whether they are keeping it fair for everyone else :)

[1]: benchmark.clickhouse.com

Apache Arrow is a pretty common memory structure these days. Datafusion is an open query engine built in Rust started by Andy Grove.

DuckDB is a great piece of software if you are

If you are looking for a query engine implemented in a safe language (Rust) I definitely suggest checking out DataFusion. It is comparable to DuckDB in performance, has all the standard built in SQL functionality, and is extensible in pretty much all areas (query language, data formats, catalogs, user defined functions, etc)

https://arrow.apache.org/datafusion/

Disclaimer I am a maintainer of DataFusion

https://github.com/jorgecarleitao/arrow2 https://github.com/apache/arrow-datafusion https://github.com/apache/arrow-ballista https://github.com/pola-rs/polars https://github.com/duckdb/duckdb

Problem comes when you want to do this inside an async context since we couldn't block an async task. https://users.rust-lang.org/t/sync-function-invoking-async/43364/6 You might need to do it in another runtime/thread. It is not recommended to do this, but sometimes it is unavoidable while implementing a third-party trait. https://github.com/apache/arrow-datafusion/issues/3777 However, I believe this isn't a problem particular to tokio, or any specific runtime.