How-to guides#

This part of the project documentation focuses on a problem-oriented approach. We'll go over how to solve common tasks.

Reading data#

Reading public, published data needs no credentials.

• Finding volumes and books searches the catalogue, filters it, and lists every book in a volume.
• Reading a published book addresses a book, pins an edition, explores an entry before downloading it, and trims a query on the server.
• Converting and plotting covers the pandas, Polars, PyArrow and ScmRun converters, the verified content cache, and getting a chart on screen.
• Reading asynchronously covers the awaited facade, fetching several books at once, and how long a client should live.

Producing data#

Producing splits into two halves. Recording captures what would be published into a local bundle, so it needs no credentials. Replay performs the writes and needs bookshelf:write, so run bookshelf auth login first.

Both guides record, and both show the replay step without running it.

• Publishing a book frames a book, registers a derived resource with its lineage, and reads the recorded manifest back before anything is written.
• Cataloguing external data covers pointers to data the platform does not store, batch registration, partial failure, and deduplication.