Columnar / Start with the bytes

Data, from the
inside out.

Understand how analytical data is stored, organized and queried. Follow a real file from bytes to values, then explore table versions and execution plans.

Choose a layer to explore.

File format

Parquet

Columns, row groups, pages and encodings inside one file.

Read the guide →

Memory and interchange

Arrow

Typed arrays, validity, offsets and record batches across process boundaries.

Read the guide →

Table format

Iceberg

Snapshots and manifests select the files that make a table.

Read the guide →

Table format

Delta Lake

Ordered commits and checkpoints reconstruct table membership.

Read the guide →

Query engine

DuckDB

Turn a local Parquet file into an answer with SQL.

Read the guide →

Execution

Query engines

Compare DuckDB and DataFusion plans and follow Arrow batches.

Read the guide →

Files, tables and search

Lance

Trace fragments and page buffers, then compare exact and indexed vector searches.

Read the guide →

Compare the layers →

01 / File formats

Parquet

Columns, row groups, encodings and compression. One small penguin dataset makes the structure tangible.

LEARN

A file, explained

Step through real bytes and see how the pieces fit together.

Read the interactive guide →

EXPLORE

Your data, locally

Inspect schema, values and metadata. Files stay in your browser.

Use the explorer →

BUILD

The TypeScript reader

See the APIs, value model and limits of the parser behind these tools.

Read the API overview →

All Parquet resources →