Hacker News Reader

Comment by 4b11b4 | Hacker News Reader

The data would be sent to the browser

Um, my data lake is measured in Pb, not Tb. How is that going to work exactly?

DuckDB has certain optimizations which allows it to read only parts of a parquet file. It can also read remote file in streaming fashion so it does not have to wait for the entire file to be downloaded or to store a large amount of data in memory. Relevant documentation: https://duckdb.org/2021/06/25/querying-parquet.html

Howdy! I work at MotherDuck and DuckDB Labs (part time as a blogger). At MotherDuck, we have both client side and server side compute! So the initial reduction from PB/TB to GB/MB can happen server side, and the results can be sliced and diced at top speed in your browser!

Please spend a sentence or two explaining the server side filtering mechanism and linking to documentation! I would like to know the conditions required for streaming queries! From the sibling comment and a search of the docs it seems like this is a Parquet only feature, which seems pretty important to note!

Not specific to Ducks but S3 select https://docs.aws.amazon.com/AmazonS3/latest/userguide/select... can filter Parquet server side on S3 and is supported by some other object stores.

Parquet is designed with predicate push-down in mind. Partitions are laid out on disk, and then blocks within files are laid out so that consumers can very, very easily narrow in on which files they need to read, before doing anymore IO than a list, or a small metadata read.

Once you know what you are reading, many parquet/arrow libraries will support streaming reads/aggregations, so the client doesn’t need to load the whole working set in memory.

This only covers very simple min / max / sum cases.

For all others you'll need to download all columns you are filtering or selecting.

Does duckdb work with delta files?

DuckDB can use the HTTP range header trick agains parquet, which means a lot of analytical questions against multi-GB files can be answered by fetching only small portions of the overall data.

Here's a post about applying this same trick to SQLite from few years ago: https://phiresky.github.io/blog/2021/hosting-sqlite-database...

Reply on news.ycombinator.com