Blosc/python-blosc2

The b2view command-line TUI (installed as a project script) allows to browse visually inspect a .b2nd or .b2z/.b2d file/dir. It has different views (tree, meta, vlmeta, data), but more love is needed to make it working in a smoother way and/or adding new features. Also, b2view is not mentioned anywhere in doc/. Users who install python-blosc2 and want to visually inspect a .b2nd or .b2z file have no way to discover it from the docs. Suggested work: Use real-world datasets (e.g. coming from parquet files in Hugging Face), and assess navegability on them. Try to improve visual aspects and browsability in general. Add a short section to doc/getting_started/overview.rst (or a new tools.rst) describing b2view, what file types it supports, and how to launch it Add a screenshot or short ASCII rendering of the TUI output Link it from the parquet-to-blosc2 docs (issue 3 above) since that tool and b2view are often used together

examples/ctable/real_world.py exists but is a fairly raw script. There is no self-contained example that clearly shows what performance gain SUMMARY indexes provide over a full scan, or how block size affects that gain — which is the most common question users will have after enabling auto-indexing. Suggested work: Write bench/ctable/summary_index_perf.py that: Generates a synthetic CTable with a few million rows and numeric columns Runs the same where() query three ways: no index, SUMMARY at chunk granularity, SUMMARY at block granularity Prints a clean results table (rows scanned, time, speedup) Includes comments explaining the trade-offs Ideally, it should work without any external dataset so it can be run immediately after install, but using an accessible dataset is also an option.

New contributors and users frequently confuse chunks, blocks, frames, and superchunks — terms that appear throughout the API but are never defined in one place. The getting-started section has no glossary. Suggested work: Create doc/getting_started/glossary.rst (and link it from doc/getting_started/index.rst) defining at minimum: SChunk, frame, chunk, block, subblock Compression parameters (cparams): codec, clevel, filters NDArray vs CTable vs SChunk — when to use each Index types: SUMMARY, BUCKET, PARTIAL, FULL, OPSI — one-sentence each Each term should link to the relevant API reference page.

The docstrings for CTable.where() and CTable.sort_by() describe the API well but contain no runnable Examples: section. New contributors exploring the API in a REPL or reading rendered API docs get no concrete illustration of the lazy-view behaviour introduced in this release. Suggested work: Add an Examples section to both docstrings (NumPy docstring format) showing: where(): filter with a simple predicate, with columns=, and with a compound condition sort_by(): sort a plain table, sort a filtered view (demonstrating that it stays lazy until iterated), and multi-key sort Both should include print() calls that make the lazy-vs-eager distinction visible

12.batcharray.ipynb covers BatchArray basics but there is no tutorial for ListArray — the newer, more flexible varlen column type. Users working with nested or ragged data (e.g. GPS waypoints, tags, embeddings of varying length) have no worked example to follow. Suggested work: Extend 12.batcharray.ipynb or add 12b.listarray.ipynb covering: Creating a ListArray with item_spec Appending, extending, and fancy-indexing Embedding a ListArray as a column inside a CTable (using the dataclass schema API) Arrow round-trip (from_arrow / to_arrow) When to prefer ListArray over BatchArray