If you query a Parquet file from your lake via DuckDB-in-browser, does DuckDB run in WASM on the web client and pull the compressed parquet to your browser where it is decompressed? Or are you connecting some DuckDB on the web client to some DuckDB component on a server somewhere?
I presume yes to the first and no to the second but just checking I have my mental model correct.
Which you can query anywhere using a ClickHouse local database engine.
When you do a query, it downloads only the required ranges of required columns from the server and does computations on your machine.
In the same way you can query any data sources with ClickHouse, either local or remote, which is a strict superset in power and usability than duckdb.
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.
For all others you'll need to download all columns you are filtering or selecting.
Here's a post about applying this same trick to SQLite from few years ago: https://phiresky.github.io/blog/2021/hosting-sqlite-database...