Type-check docstring code examples via companion files and unified sync script#2119
Open
jonathanhefner wants to merge 6 commits intomodelcontextprotocol:mainfrom
Open
Type-check docstring code examples via companion files and unified sync script#2119jonathanhefner wants to merge 6 commits intomodelcontextprotocol:mainfrom
jonathanhefner wants to merge 6 commits intomodelcontextprotocol:mainfrom
Conversation
The new script is a superset of the old one. It continues to handle the existing `<!-- snippet-source -->` markers in `README.v2.md`, and adds support for syncing code snippets into Python docstrings and markdown docs under `docs/`. New capabilities beyond the old script: - Region extraction from example files using `# region` / `# endregion` markers, so a single example file can provide multiple snippets - All source paths resolve relative to the repository root - Scans `src/**/*.py` and `docs/**/*.md` in addition to the README - Caches file contents and extracted regions for efficiency The marker format uses HTML comments (`<!-- snippet-source -->` / `<!-- /snippet-source -->`), which are invisible when rendered by `mkdocstrings` and do not interfere with `pymdownx.superfences` code fence parsing.
Add support for `<!-- snippet-source #region_name -->` markers that omit the companion file path. The path is derived from the target file's location using the mapping `src/X` → `examples/snippets/docstrings/X`. This eliminates line-length violations on marker lines in docstrings, since the full companion path no longer needs to be embedded in every marker. The full-path form continues to work for non-`src/` targets like markdown files.
Move inline code examples from docstrings in `src/mcp/` into standalone companion files at `examples/snippets/docstrings/mcp/`, mirroring the source tree structure. The `scripts/sync_snippets.py` script keeps the docstring content in sync with the companion files via `<!-- snippet-source #RegionName -->` markers. This ensures all docstring examples are checked by pyright and ruff, catching type errors and style drift that would otherwise go unnoticed in raw docstring text. The pattern follows the TypeScript SDK's approach of one companion file per source file, with each example in a named function whose parameters supply typed context. 19 companion files cover 42 code examples across the public API surface: `MCPServer`, `Client`, `ClientSession`, `Context`, `ResponseRouter`, `ServerTaskContext`, `ExperimentalTaskHandlers`, `ClientCredentialsOAuthProvider`, `PrivateKeyJWTOAuthProvider`, `SignedJWTParameters`, and others. All source markers use path-less form (`#RegionName`) — the companion path is derived automatically from the target file location. Region names follow `ClassName_methodName_variant` without abbreviation. Pyright execution environment for the companion files suppresses only "unused artifact" diagnostics inherent to example code (`reportUnusedFunction`, `reportUnusedVariable`, `reportAbstractUsage`, `reportUnusedClass`, `reportPrivateUsage`). All actual type-checking rules remain enabled.
Add a "Docstring Code Examples" section covering the companion file system: directory layout, `<!-- snippet-source #RegionName -->` markers, `ClassName_methodName_variant` naming, the function-parameter pattern for typed dependencies, and the prohibition on type-suppression comments inside regions.
`sync_snippets.py` also syncs snippet-source markers in `docs/**/*.md` and `README.v2.md`, not just `src/` docstrings. Add a short section after "Docstring Code Examples" noting that these files use explicit paths (path-less `#Region` markers are only supported in `src/` files).
The previous documentation had two flat sections where the docstring section contained all the detail and the markdown section referred to it vaguely. Reorganize into a `## Code Snippet System` intro covering shared concepts (marker format, region extraction, naming conventions, function wrappers, typed params, `# type: ignore` prohibition, pyright workflow) with `### Markdown Code Examples` and `### Docstring Code Examples` subsections covering only what is unique to each target type.
jonathanhefner
force-pushed
the
sync-snippets-script
branch
from
4551297 to
c0e1559
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Docstring code examples in
src/mcp/were previously raw text — invisible to pyright and ruff, so type errors and style drift went unnoticed. This PR introduces a system for keeping those examples in standalone, type-checked companion files that are synced back into the docstrings automatically.The old
update_readme_snippets.pyis replaced withsync_snippets.py, a superset that handlesREADME.v2.md,docs/**/*.md, andsrc/**/*.pydocstrings. It adds region extraction (# region/# endregionmarkers) so a single companion file can supply multiple snippets. For docstrings specifically,<!-- snippet-source #RegionName -->markers derive the companion path automatically from the target file's location (src/mcp/foo.py→examples/snippets/docstrings/mcp/foo.py), avoiding line-length violations from embedding full paths.42 code examples across the public API surface (
MCPServer,Client,ClientSession,Context,ResponseRouter, task support, OAuth providers, etc.) are extracted into 19 companion files underexamples/snippets/docstrings/mcp/, mirroring the source tree. Each example is wrapped in a named function with typed parameters, so pyright and ruff check them on every CI run — catching signature changes, renamed parameters, and moved imports at CI time rather than when a user copies a broken example.Conventions for the snippet system (region naming, function-parameter pattern,
# type: ignoreprohibition, editing workflow) are documented inCLAUDE.md.@Kludex, I was told you were thinking about moving
examples/snippets/todocs/snippets/. Would you prefer that docstring snippets live indocs/snippets/docstrings/? (Or anywhere else other thanexamples/snippets/docstrings/?)