simonw / datasette

GitHub metadata describes the project as "An open source multi-tool for exploring and publishing data", and gives datasette.io as its homepage. The primary language is Python; the byte breakdown lists about 2.47 million Python bytes alongside JavaScript, HTML, CSS, Shell, C, Dockerfile and Just files. Repository topics include asgi, automatic-api, csv, datasets, datasette, docker, json, python, sql and sqlite. The license is Apache-2.0, and a LICENSE file sits at the repository root. Structurally, the analyzed snapshot has 333 files: a datasette package (135 files) plus tests (102), docs (50), demos (7) and a few other top-level directories. The latest listed release is 1.0a39 (published 2026-09-11), and the repository had 11,460 stars on 2026-09-14.

The only launch code shown in datasette/__main__.py is a main guard that calls cli():

if __name__ == "__main__":
    cli()

A bounded static execution path records that __main__ hands off to cli in datasette/cli.py and stops there as a leaf. datasette/app.py and datasette/cli.py are both flagged as likely executable entrypoints (filename heuristic). app.py is labeled an entry/orchestration candidate with 356 incoming and 197 outgoing static calls. It contains the two large classes Datasette (lines 394-3037) and DatasetteRouter (lines 3040-3314). Note the limits: every traced execution path starting at app.py ends at an unresolved external boundary such as logging.getLogger, contextvars.ContextVar, collections.namedtuple or Setting, so deeper runtime wiring is not established.

• Top-level layout separates the application package datasette from the docs and tests subsystems.
• datasette/utils/__init__.py is the most connected Python module (519 incoming / 53 outgoing calls) and defines CustomJSONEncoder and sqlite_timelimit.
• datasette/utils/asgi.py is next (369 incoming) and defines Request, AsgiLifespan, AsgiStream and AsgiWriter, plus exception classes like NotFound and Forbidden.
• datasette/views/table.py has the highest outgoing degree among the listed Python modules (283 out / 46 in), which suggests it coordinates many modules; it defines TableContext and Row.
• datasette/__init__.py (348 in / 0 out) and datasette/resources.py (227 in / 0 out) look like leaf/data-boundary modules; resources.py defines DatabaseResource, TableResource and QueryResource.
• Front-end assets live under datasette/static; the analyzer reports the highest overall degree for edit-tools.js (808/808) and table.js (83/83), but JavaScript analysis is only a bounded heuristic.

Manifest evidence in this packet covers JavaScript tooling only. package.json, one of two detected build/dependency manifests, lists runtime dependencies @codemirror/lang-sql ^6.3.3, codemirror ^6.0.1, rollup ^3.30.0, @rollup/plugin-node-resolve ^15.0.1 and @rollup/plugin-terser ^0.1.0, plus prettier ^3.0.0 as a dev dependency. Dependency parsing has a hard limit here: only requirements-style manifests were parsed, and pyproject.toml dependency tables were not, so the Python package's own dependencies are not established by this analysis. The tests subsystem contains 102 test files. Sampled file tests/test_api_write.py defines helpers ds_write and write_token and tests such as test_structured_writes_require_ordinary_tables and test_base64_write_api_insert_upsert_update_decode_blobs. tests/build_small_spatialite_db.py is a project-important likely entrypoint whose __main__ guard calls generate_it("spatialite.db"), generating a SQLite test database.

1. README.md - marked project-level important; the usual first stop.
2. pyproject.toml and package.json - packaging and tooling manifests, both flagged important.
3. datasette/__main__.py, then datasette/cli.py - the launch chain; the static path runs from the __main__ guard to cli.
4. datasette/app.py - central orchestration module, holding the Datasette and DatasetteRouter classes.
5. datasette/utils/__init__.py and datasette/utils/asgi.py - the highest incoming-call counts among the listed Python modules (519 and 369; edit-tools.js reports 808 incoming overall).
6. datasette/views/table.py - view logic with the highest outgoing call degree among Python modules (283; edit-tools.js reports 808 outgoing overall), a good window into how requests reach the rest of the code.

When changing code, start tracing at the detected entrypoint datasette/app.py and check downstream effects by hand, because dynamic behaviour is not resolved statically.

• Static only: reflection, dynamic imports, monkey-patching, generated code, framework runtime wiring and dynamic dispatch are not resolved, so how plugins are discovered and loaded cannot be established from this packet.
• Install, run and build inference is partial; installation commands are not verified, and the packet contains no runtime or execution results.
• Of 21,338 static relations, 15,850 are external or unresolved and 5,488 resolve locally.
• Two files were inventoried but not parsed (oversized, minified or vendored); bounded analysis also skips files over 400 kB, minified bundles and vendored directories.
• Dependency coverage is partial: pyproject.toml dependency tables were not parsed.
• README claim extraction is line-based and may capture code lines, so any README-derived statement is author_claimed at best.

• Datasette (datasette/app.py, lines 394-3037) and DatasetteRouter (lines 3040-3314): the two large classes in app.py, which the analyzer labels an entry/orchestration candidate.
• DatabaseResource, TableResource, QueryResource (datasette/resources.py): classes in a module labeled a leaf/data-boundary candidate.
• Request, AsgiLifespan, AsgiStream, AsgiWriter and the NotFound/Forbidden/BadRequest/PayloadTooLarge exceptions (datasette/utils/asgi.py): classes in the module with the second-highest incoming call count among the listed Python modules (369).
• CustomJSONEncoder and sqlite_timelimit (datasette/utils/__init__.py) plus TableContext and Row (datasette/views/table.py): symbols in the listed Python modules with the most incoming calls (519) and the most outgoing calls (283).