From 6b3a1f973419f040017b0a69bb0b974dc80abbe3 Mon Sep 17 00:00:00 2001 From: Claude Date: Tue, 17 Feb 2026 22:58:11 +0000 Subject: [PATCH 001/472] Add column types design document Analyzes three design options for how Datasette could support column types (datetime, file references, URLs, etc.) that cross-cut rendering, editing, validation, and faceting. Recommends a hybrid approach: use the existing metadata_columns system as the foundation with a "type" metadata key convention, and optionally layer a ColumnType plugin registry on top for plugins that want to provide a full bundled type experience. https://claude.ai/code/session_01TitjhUZrRR4hzS3V5uoQoi --- docs/design/column-types.md | 319 ++++++++++++++++++++++++++++++++++++ 1 file changed, 319 insertions(+) create mode 100644 docs/design/column-types.md diff --git a/docs/design/column-types.md b/docs/design/column-types.md new file mode 100644 index 00000000..f42ef2d8 --- /dev/null +++ b/docs/design/column-types.md @@ -0,0 +1,319 @@ +# Column Types Design + +## Current State + +Datasette knows three things about columns today: + +1. **SQLite schema info** - stored in `catalog_columns` (name, SQLite type affinity, notnull, default, pk, hidden) +2. **Column metadata** - stored in `metadata_columns` (arbitrary key/value pairs per column, currently only `description` is used) +3. **Foreign keys** - stored in `catalog_foreign_keys` (detected from schema, rendered as links automatically) + +Rendering customization happens entirely through the `render_cell` plugin hook: plugins inspect the value/column/table and return HTML if they want to handle it. There is no structured way for Datasette or its plugins to declare "this column holds timestamps" or "this column holds file IDs" and have the system respond to that declaration across multiple concerns. + +## The Problem + +Several features need to know *what kind of data* a column holds, beyond what SQLite's type affinity tells us: + +- **datasette-files** needs columns that hold `df-{ULID}` file references to render as file previews, offer file upload widgets on edit forms, and validate on insert +- **Timestamps** need to render with locale-aware formatting, offer date-picker widgets on forms, enable date faceting, and sort correctly +- **URLs** could render as clickable links, be validated on input, and be previewed +- **Email addresses**, **JSON**, **Markdown**, **geographic coordinates** - same pattern + +Each of these cross-cuts multiple concerns: rendering, editing, validation, faceting, filtering, API representation. The `render_cell` hook only covers one of those concerns (HTML table rendering). + +## What Could Column Types Do? + +A column type is a named bundle of behaviors that applies to a column. Here's what each behavior might look like: + +### Rendering (display) +- HTML table cell rendering (today's `render_cell`) +- Row detail page rendering (possibly different from table view - a map for coordinates, an image preview for files) +- API representation (should a timestamp column include an ISO 8601 string AND a Unix epoch? Should a file column include the file metadata inline?) + +### Editing (forms) +- What widget to show when inserting/updating a row (date picker, file upload, textarea for markdown, dropdown for enums) +- What the insert API should accept (a file upload multipart for file columns, ISO 8601 strings for timestamps) + +### Validation +- Reject values that don't match the expected format on write +- Provide clear error messages ("Expected an ISO 8601 datetime, got 'next tuesday'") + +### Faceting and Filtering +- Custom facet types (date faceting already exists as a special case in `facets.py`) +- Custom filter operations ("before", "after" for dates; "contains domain" for URLs) + +### Type Coercion +- Automatically convert input values (parse "2024-01-15" into the appropriate storage format, resize an uploaded image) + +### Sorting +- Custom sort behavior (natural sort for version strings, locale-aware sort for text) + +### Column Header / Schema +- Expose the column type in the JSON API schema (`"type": "datetime"` rather than just `"type": "TEXT"`) +- Show a type indicator in the table UI (an icon or label next to the column name) + +## Design Options + +### Option A: Column metadata is enough (extend what exists) + +The `metadata_columns` table already supports arbitrary key/value pairs per column. A "column type" is just `key="type", value="datetime"`. Plugins register themselves as handlers for specific type values and implement whatever behaviors they need. + +**How it works:** + +```yaml +databases: + mydb: + tables: + events: + columns: + created_at: + description: "When the event was created" + type: datetime + photo: + description: "Event photo" + type: file + type:source: event-photos # type-specific config +``` + +Stored in `metadata_columns` as: +``` +(mydb, events, created_at, type, datetime) +(mydb, events, created_at, description, "When the event was created") +(mydb, events, photo, type, file) +(mydb, events, photo, type:source, event-photos) +``` + +Plugins check `await datasette.get_column_metadata(db, table, col)` to see if a column has a type they handle. The existing `render_cell` hook already receives the `datasette` instance, so plugins can look up column metadata there. + +Runtime configuration works via `await datasette.set_column_metadata(db, table, col, "type", "datetime")` - already implemented. + +**New hooks needed:** + +```python +# Form widget for editing +def edit_cell_widget(column, table, database, datasette, request, current_value): + """Return HTML for a form widget for this column""" + +# Validation on write +def validate_cell(column, table, database, datasette, value): + """Return None if valid, or an error message string""" + +# Custom facet registration (already somewhat exists via register_facet_classes) +``` + +**Pros:** +- No new concepts - just more metadata keys and more hooks +- Runtime modification already works (`set_column_metadata`) +- Incremental - can adopt one behavior at a time +- Plugin authors already understand metadata + +**Cons:** +- No formal contract for what a "type" provides - each plugin picks and chooses which hooks to implement +- Column metadata is flat key/value (strings only) - complex type config needs conventions (like `type:source` prefix, or JSON-in-a-string) +- Discovery is hard - there's no `get_registered_column_types()` to enumerate what's available +- No way to validate that a column's type value actually corresponds to a registered plugin + +### Option B: First-class column type registry + +Add a new `register_column_types` plugin hook that lets plugins formally declare column types with all their behaviors in one place. + +**How it works:** + +```python +from datasette import hookimpl +from datasette.column_types import ColumnType + +class DateTimeColumnType(ColumnType): + name = "datetime" + label = "Date/Time" + description = "ISO 8601 datetime values" + + def render_cell(self, value, row, column, table, database, request): + if value is None: + return None + dt = parse_datetime(value) + return Markup(f'') + + def render_cell_detail(self, value, row, column, table, database, request): + """Richer rendering for the row detail page""" + ... + + def edit_widget(self, value, column, table, database, request): + return Markup(f'') + + def validate(self, value, column, table, database): + try: + parse_datetime(value) + return None + except ValueError: + return f"Invalid datetime: {value}" + + def facet_class(self): + return DateTimeFacet + + def api_schema(self): + return {"type": "string", "format": "date-time"} + +@hookimpl +def register_column_types(): + return [DateTimeColumnType()] +``` + +Type assignments still live in column metadata (`metadata_columns` with `key="type"`), but now the type value must correspond to a registered `ColumnType`. Datasette can validate this and provide a UI for selecting from available types. + +**Configuration:** + +```yaml +databases: + mydb: + tables: + events: + columns: + created_at: + type: datetime + photo: + type: file + type_config: + source: event-photos + accept: "image/*" +``` + +The `type_config` is passed to the ColumnType instance, which can define what config it accepts. + +**Runtime UI:** + +Because types are registered, Datasette can offer a "Set column type" action: +1. User clicks column header -> "Configure column type" +2. Datasette shows a dropdown of registered types (datetime, file, url, markdown, ...) +3. User selects one -> type-specific config form appears (the ColumnType can provide this) +4. Saves to `metadata_columns` + +**Pros:** +- Clear contract - a ColumnType is a concrete thing with known capabilities +- Discovery - `datasette.column_types` gives you the full list +- UI-friendly - Datasette can build configuration UIs automatically +- Composition - a ColumnType bundles all related behaviors together +- Validation - Datasette can reject unknown type names +- Type-specific configuration is structured (not flat key/value hacks) + +**Cons:** +- New concept to learn +- More opinionated - may not suit all plugin patterns +- ColumnType methods become an API surface that's harder to evolve +- Could feel heavyweight for simple cases (just wanting to tweak cell rendering) + +### Option C: Hybrid - metadata-driven with optional type registry + +Use column metadata as the storage and source of truth, but add an optional registry for plugins that want to provide a full "type" experience. + +**How it works:** + +Column metadata remains the foundation. The `type` key in column metadata is the link between the two systems. If a plugin registers a ColumnType with `name="datetime"`, and a column has `metadata type=datetime`, then Datasette delegates to that ColumnType for all registered behaviors. + +But plugins can also just implement individual hooks (`render_cell`, `validate_cell`, etc.) and check column metadata themselves - the registry is not required. + +```python +# Full type registration (optional) +@hookimpl +def register_column_types(): + return [DateTimeColumnType()] + +# OR just use individual hooks (also fine) +@hookimpl +def render_cell(value, column, table, database, datasette, request): + metadata = await datasette.get_column_metadata(database, table, column) + if metadata.get("type") == "datetime": + return format_datetime(value) +``` + +**Datasette's behavior when rendering a cell:** +1. Check if the column has `type` metadata +2. If yes, and a registered ColumnType matches, call its `render_cell` +3. Then run `render_cell` plugin hooks as today (plugins can override or augment) +4. Fall back to default rendering + +**Pros:** +- Backward compatible - existing render_cell plugins keep working +- Incremental adoption - plugins can start with metadata-only, evolve to ColumnType later +- Runtime config works the same either way (it's all metadata) +- The registry is additive, not required + +**Cons:** +- Two ways to do things - potential confusion +- Ordering/priority between ColumnType rendering and render_cell hooks needs clear rules + +## Key Design Questions + +### 1. Should column types be auto-detected? + +Datasette could infer types from: +- SQLite declared types (`DATETIME`, `TIMESTAMP` in the schema) +- Column names (`created_at`, `updated_at` suggest timestamps; `email` suggests email) +- Data patterns (first N rows match ISO 8601 format) + +Auto-detection is convenient but can be wrong. Suggestion: auto-detect as *suggestions* shown in the UI, but don't apply without user confirmation. Or: auto-detect for well-known SQLite type declarations (`DATETIME` -> datetime type) since the schema author's intent is clear. + +### 2. Where does type configuration live? + +All options above use `metadata_columns` for storage. The question is whether `type_config` should also be in `metadata_columns` (as prefixed keys like `type:source`) or in a separate storage mechanism. + +Using `metadata_columns` with a convention like `type_config:key` keeps things simple. Alternatively, store a single JSON blob: `key="type_config", value='{"source": "event-photos"}'`. + +Recommendation: store type config as a JSON string in `metadata_columns` with `key="type_config"`. This avoids namespace pollution and lets types define arbitrarily structured config. + +### 3. How does this interact with the edit/insert API? + +Datasette's table create/insert API (`POST /-/create`, `POST /db/table/-/insert`) currently accepts raw values. With column types: + +- The API could accept type-appropriate values and coerce them (accept `"2024-01-15"` for a datetime column) +- File columns would need multipart upload support +- Validation errors need to surface clearly in API responses + +This is where the `validate_cell` hook (or ColumnType.validate method) becomes important - it runs before the write and returns structured errors. + +### 4. Should types affect the JSON API response? + +Today the API returns raw SQLite values. Options: +- Add a `column_types` key to the response metadata listing the assigned types +- Optionally transform values (format datetimes, inline file metadata) +- Add a schema endpoint describing column types + +Recommendation: expose type information in response metadata, but don't transform values by default (breaking change). Offer an `?_extra=typed_values` or `?_shape=typed` opt-in. + +### 5. How do types interact with faceting? + +The existing `register_facet_classes` hook lets plugins add facet types. If a column has `type=datetime`, Datasette could auto-suggest the date facet instead of requiring `?_facet_date=col`. This is a natural fit for the ColumnType registry (Option B/C) where a type can declare its preferred facet class. + +## Recommendation + +**Option C (hybrid)** seems like the best fit for Datasette's design philosophy: + +1. **Column metadata is the foundation.** It already exists, already supports runtime modification, and is already persisted. Making the `type` metadata key a first-class convention (documented, supported in YAML, shown in the UI) is the minimal viable step. + +2. **The ColumnType registry is the ergonomic layer.** It makes plugin development nicer and enables Datasette to build UIs for type selection and configuration. But it's optional - simple plugins can just check metadata in their hooks. + +3. **Incremental delivery.** Ship the metadata convention and a few new hooks first (`edit_cell_widget`, `validate_cell`). Add the ColumnType registry later when patterns solidify. + +### Concrete next steps + +1. **Document the `type` metadata key convention.** Even before adding new hooks, plugins can start using `set_column_metadata(db, table, col, "type", "datetime")` and checking it in `render_cell`. + +2. **Add `type` to the YAML metadata schema.** Currently `columns:` in metadata YAML only supports description strings. Extend it to support: + ```yaml + columns: + created_at: + description: "When created" + type: datetime + # Short form still works for description-only + name: "The person's name" + ``` + +3. **Add a `column_actions` plugin hook.** This lets plugins add actions to column headers (like "Set type to datetime") mirroring the existing `table_actions` pattern. + +4. **Add a `validate_cell` hook.** Called before writes, receives column metadata. + +5. **Add an `edit_cell_widget` hook.** Returns form HTML for the table editing UI. + +6. **Ship built-in datetime and URL column types** as either core code or a bundled plugin, to prove the pattern works. + +7. **Later: add `register_column_types` hook** once the per-behavior hooks are proven and the ColumnType base class API stabilizes. From 7a66456615cad38d9e70267a14ca30dcc4bca701 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Fri, 20 Feb 2026 11:19:19 -0800 Subject: [PATCH 002/472] black --version --- .github/workflows/test.yml | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/.github/workflows/test.yml b/.github/workflows/test.yml index b1ba3232..a0f5477b 100644 --- a/.github/workflows/test.yml +++ b/.github/workflows/test.yml @@ -34,7 +34,9 @@ jobs: # And the test that exceeds a localhost HTTPS server tests/test_datasette_https_server.sh - name: Black - run: black --check . + run: | + black --version + black --check . - name: Ruff run: ruff check datasette tests - name: Check if cog needs to be run From 2f0e64df681c7bf65e8ce3065380be36a4ccd266 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Fri, 20 Feb 2026 11:24:52 -0800 Subject: [PATCH 003/472] black==26.1.0 I'm getting CI failures for Black, maybe this will help --- pyproject.toml | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/pyproject.toml b/pyproject.toml index d9ef2a73..2ab2ce10 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -61,7 +61,7 @@ dev = [ "pytest-xdist>=2.2.1", "pytest-asyncio>=1.2.0", "beautifulsoup4>=4.8.1", - "black==25.11.0", + "black==26.1.0", "blacken-docs==1.20.0", "pytest-timeout>=1.4.2", "trustme>=0.7", From 6a2c27b15b300ba1b924ce00a61532943482392e Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Fri, 20 Feb 2026 11:28:39 -0800 Subject: [PATCH 004/472] blacken-docs --- docs/plugin_hooks.rst | 13 ++++--------- docs/spatialite.rst | 6 ++---- docs/testing_plugins.rst | 8 ++------ 3 files changed, 8 insertions(+), 19 deletions(-) diff --git a/docs/plugin_hooks.rst b/docs/plugin_hooks.rst index 068469a8..fa335368 100644 --- a/docs/plugin_hooks.rst +++ b/docs/plugin_hooks.rst @@ -1074,11 +1074,9 @@ You can also return an async function, which will be awaited on startup. Use thi async def inner(): db = datasette.get_database() if "my_table" not in await db.table_names(): - await db.execute_write( - """ + await db.execute_write(""" create table my_table (mycol text) - """ - ) + """) return inner @@ -1561,7 +1559,6 @@ The resolver will automatically apply the most specific rule. from datasette import hookimpl from datasette.permissions import PermissionSQL - TRUSTED = {"alice", "bob"} @@ -2261,8 +2258,7 @@ This example logs events to a ``datasette_events`` table in a database called `` def startup(datasette): async def inner(): db = datasette.get_database("events") - await db.execute_write( - """ + await db.execute_write(""" create table if not exists datasette_events ( id integer primary key, event_type text, @@ -2270,8 +2266,7 @@ This example logs events to a ``datasette_events`` table in a database called `` actor text, properties text ) - """ - ) + """) return inner diff --git a/docs/spatialite.rst b/docs/spatialite.rst index fbe0d75f..c93c1e00 100644 --- a/docs/spatialite.rst +++ b/docs/spatialite.rst @@ -90,12 +90,10 @@ Here's a recipe for taking a table with existing latitude and longitude columns, "SELECT AddGeometryColumn('museums', 'point_geom', 4326, 'POINT', 2);" ) # Now update that geometry column with the lat/lon points - conn.execute( - """ + conn.execute(""" UPDATE museums SET point_geom = GeomFromText('POINT('||"longitude"||' '||"latitude"||')',4326); - """ - ) + """) # Now add a spatial index to that column conn.execute( 'select CreateSpatialIndex("museums", "point_geom");' diff --git a/docs/testing_plugins.rst b/docs/testing_plugins.rst index fc1aa6f6..b0713e7c 100644 --- a/docs/testing_plugins.rst +++ b/docs/testing_plugins.rst @@ -233,15 +233,11 @@ As an example, here's a very simple plugin which executes an HTTP response and r async def fetch_url(datasette, request): if request.method == "GET": - return Response.html( - """ + return Response.html("""
-
""".format( - request.scope["csrftoken"]() - ) - ) + """.format(request.scope["csrftoken"]())) vars = await request.post_vars() url = vars["url"] return Response.text(httpx.get(url).text) From c96dc5ce2656607b9e81743acf600f8fd5f6a795 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Wed, 25 Feb 2026 16:32:45 -0800 Subject: [PATCH 005/472] register_token_handler() plugin hook for custom API token backends (#2650) Closes #2649 * Add register_token_handler plugin hook for pluggable token backends Adds a new register_token_handler hook that allows plugins to provide custom token creation and verification backends. This enables plugins like datasette-oauth to issue tokens without depending on specific backend plugins like datasette-auth-tokens. Key changes: - New datasette/tokens.py with TokenHandler base class and SignedTokenHandler (the default signed-token implementation moved here) - New register_token_handler hookspec in hookspecs.py - Datasette.create_token() is now async and delegates to token handlers - New Datasette.verify_token() method tries all handlers in sequence - handler= parameter on create_token() to select a specific backend - TokenHandler exported from datasette package for plugin use - Fixed actor_from_request loop to await all coroutines (avoids warnings) * Add documentation and hook test for register_token_handler Fixes CI failures: the new hook needs a section in docs/plugin_hooks.rst (checked by test_plugin_hooks_are_documented) and a test_hook_* function in test_plugins.py (checked by test_plugin_hooks_have_tests). * Register tokens module as separate default plugin Instead of re-exporting hookimpls from default_permissions/__init__.py, register datasette.default_permissions.tokens as its own DEFAULT_PLUGINS entry. Cleaner and avoids confusing import-for-side-effect patterns. * Replace restrict_x params with TokenRestrictions dataclass Consolidates the three separate restrict_all, restrict_database, and restrict_resource parameters into a single TokenRestrictions dataclass. Cleaner API surface for both Datasette.create_token() and TokenHandler.create_token(). Also clarifies docs re: default handler selection via pluggy ordering. * Add builder methods to TokenRestrictions Adds allow_all(), allow_database(), and allow_resource() methods that return self for chaining. Callers no longer need to manipulate nested dicts directly: restrictions = (TokenRestrictions() .allow_all("view-instance") .allow_database("mydb", "create-table") .allow_resource("mydb", "mytable", "insert-row")) * docs: add 1.0a25 upgrade guide section for create_token() signature change Ref: https://github.com/simonw/datasette/issues/2649#issuecomment-3962639393 * docs: note that create_token() is now async in upgrade guide * docs: update internals, plugin_hooks, authentication for new token API - internals.rst: new async create_token() signature with restrictions and handler params, add TokenRestrictions reference docs - plugin_hooks.rst: show full create_token signature in TokenHandler example, note list returns and error cases - authentication.rst: cross-reference TokenRestrictions from the restrictions section * style: apply black formatting to token handler files * docs: fix RST heading underline length in internals.rst * tests: add restrictions round-trip and expiration tests for token handler Covers allow_database/allow_resource builders, _r payload encoding, and token_expires in verified actors. Coverage 76% -> 90%. * tests: add test for signed tokens disabled * fix: add TokenRestrictions TYPE_CHECKING import to fix ruff F821 * docs: regenerate plugins.rst with cog * docs: reformat code blocks in plugin_hooks.rst with blacken-docs * docs: add await .verify_token() to internals.rst * tests: rewrite register_token_handler test to use real plugin handler Adds a HardcodedTokenHandler to the test plugins dir that creates tokens like dstok_hardcoded_token_1. The test now exercises creating tokens via the default handler (which is the plugin's hardcoded one), by explicitly naming the hardcoded handler, and by explicitly naming the signed handler -- then verifies each token round-trips correctly. * tests: clarify test_token_handler_via_http tests the default signed handler * fix: use handler="signed" explicitly where signed tokens are expected The HardcodedTokenHandler in my_plugin.py gets globally registered, so create_token() without a handler name picks it up as the default. Fix the create-token view, CLI, and tests to explicitly request the signed handler where they depend on signed token behavior. * fix: use handler="signed" in test_create_table_permissions https://claude.ai/code/session_013cQFiDQjYRrRBH2biFfKuS --- datasette/__init__.py | 1 + datasette/app.py | 102 +++++--- datasette/cli.py | 30 ++- datasette/default_permissions/__init__.py | 1 - datasette/default_permissions/tokens.py | 85 ++---- datasette/hookspecs.py | 5 + datasette/plugins.py | 1 + datasette/tokens.py | 180 +++++++++++++ datasette/views/special.py | 34 +-- docs/authentication.rst | 1 + docs/internals.rst | 81 ++++-- docs/plugin_hooks.rst | 59 +++++ docs/plugins.rst | 11 +- docs/upgrade_guide.md | 40 +++ tests/fixtures.py | 1 + tests/plugins/my_plugin.py | 27 ++ tests/test_api_write.py | 9 +- tests/test_permissions.py | 2 +- tests/test_plugins.py | 32 +++ tests/test_token_handler.py | 301 ++++++++++++++++++++++ 20 files changed, 839 insertions(+), 164 deletions(-) create mode 100644 datasette/tokens.py create mode 100644 tests/test_token_handler.py diff --git a/datasette/__init__.py b/datasette/__init__.py index 47d2b4f6..eb18e59e 100644 --- a/datasette/__init__.py +++ b/datasette/__init__.py @@ -1,6 +1,7 @@ from datasette.permissions import Permission # noqa from datasette.version import __version_info__, __version__ # noqa from datasette.events import Event # noqa +from datasette.tokens import TokenHandler, TokenRestrictions # noqa from datasette.utils.asgi import Forbidden, NotFound, Request, Response # noqa from datasette.utils import actor_matches_allow # noqa from datasette.views import Context # noqa diff --git a/datasette/app.py b/datasette/app.py index 6efaa430..2df6e4e8 100644 --- a/datasette/app.py +++ b/datasette/app.py @@ -7,6 +7,7 @@ from typing import TYPE_CHECKING, Any, Dict, Iterable, List if TYPE_CHECKING: from datasette.permissions import Resource + from datasette.tokens import TokenRestrictions import asgi_csrf import collections import dataclasses @@ -713,44 +714,70 @@ class Datasette: """ return _in_datasette_client.get() - def create_token( + def _token_handlers(self): + """Collect all registered token handlers from plugins.""" + from datasette.tokens import TokenHandler + + handlers = [] + for result in pm.hook.register_token_handler(datasette=self): + if isinstance(result, TokenHandler): + handlers.append(result) + elif isinstance(result, list): + handlers.extend(h for h in result if isinstance(h, TokenHandler)) + return handlers + + async def create_token( self, actor_id: str, *, expires_after: int | None = None, - restrict_all: Iterable[str] | None = None, - restrict_database: Dict[str, Iterable[str]] | None = None, - restrict_resource: Dict[str, Dict[str, Iterable[str]]] | None = None, - ): - token = {"a": actor_id, "t": int(time.time())} - if expires_after: - token["d"] = expires_after + restrictions: "TokenRestrictions | None" = None, + handler: str | None = None, + ) -> str: + """ + Create an API token for the given actor. - def abbreviate_action(action): - # rename to abbr if possible - action_obj = self.actions.get(action) - if not action_obj: - return action - return action_obj.abbr or action + Uses the first registered token handler by default, or a specific + handler if ``handler`` is provided (matched by handler name). - if expires_after: - token["d"] = expires_after - if restrict_all or restrict_database or restrict_resource: - token["_r"] = {} - if restrict_all: - token["_r"]["a"] = [abbreviate_action(a) for a in restrict_all] - if restrict_database: - token["_r"]["d"] = {} - for database, actions in restrict_database.items(): - token["_r"]["d"][database] = [abbreviate_action(a) for a in actions] - if restrict_resource: - token["_r"]["r"] = {} - for database, resources in restrict_resource.items(): - for resource, actions in resources.items(): - token["_r"]["r"].setdefault(database, {})[resource] = [ - abbreviate_action(a) for a in actions - ] - return "dstok_{}".format(self.sign(token, namespace="token")) + Pass a :class:`TokenRestrictions` to limit which actions the token + can perform. + """ + handlers = self._token_handlers() + if not handlers: + raise RuntimeError("No token handlers are registered") + + if handler is not None: + matched = [h for h in handlers if h.name == handler] + if not matched: + available = [h.name for h in handlers] + raise ValueError( + f"Token handler {handler!r} not found. " + f"Available handlers: {available}" + ) + chosen = matched[0] + else: + chosen = handlers[0] + + return await chosen.create_token( + self, + actor_id, + expires_after=expires_after, + restrictions=restrictions, + ) + + async def verify_token(self, token: str) -> dict | None: + """ + Verify an API token by trying all registered token handlers. + + Returns an actor dict from the first handler that recognizes the + token, or None if no handler accepts it. + """ + for token_handler in self._token_handlers(): + result = await token_handler.verify_token(self, token) + if result is not None: + return result + return None def get_database(self, name=None, route=None): if route is not None: @@ -2159,10 +2186,13 @@ class DatasetteRouter: # Handle authentication default_actor = scope.get("actor") or None actor = None - for actor in pm.hook.actor_from_request(datasette=self.ds, request=request): - actor = await await_me_maybe(actor) - if actor: - break + results = pm.hook.actor_from_request(datasette=self.ds, request=request) + for result in results: + result = await await_me_maybe(result) + if result and actor is None: + actor = result + # Don't break β€” we must await all coroutines to avoid + # "coroutine was never awaited" warnings scope_modifications["actor"] = actor or default_actor scope = dict(scope, **scope_modifications) diff --git a/datasette/cli.py b/datasette/cli.py index 121911ab..b473fbb7 100644 --- a/datasette/cli.py +++ b/datasette/cli.py @@ -832,21 +832,23 @@ def create_token( err=True, ) - restrict_database = {} - for database, action in databases: - restrict_database.setdefault(database, []).append(action) - restrict_resource = {} - for database, resource, action in resources: - restrict_resource.setdefault(database, {}).setdefault(resource, []).append( - action - ) + from datasette.tokens import TokenRestrictions - token = ds.create_token( - id, - expires_after=expires_after, - restrict_all=alls, - restrict_database=restrict_database, - restrict_resource=restrict_resource, + restrictions = TokenRestrictions() + for action in alls: + restrictions.allow_all(action) + for database, action in databases: + restrictions.allow_database(database, action) + for database, resource, action in resources: + restrictions.allow_resource(database, resource, action) + + token = run_sync( + lambda: ds.create_token( + id, + expires_after=expires_after, + restrictions=restrictions, + handler="signed", + ) ) click.echo(token) if debug: diff --git a/datasette/default_permissions/__init__.py b/datasette/default_permissions/__init__.py index 40373fa7..4ebe6147 100644 --- a/datasette/default_permissions/__init__.py +++ b/datasette/default_permissions/__init__.py @@ -37,7 +37,6 @@ from .defaults import ( default_action_permissions_sql as default_action_permissions_sql, DEFAULT_ALLOW_ACTIONS as DEFAULT_ALLOW_ACTIONS, ) -from .tokens import actor_from_signed_api_token as actor_from_signed_api_token @hookimpl diff --git a/datasette/default_permissions/tokens.py b/datasette/default_permissions/tokens.py index 474b0c23..7a359dc6 100644 --- a/datasette/default_permissions/tokens.py +++ b/datasette/default_permissions/tokens.py @@ -1,44 +1,35 @@ """ Token authentication for Datasette. -Handles signed API tokens (dstok_ prefix). +Registers the default SignedTokenHandler and delegates token verification +to datasette.verify_token() so all registered handlers are tried. """ from __future__ import annotations -import time from typing import TYPE_CHECKING, Optional if TYPE_CHECKING: from datasette.app import Datasette -import itsdangerous - from datasette import hookimpl +from datasette.tokens import SignedTokenHandler + + +@hookimpl +def register_token_handler(datasette: "Datasette"): + """Register the default signed token handler.""" + return SignedTokenHandler() @hookimpl(specname="actor_from_request") -def actor_from_signed_api_token(datasette: "Datasette", request) -> Optional[dict]: +async def actor_from_signed_api_token( + datasette: "Datasette", request +) -> Optional[dict]: """ - Authenticate requests using signed API tokens (dstok_ prefix). - - Token structure (signed JSON): - { - "a": "actor_id", # Actor ID - "t": 1234567890, # Timestamp (Unix epoch) - "d": 3600, # Optional: Duration in seconds - "_r": {...} # Optional: Restrictions - } + Authenticate requests using API tokens by delegating to all registered + token handlers via datasette.verify_token(). """ - prefix = "dstok_" - - # Check if tokens are enabled - if not datasette.setting("allow_signed_tokens"): - return None - - max_signed_tokens_ttl = datasette.setting("max_signed_tokens_ttl") - - # Get authorization header authorization = request.headers.get("authorization") if not authorization: return None @@ -46,50 +37,4 @@ def actor_from_signed_api_token(datasette: "Datasette", request) -> Optional[dic return None token = authorization[len("Bearer ") :] - if not token.startswith(prefix): - return None - - # Remove prefix and verify signature - token = token[len(prefix) :] - try: - decoded = datasette.unsign(token, namespace="token") - except itsdangerous.BadSignature: - return None - - # Validate timestamp - if "t" not in decoded: - return None - created = decoded["t"] - if not isinstance(created, int): - return None - - # Handle duration/expiry - duration = decoded.get("d") - if duration is not None and not isinstance(duration, int): - return None - - # Apply max TTL if configured - if (duration is None and max_signed_tokens_ttl) or ( - duration is not None - and max_signed_tokens_ttl - and duration > max_signed_tokens_ttl - ): - duration = max_signed_tokens_ttl - - # Check expiry - if duration: - if time.time() - created > duration: - return None - - # Build actor dict - actor = {"id": decoded["a"], "token": "dstok"} - - # Copy restrictions if present - if "_r" in decoded: - actor["_r"] = decoded["_r"] - - # Add expiry timestamp if applicable - if duration: - actor["token_expires"] = created + duration - - return actor + return await datasette.verify_token(token) diff --git a/datasette/hookspecs.py b/datasette/hookspecs.py index 89be6a65..64901900 100644 --- a/datasette/hookspecs.py +++ b/datasette/hookspecs.py @@ -222,6 +222,11 @@ def top_canned_query(datasette, request, database, query_name): """HTML to include at the top of the canned query page""" +@hookspec +def register_token_handler(datasette): + """Return a TokenHandler instance for token creation and verification""" + + @hookspec def write_wrapper(datasette, database, request, transaction): """Called when a write function is about to execute. diff --git a/datasette/plugins.py b/datasette/plugins.py index e9818885..992137bd 100644 --- a/datasette/plugins.py +++ b/datasette/plugins.py @@ -23,6 +23,7 @@ DEFAULT_PLUGINS = ( "datasette.sql_functions", "datasette.actor_auth_cookie", "datasette.default_permissions", + "datasette.default_permissions.tokens", "datasette.default_actions", "datasette.default_magic_parameters", "datasette.blob_renderer", diff --git a/datasette/tokens.py b/datasette/tokens.py new file mode 100644 index 00000000..5a12d8e0 --- /dev/null +++ b/datasette/tokens.py @@ -0,0 +1,180 @@ +""" +Token handler system for Datasette. + +Provides a base class for token handlers and the default signed token handler. +Plugins can implement register_token_handler to provide custom token backends +(e.g. database-backed tokens that can be revoked and audited). +""" + +from __future__ import annotations + +import dataclasses +import time +from typing import TYPE_CHECKING, Optional + +import itsdangerous + +if TYPE_CHECKING: + from datasette.app import Datasette + + +@dataclasses.dataclass +class TokenRestrictions: + """ + Restrictions to apply to a token, limiting which actions it can perform. + + Use the builder methods to construct restrictions:: + + restrictions = (TokenRestrictions() + .allow_all("view-instance") + .allow_database("mydb", "create-table") + .allow_resource("mydb", "mytable", "insert-row")) + """ + + all: list[str] = dataclasses.field(default_factory=list) + database: dict[str, list[str]] = dataclasses.field(default_factory=dict) + resource: dict[str, dict[str, list[str]]] = dataclasses.field(default_factory=dict) + + def allow_all(self, action: str) -> "TokenRestrictions": + """Allow an action across all databases and resources.""" + self.all.append(action) + return self + + def allow_database(self, database: str, action: str) -> "TokenRestrictions": + """Allow an action on a specific database.""" + self.database.setdefault(database, []).append(action) + return self + + def allow_resource( + self, database: str, resource: str, action: str + ) -> "TokenRestrictions": + """Allow an action on a specific resource within a database.""" + self.resource.setdefault(database, {}).setdefault(resource, []).append(action) + return self + + +class TokenHandler: + """ + Base class for token handlers. + + Subclass this and implement create_token() and verify_token() to provide + a custom token backend. Return an instance from the register_token_handler hook. + """ + + name: str = "" + + async def create_token( + self, + datasette: "Datasette", + actor_id: str, + *, + expires_after: Optional[int] = None, + restrictions: Optional[TokenRestrictions] = None, + ) -> str: + """Create and return a token string for the given actor.""" + raise NotImplementedError + + async def verify_token(self, datasette: "Datasette", token: str) -> Optional[dict]: + """ + Verify a token and return an actor dict, or None if this handler + does not recognize the token. + """ + raise NotImplementedError + + +class SignedTokenHandler(TokenHandler): + """ + Default token handler using itsdangerous signed tokens (dstok_ prefix). + """ + + name = "signed" + + async def create_token( + self, + datasette: "Datasette", + actor_id: str, + *, + expires_after: Optional[int] = None, + restrictions: Optional[TokenRestrictions] = None, + ) -> str: + if not datasette.setting("allow_signed_tokens"): + raise ValueError( + "Signed tokens are not enabled for this Datasette instance" + ) + + token = {"a": actor_id, "t": int(time.time())} + + def abbreviate_action(action): + action_obj = datasette.actions.get(action) + if not action_obj: + return action + return action_obj.abbr or action + + if expires_after: + token["d"] = expires_after + if restrictions and ( + restrictions.all or restrictions.database or restrictions.resource + ): + token["_r"] = {} + if restrictions.all: + token["_r"]["a"] = [abbreviate_action(a) for a in restrictions.all] + if restrictions.database: + token["_r"]["d"] = {} + for database, actions in restrictions.database.items(): + token["_r"]["d"][database] = [abbreviate_action(a) for a in actions] + if restrictions.resource: + token["_r"]["r"] = {} + for database, resources in restrictions.resource.items(): + for resource, actions in resources.items(): + token["_r"]["r"].setdefault(database, {})[resource] = [ + abbreviate_action(a) for a in actions + ] + return "dstok_{}".format(datasette.sign(token, namespace="token")) + + async def verify_token(self, datasette: "Datasette", token: str) -> Optional[dict]: + prefix = "dstok_" + + if not datasette.setting("allow_signed_tokens"): + return None + + max_signed_tokens_ttl = datasette.setting("max_signed_tokens_ttl") + + if not token.startswith(prefix): + return None + + raw = token[len(prefix) :] + try: + decoded = datasette.unsign(raw, namespace="token") + except itsdangerous.BadSignature: + return None + + if "t" not in decoded: + return None + created = decoded["t"] + if not isinstance(created, int): + return None + + duration = decoded.get("d") + if duration is not None and not isinstance(duration, int): + return None + + if (duration is None and max_signed_tokens_ttl) or ( + duration is not None + and max_signed_tokens_ttl + and duration > max_signed_tokens_ttl + ): + duration = max_signed_tokens_ttl + + if duration: + if time.time() - created > duration: + return None + + actor = {"id": decoded["a"], "token": "dstok"} + + if "_r" in decoded: + actor["_r"] = decoded["_r"] + + if duration: + actor["token_expires"] = created + duration + + return actor diff --git a/datasette/views/special.py b/datasette/views/special.py index 640c82eb..dbe5eab1 100644 --- a/datasette/views/special.py +++ b/datasette/views/special.py @@ -710,42 +710,36 @@ class CreateTokenView(BaseView): errors.append("Invalid expire duration unit") # Are there any restrictions? - restrict_all = [] - restrict_database = {} - restrict_resource = {} + from datasette.tokens import TokenRestrictions + + restrictions = TokenRestrictions() for key in form: if key.startswith("all:") and key.count(":") == 1: - restrict_all.append(key.split(":")[1]) + restrictions.allow_all(key.split(":")[1]) elif key.startswith("database:") and key.count(":") == 2: bits = key.split(":") - database = tilde_decode(bits[1]) - action = bits[2] - restrict_database.setdefault(database, []).append(action) + restrictions.allow_database(tilde_decode(bits[1]), bits[2]) elif key.startswith("resource:") and key.count(":") == 3: bits = key.split(":") - database = tilde_decode(bits[1]) - resource = tilde_decode(bits[2]) - action = bits[3] - restrict_resource.setdefault(database, {}).setdefault( - resource, [] - ).append(action) + restrictions.allow_resource( + tilde_decode(bits[1]), tilde_decode(bits[2]), bits[3] + ) - token = self.ds.create_token( + token = await self.ds.create_token( request.actor["id"], expires_after=expires_after, - restrict_all=restrict_all, - restrict_database=restrict_database, - restrict_resource=restrict_resource, + restrictions=restrictions, + handler="signed", ) token_bits = self.ds.unsign(token[len("dstok_") :], namespace="token") await self.ds.track_event( CreateTokenEvent( actor=request.actor, expires_after=expires_after, - restrict_all=restrict_all, - restrict_database=restrict_database, - restrict_resource=restrict_resource, + restrict_all=restrictions.all, + restrict_database=restrictions.database, + restrict_resource=restrictions.resource, ) ) context = await self.shared(request) diff --git a/docs/authentication.rst b/docs/authentication.rst index 69a6f606..1b949f9a 100644 --- a/docs/authentication.rst +++ b/docs/authentication.rst @@ -1072,6 +1072,7 @@ cannot grant new access. If the underlying actor is denied by ``allow`` rules in ``datasette.yaml`` or by a plugin, a token that lists that resource in its ``"_r"`` section will still be denied. +To create tokens with restrictions in Python code, use the :ref:`TokenRestrictions ` builder and pass it to :ref:`datasette.create_token() `. .. _permissions_plugins: diff --git a/docs/internals.rst b/docs/internals.rst index 0491c1f7..7d607bfe 100644 --- a/docs/internals.rst +++ b/docs/internals.rst @@ -673,8 +673,8 @@ This example checks if the user can access a specific table, and sets ``private` .. _datasette_create_token: -.create_token(actor_id, expires_after=None, restrict_all=None, restrict_database=None, restrict_resource=None) --------------------------------------------------------------------------------------------------------------- +await .create_token(actor_id, expires_after=None, restrictions=None, handler=None) +---------------------------------------------------------------------------------- ``actor_id`` - string The ID of the actor to create a token for. @@ -682,16 +682,13 @@ This example checks if the user can access a specific table, and sets ``private` ``expires_after`` - int, optional The number of seconds after which the token should expire. -``restrict_all`` - iterable, optional - A list of actions that this token should be restricted to across all databases and resources. +``restrictions`` - :ref:`TokenRestrictions `, optional + A :ref:`TokenRestrictions ` object limiting which actions the token can perform. -``restrict_database`` - dict, optional - For restricting actions within specific databases, e.g. ``{"mydb": ["view-table", "view-query"]}``. +``handler`` - string, optional + The name of a specific token handler to use. If omitted, the first registered handler is used. See :ref:`plugin_hook_register_token_handler`. -``restrict_resource`` - dict, optional - For restricting actions to specific resources (tables, SQL views and :ref:`canned_queries`) within a database. For example: ``{"mydb": {"mytable": ["insert-row", "update-row"]}}``. - -This method returns a signed :ref:`API token ` of the format ``dstok_...`` which can be used to authenticate requests to the Datasette API. +This is an ``async`` method that returns an :ref:`API token ` string which can be used to authenticate requests to the Datasette API. The default ``SignedTokenHandler`` returns tokens of the format ``dstok_...``. All tokens must have an ``actor_id`` string indicating the ID of the actor which the token will act on behalf of. @@ -699,28 +696,72 @@ Tokens default to lasting forever, but can be set to expire after a given number .. code-block:: python - token = datasette.create_token( + token = await datasette.create_token( actor_id="user1", expires_after=3600, ) -The three ``restrict_*`` arguments can be used to create a token that has additional restrictions beyond what the associated actor is allowed to do. +.. _TokenRestrictions: + +TokenRestrictions +~~~~~~~~~~~~~~~~~ + +The ``TokenRestrictions`` class uses a builder pattern to specify which actions a token is allowed to perform. Import it from ``datasette.tokens``: + +.. code-block:: python + + from datasette.tokens import TokenRestrictions + + restrictions = ( + TokenRestrictions() + .allow_all("view-instance") + .allow_all("view-table") + .allow_database("docs", "view-query") + .allow_resource("docs", "attachments", "insert-row") + .allow_resource("docs", "attachments", "update-row") + ) + +The builder methods are: + +- ``allow_all(action)`` - allow an action across all databases and resources +- ``allow_database(database, action)`` - allow an action on a specific database +- ``allow_resource(database, resource, action)`` - allow an action on a specific resource (table, SQL view or :ref:`canned query `) within a database + +Each method returns the ``TokenRestrictions`` instance so calls can be chained. The following example creates a token that can access ``view-instance`` and ``view-table`` across everything, can additionally use ``view-query`` for anything in the ``docs`` database and is allowed to execute ``insert-row`` and ``update-row`` in the ``attachments`` table in that database: .. code-block:: python - token = datasette.create_token( + token = await datasette.create_token( actor_id="user1", - restrict_all=("view-instance", "view-table"), - restrict_database={"docs": ("view-query",)}, - restrict_resource={ - "docs": { - "attachments": ("insert-row", "update-row") - } - }, + restrictions=( + TokenRestrictions() + .allow_all("view-instance") + .allow_all("view-table") + .allow_database("docs", "view-query") + .allow_resource("docs", "attachments", "insert-row") + .allow_resource("docs", "attachments", "update-row") + ), ) +.. _datasette_verify_token: + +await .verify_token(token) +-------------------------- + +``token`` - string + The token string to verify. + +This is an ``async`` method that verifies an API token by trying each registered token handler in order. Returns an actor dictionary from the first handler that recognizes the token, or ``None`` if no handler accepts it. + +.. code-block:: python + + actor = await datasette.verify_token(token) + if actor: + # Token was valid + print(actor["id"]) + .. _datasette_get_database: .get_database(name) diff --git a/docs/plugin_hooks.rst b/docs/plugin_hooks.rst index fa335368..b9701f7c 100644 --- a/docs/plugin_hooks.rst +++ b/docs/plugin_hooks.rst @@ -2334,3 +2334,62 @@ The plugin can then call ``datasette.track_event(...)`` to send a ``ban-user`` e BanUserEvent(user={"id": 1, "username": "cleverbot"}) ) +.. _plugin_hook_register_token_handler: + +register_token_handler(datasette) +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +``datasette`` - :ref:`internals_datasette` + You can use this to access plugin configuration options via ``datasette.plugin_config(your_plugin_name)``. + +Return a ``TokenHandler`` instance to provide a custom token creation and verification backend. This hook can return a single ``TokenHandler`` or a list of them. + +The default ``SignedTokenHandler`` uses itsdangerous signed tokens (``dstok_`` prefix). Plugins can provide alternative backends such as database-backed tokens that support revocation and auditing. + +.. code-block:: python + + from datasette import hookimpl, TokenHandler + + + class DatabaseTokenHandler(TokenHandler): + name = "database" + + async def create_token( + self, + datasette, + actor_id, + *, + expires_after=None, + restrictions=None + ): + # Store token in database and return token string + ... + + async def verify_token(self, datasette, token): + # Look up token in database, return actor dict or None + ... + + + @hookimpl + def register_token_handler(datasette): + return DatabaseTokenHandler() + +The ``create_token`` method receives a ``restrictions`` argument which will be a :ref:`TokenRestrictions ` instance or ``None``. + +Tokens can then be created and verified using :ref:`datasette.create_token() ` and ``datasette.verify_token()``, which delegate to the registered handlers. If no ``handler`` is specified, the first handler is used according to `pluggy call-time ordering `_. Use the ``handler`` parameter to select a specific backend by name: + +.. code-block:: python + + # Uses first registered handler (default) + token = await datasette.create_token("user123") + + # Uses a specific handler by name + token = await datasette.create_token( + "user123", handler="database" + ) + + # Verification tries all handlers + actor = await datasette.verify_token(token) + +If no handlers are registered, ``create_token()`` raises ``RuntimeError``. If the requested ``handler`` name is not found, it raises ``ValueError``. + diff --git a/docs/plugins.rst b/docs/plugins.rst index d5a98923..60bdc111 100644 --- a/docs/plugins.rst +++ b/docs/plugins.rst @@ -231,12 +231,21 @@ If you run ``datasette plugins --all`` it will include default plugins that ship "templates": false, "version": null, "hooks": [ - "actor_from_request", "canned_queries", "permission_resources_sql", "skip_csrf" ] }, + { + "name": "datasette.default_permissions.tokens", + "static": false, + "templates": false, + "version": null, + "hooks": [ + "actor_from_request", + "register_token_handler" + ] + }, { "name": "datasette.events", "static": false, diff --git a/docs/upgrade_guide.md b/docs/upgrade_guide.md index a3c321a4..861a8795 100644 --- a/docs/upgrade_guide.md +++ b/docs/upgrade_guide.md @@ -114,3 +114,43 @@ Instead, one should use the following methods on a Datasette class: ```{include} upgrade-1.0a20.md :heading-offset: 1 ``` + +(upgrade_guide_v1_a25)= +### Datasette 1.0a25: `create_token()` signature change + +`datasette.create_token()` is now an `async` method (previously it was synchronous). The `restrict_all`, `restrict_database`, and `restrict_resource` keyword arguments have been replaced by a single `restrictions` parameter that accepts a {ref}`TokenRestrictions ` object. + +Old code: + +```python +token = datasette.create_token( + actor_id="user1", + restrict_all=["view-instance", "view-table"], + restrict_database={"docs": ["view-query"]}, + restrict_resource={ + "docs": { + "attachments": ["insert-row", "update-row"] + } + }, +) +``` + +New code: + +```python +from datasette.tokens import TokenRestrictions + +token = await datasette.create_token( + actor_id="user1", + restrictions=( + TokenRestrictions() + .allow_all("view-instance") + .allow_all("view-table") + .allow_database("docs", "view-query") + .allow_resource("docs", "attachments", "insert-row") + .allow_resource("docs", "attachments", "update-row") + ), +) +``` + +The `datasette create-token` CLI command is unchanged. diff --git a/tests/fixtures.py b/tests/fixtures.py index 9f99519a..1f6c491d 100644 --- a/tests/fixtures.py +++ b/tests/fixtures.py @@ -51,6 +51,7 @@ EXPECTED_PLUGINS = [ "register_facet_classes", "register_magic_parameters", "register_routes", + "register_token_handler", "render_cell", "row_actions", "skip_csrf", diff --git a/tests/plugins/my_plugin.py b/tests/plugins/my_plugin.py index 20e7d111..77079557 100644 --- a/tests/plugins/my_plugin.py +++ b/tests/plugins/my_plugin.py @@ -1,6 +1,7 @@ import asyncio from datasette import hookimpl from datasette.facets import Facet +from datasette.tokens import TokenHandler from datasette import tracer from datasette.permissions import Action from datasette.resources import DatabaseResource @@ -586,3 +587,29 @@ def permission_resources_sql(datasette, actor, action): return PermissionSQL.allow(reason=f"todomvc actor allowed for {action}") return None + + +class HardcodedTokenHandler(TokenHandler): + name = "hardcoded" + _counter = 0 + + async def create_token( + self, + datasette, + actor_id, + *, + expires_after=None, + restrictions=None, + ): + HardcodedTokenHandler._counter += 1 + return f"dstok_hardcoded_token_{HardcodedTokenHandler._counter}" + + async def verify_token(self, datasette, token): + if token.startswith("dstok_hardcoded_token_"): + return {"id": "hardcoded-actor", "token": "hardcoded"} + return None + + +@hookimpl +def register_token_handler(datasette): + return HardcodedTokenHandler() diff --git a/tests/test_api_write.py b/tests/test_api_write.py index 05835e51..e59c4295 100644 --- a/tests/test_api_write.py +++ b/tests/test_api_write.py @@ -1362,7 +1362,14 @@ async def test_create_table( async def test_create_table_permissions( ds_write, permissions, body, expected_status, expected_errors ): - token = ds_write.create_token("root", restrict_all=["view-instance"] + permissions) + from datasette.tokens import TokenRestrictions + + restrictions = TokenRestrictions() + for action in ["view-instance"] + permissions: + restrictions.allow_all(action) + token = await ds_write.create_token( + "root", handler="signed", restrictions=restrictions + ) response = await ds_write.client.post( "/data/-/create", json=body, diff --git a/tests/test_permissions.py b/tests/test_permissions.py index 96c0cf6f..42a19ca4 100644 --- a/tests/test_permissions.py +++ b/tests/test_permissions.py @@ -1657,7 +1657,7 @@ async def test_permission_check_view_requires_debug_permission(): # Root user should have access (root has all permissions) ds_with_root = Datasette() ds_with_root.root_enabled = True - root_token = ds_with_root.create_token("root") + root_token = await ds_with_root.create_token("root", handler="signed") response = await ds_with_root.client.get( "/-/check.json?action=view-instance", headers={"Authorization": f"Bearer {root_token}"}, diff --git a/tests/test_plugins.py b/tests/test_plugins.py index 754b199c..fa9d1a1f 100644 --- a/tests/test_plugins.py +++ b/tests/test_plugins.py @@ -1566,6 +1566,38 @@ async def test_hook_register_events(): assert any(k.__name__ == "OneEvent" for k in datasette.event_classes) +@pytest.mark.asyncio +async def test_hook_register_token_handler(ds_client): + handlers = ds_client.ds._token_handlers() + handler_names = [h.name for h in handlers] + # Both the default signed handler and the test hardcoded handler + assert "signed" in handler_names + assert "hardcoded" in handler_names + + # Create a token using the hardcoded handler (first registered from plugins dir) + token = await ds_client.ds.create_token("test-user") + assert token.startswith("dstok_hardcoded_token_") + + # Verify it + actor = await ds_client.ds.verify_token(token) + assert actor["id"] == "hardcoded-actor" + assert actor["token"] == "hardcoded" + + # Create a token by explicitly requesting the hardcoded handler by name + token2 = await ds_client.ds.create_token("test-user", handler="hardcoded") + assert token2.startswith("dstok_hardcoded_token_") + actor2 = await ds_client.ds.verify_token(token2) + assert actor2["id"] == "hardcoded-actor" + + # Create a token by explicitly requesting the signed handler by name + signed_token = await ds_client.ds.create_token("test-user", handler="signed") + assert signed_token.startswith("dstok_") + assert not signed_token.startswith("dstok_hardcoded_token_") + signed_actor = await ds_client.ds.verify_token(signed_token) + assert signed_actor["id"] == "test-user" + assert signed_actor["token"] == "dstok" + + @pytest.mark.asyncio async def test_hook_write_wrapper(): datasette = Datasette(memory=True) diff --git a/tests/test_token_handler.py b/tests/test_token_handler.py new file mode 100644 index 00000000..83f09046 --- /dev/null +++ b/tests/test_token_handler.py @@ -0,0 +1,301 @@ +""" +Tests for the register_token_handler plugin hook. +""" + +from datasette.app import Datasette +from datasette.hookspecs import hookimpl +from datasette.plugins import pm +from datasette.tokens import TokenHandler, TokenRestrictions, SignedTokenHandler +import pytest + + +@pytest.fixture +def datasette(): + return Datasette() + + +@pytest.mark.asyncio +async def test_default_signed_handler_registered(datasette): + """The default SignedTokenHandler should be registered automatically.""" + handlers = datasette._token_handlers() + assert len(handlers) >= 1 + assert any(isinstance(h, SignedTokenHandler) for h in handlers) + assert any(h.name == "signed" for h in handlers) + + +@pytest.mark.asyncio +async def test_create_token_default(datasette): + """create_token() with handler='signed' should create a signed token.""" + token = await datasette.create_token("test_actor", handler="signed") + assert token.startswith("dstok_") + + +@pytest.mark.asyncio +async def test_create_token_with_restrictions(datasette): + """create_token() should handle restriction parameters.""" + token = await datasette.create_token( + "test_actor", + handler="signed", + expires_after=3600, + restrictions=TokenRestrictions().allow_all("view-instance"), + ) + assert token.startswith("dstok_") + # Verify the token contains the expected data + decoded = datasette.unsign(token[len("dstok_") :], namespace="token") + assert decoded["a"] == "test_actor" + assert decoded["d"] == 3600 + assert "_r" in decoded + assert "a" in decoded["_r"] + + +@pytest.mark.asyncio +async def test_verify_token_default(datasette): + """verify_token() should verify signed tokens.""" + token = await datasette.create_token("test_actor", handler="signed") + actor = await datasette.verify_token(token) + assert actor is not None + assert actor["id"] == "test_actor" + assert actor["token"] == "dstok" + + +@pytest.mark.asyncio +async def test_verify_token_unknown_returns_none(datasette): + """verify_token() should return None for unrecognized tokens.""" + result = await datasette.verify_token("unknown_token_format_xyz") + assert result is None + + +@pytest.mark.asyncio +async def test_verify_token_bad_signature_returns_none(datasette): + """verify_token() should return None for tokens with bad signatures.""" + result = await datasette.verify_token("dstok_tampered_data_here") + assert result is None + + +@pytest.mark.asyncio +async def test_create_token_with_named_handler(datasette): + """create_token(handler='signed') should select the signed handler.""" + token = await datasette.create_token("test_actor", handler="signed") + assert token.startswith("dstok_") + + +@pytest.mark.asyncio +async def test_create_token_unknown_handler_raises(datasette): + """create_token(handler='nonexistent') should raise ValueError.""" + with pytest.raises(ValueError, match="Token handler 'nonexistent' not found"): + await datasette.create_token("test_actor", handler="nonexistent") + + +@pytest.mark.asyncio +async def test_custom_token_handler(datasette): + """A custom token handler should be usable for both create and verify.""" + + class CustomHandler(TokenHandler): + name = "custom" + + async def create_token(self, datasette, actor_id, **kwargs): + return f"custom_{actor_id}" + + async def verify_token(self, datasette, token): + if token.startswith("custom_"): + return {"id": token[len("custom_") :], "token": "custom"} + return None + + class Plugin: + __name__ = "CustomTokenPlugin" + + @staticmethod + @hookimpl + def register_token_handler(datasette): + return CustomHandler() + + pm.register(Plugin(), name="test_custom_handler") + try: + handlers = datasette._token_handlers() + assert any(h.name == "custom" for h in handlers) + + # Create with custom handler + token = await datasette.create_token("alice", handler="custom") + assert token == "custom_alice" + + # Verify custom token + actor = await datasette.verify_token("custom_alice") + assert actor is not None + assert actor["id"] == "alice" + assert actor["token"] == "custom" + + # Signed tokens should still work + signed_token = await datasette.create_token("bob", handler="signed") + assert signed_token.startswith("dstok_") + actor = await datasette.verify_token(signed_token) + assert actor["id"] == "bob" + finally: + pm.unregister(name="test_custom_handler") + + +@pytest.mark.asyncio +async def test_verify_token_tries_all_handlers(datasette): + """verify_token() should try each handler until one matches.""" + + class HandlerA(TokenHandler): + name = "handler_a" + + async def create_token(self, datasette, actor_id, **kwargs): + return f"a_{actor_id}" + + async def verify_token(self, datasette, token): + if token.startswith("a_"): + return {"id": token[2:], "token": "handler_a"} + return None + + class HandlerB(TokenHandler): + name = "handler_b" + + async def create_token(self, datasette, actor_id, **kwargs): + return f"b_{actor_id}" + + async def verify_token(self, datasette, token): + if token.startswith("b_"): + return {"id": token[2:], "token": "handler_b"} + return None + + class PluginA: + __name__ = "PluginA" + + @staticmethod + @hookimpl + def register_token_handler(datasette): + return HandlerA() + + class PluginB: + __name__ = "PluginB" + + @staticmethod + @hookimpl + def register_token_handler(datasette): + return HandlerB() + + pm.register(PluginA(), name="test_handler_a") + pm.register(PluginB(), name="test_handler_b") + try: + # Both handler tokens should verify + actor_a = await datasette.verify_token("a_alice") + assert actor_a is not None + assert actor_a["id"] == "alice" + assert actor_a["token"] == "handler_a" + + actor_b = await datasette.verify_token("b_bob") + assert actor_b is not None + assert actor_b["id"] == "bob" + assert actor_b["token"] == "handler_b" + + # Unknown token should return None + assert await datasette.verify_token("c_charlie") is None + finally: + pm.unregister(name="test_handler_a") + pm.unregister(name="test_handler_b") + + +@pytest.mark.asyncio +async def test_token_handler_via_http(datasette): + """Default signed tokens should work through HTTP auth.""" + token = await datasette.create_token("http_user", handler="signed") + response = await datasette.client.get( + "/-/actor.json", + headers={"Authorization": f"Bearer {token}"}, + ) + assert response.status_code == 200 + actor = response.json()["actor"] + assert actor["id"] == "http_user" + assert actor["token"] == "dstok" + + +@pytest.mark.asyncio +async def test_custom_handler_via_http(datasette): + """Custom handler tokens should work through HTTP auth.""" + + class CustomHandler(TokenHandler): + name = "custom_http" + + async def create_token(self, datasette, actor_id, **kwargs): + return f"chttp_{actor_id}" + + async def verify_token(self, datasette, token): + if token.startswith("chttp_"): + return {"id": token[len("chttp_") :], "token": "custom_http"} + return None + + class Plugin: + __name__ = "CustomHTTPPlugin" + + @staticmethod + @hookimpl + def register_token_handler(datasette): + return CustomHandler() + + pm.register(Plugin(), name="test_custom_http") + try: + token = await datasette.create_token("web_user", handler="custom_http") + response = await datasette.client.get( + "/-/actor.json", + headers={"Authorization": f"Bearer {token}"}, + ) + assert response.status_code == 200 + actor = response.json()["actor"] + assert actor["id"] == "web_user" + assert actor["token"] == "custom_http" + finally: + pm.unregister(name="test_custom_http") + + +@pytest.mark.asyncio +async def test_token_handler_base_class_raises(): + """TokenHandler base class methods should raise NotImplementedError.""" + handler = TokenHandler() + ds = Datasette() + with pytest.raises(NotImplementedError): + await handler.create_token(ds, "test") + with pytest.raises(NotImplementedError): + await handler.verify_token(ds, "test") + + +@pytest.mark.asyncio +async def test_restrictions_round_trip(datasette): + """Tokens with database/resource restrictions should round-trip correctly.""" + restrictions = ( + TokenRestrictions() + .allow_all("view-instance") + .allow_database("docs", "view-query") + .allow_resource("docs", "attachments", "insert-row") + ) + token = await datasette.create_token( + "test_actor", handler="signed", restrictions=restrictions + ) + actor = await datasette.verify_token(token) + assert actor is not None + assert actor["id"] == "test_actor" + assert actor["_r"]["a"] == ["view-instance"] + assert actor["_r"]["d"] == {"docs": ["view-query"]} + assert actor["_r"]["r"] == {"docs": {"attachments": ["insert-row"]}} + + +@pytest.mark.asyncio +async def test_expires_after_round_trip(datasette): + """Tokens with expires_after should include token_expires in the actor.""" + token = await datasette.create_token( + "test_actor", handler="signed", expires_after=3600 + ) + actor = await datasette.verify_token(token) + assert actor is not None + assert actor["id"] == "test_actor" + assert "token_expires" in actor + + +@pytest.mark.asyncio +async def test_signed_tokens_disabled(): + """create_token and verify_token should fail/skip when signed tokens are disabled.""" + ds = Datasette(settings={"allow_signed_tokens": False}) + with pytest.raises(ValueError, match="Signed tokens are not enabled"): + await ds.create_token("test_actor", handler="signed") + # verify_token should return None rather than raising + assert await ds.verify_token("dstok_anything") is None From 24d801b7f799912cb4eb897a97e4f4a9fe76b966 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Wed, 25 Feb 2026 16:33:27 -0800 Subject: [PATCH 006/472] Respect metadata-defined facet ordering in sorted_facet_results (#2648) * Preserve metadata-defined facet ordering on table pages When facets are explicitly defined in table metadata/config, they now appear in the order specified in the configuration rather than being sorted by result count. Request-added facets still appear after metadata-defined facets, sorted by count as before. * Document metadata-defined facet ordering behavior * Apply black formatting https://claude.ai/code/session_01PbSHtjsUpNk3Fx7xjvVqDb --- datasette/views/table.py | 34 ++++++++++++++++++++++++++----- docs/facets.rst | 2 ++ tests/test_facets.py | 44 ++++++++++++++++++++++++++++++++++++---- 3 files changed, 71 insertions(+), 9 deletions(-) diff --git a/datasette/views/table.py b/datasette/views/table.py index 594e925e..e1e5507f 100644 --- a/datasette/views/table.py +++ b/datasette/views/table.py @@ -1580,11 +1580,35 @@ async def table_view_data( ] async def extra_sorted_facet_results(extra_facet_results): - return sorted( - extra_facet_results["results"].values(), - key=lambda f: (len(f["results"]), f["name"]), - reverse=True, - ) + facet_configs = table_metadata.get("facets", []) + if facet_configs: + # Build ordered list of facet names from metadata config + metadata_facet_names = [] + for fc in facet_configs: + if isinstance(fc, str): + metadata_facet_names.append(fc) + elif isinstance(fc, dict): + metadata_facet_names.append(list(fc.values())[0]) + metadata_order = {name: i for i, name in enumerate(metadata_facet_names)} + metadata_facets = [] + request_facets = [] + for f in extra_facet_results["results"].values(): + if f["name"] in metadata_order: + metadata_facets.append(f) + else: + request_facets.append(f) + metadata_facets.sort(key=lambda f: metadata_order[f["name"]]) + request_facets.sort( + key=lambda f: (len(f["results"]), f["name"]), + reverse=True, + ) + return metadata_facets + request_facets + else: + return sorted( + extra_facet_results["results"].values(), + key=lambda f: (len(f["results"]), f["name"]), + reverse=True, + ) async def extra_table_definition(): return await db.get_table_definition(table_name) diff --git a/docs/facets.rst b/docs/facets.rst index 15fe7227..2a135b69 100644 --- a/docs/facets.rst +++ b/docs/facets.rst @@ -153,6 +153,8 @@ Here's an example that turns on faceting by default for the ``qLegalStatus`` col Facets defined in this way will always be shown in the interface and returned in the API, regardless of the ``_facet`` arguments passed to the view. +Facets defined in metadata will be displayed in the order they are listed in the configuration. Any additional facets added via query string parameters (e.g. ``?_facet=column_name``) will appear after the metadata-defined facets, sorted by the number of unique values. + You can specify :ref:`array ` or :ref:`date ` facets in metadata using JSON objects with a single key of ``array`` or ``date`` and a value specifying the column, like this: .. [[[cog diff --git a/tests/test_facets.py b/tests/test_facets.py index a2b505ec..8c22ffce 100644 --- a/tests/test_facets.py +++ b/tests/test_facets.py @@ -623,12 +623,48 @@ def test_other_types_of_facet_in_metadata(): } ) as client: response = client.get("/fixtures/facetable") - for fragment in ( - "created (date)\n", - "tags (array)\n", + fragments = ( "state\n", - ): + "tags (array)\n", + "created (date)\n", + ) + for fragment in fragments: assert fragment in response.text + # Verify they appear in the metadata-defined order + positions = [response.text.index(f) for f in fragments] + assert positions == sorted( + positions + ), "Facets should appear in metadata-defined order" + + +def test_metadata_facet_ordering(): + with make_app_client( + metadata={ + "databases": { + "fixtures": { + "tables": { + "facetable": { + "facets": ["state", {"array": "tags"}, {"date": "created"}] + } + } + } + } + } + ) as client: + # JSON response should have facets in the metadata-defined order + response = client.get("/fixtures/facetable.json?_extra=sorted_facet_results") + data = response.json + facet_names = [f["name"] for f in data["sorted_facet_results"]] + assert facet_names == ["state", "tags", "created"] + + # With an additional request-based facet, metadata facets come first + # in their defined order, followed by request-based facets + response2 = client.get( + "/fixtures/facetable.json?_extra=sorted_facet_results&_facet=_city_id" + ) + data2 = response2.json + facet_names2 = [f["name"] for f in data2["sorted_facet_results"]] + assert facet_names2 == ["state", "tags", "created", "_city_id"] @pytest.mark.asyncio From 2bc1dd2275978e75622c5764729a4273ebac957e Mon Sep 17 00:00:00 2001 From: Daniel Bates Date: Wed, 25 Feb 2026 16:46:29 -0800 Subject: [PATCH 007/472] Fix --reload interpreting 'serve' command as a file argument (#2646) When hupper spawns the worker process, it calls the function specified by worker_path directly. Using "datasette.cli.serve" causes Click to parse sys.argv without going through the CLI group, so the literal word "serve" from the original command gets treated as a positional file argument. Change the worker path to "datasette.cli.cli" so the worker process goes through the Click group dispatcher, which properly recognizes "serve" as a subcommand and strips it from the argument list. Closes #2123 Co-authored-by: Claude Opus 4.6 Co-authored-by: Simon Willison --- datasette/cli.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/datasette/cli.py b/datasette/cli.py index b473fbb7..db777fe8 100644 --- a/datasette/cli.py +++ b/datasette/cli.py @@ -547,7 +547,7 @@ def serve( if reload: import hupper - reloader = hupper.start_reloader("datasette.cli.serve") + reloader = hupper.start_reloader("datasette.cli.cli") if immutable: reloader.watch_files(immutable) if config: From 1246c6576bb2f1ba9dc5c7d9811427d00d440976 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Wed, 25 Feb 2026 16:49:14 -0800 Subject: [PATCH 008/472] Release 1.0a25 Refs #2636, #2641, #2646, #2647, #2650 --- docs/changelog.rst | 41 +++++++++++++++++++++++++++++++++++++++++ docs/contributing.rst | 1 + docs/upgrade-1.0a20.md | 1 - docs/upgrade_guide.md | 1 + 4 files changed, 43 insertions(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 67ceeece..c0467793 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,47 @@ Changelog ========= +.. _v1_0_a25: + +1.0a25 (2026-02-25) +------------------- + +``write_wrapper`` plugin hook for intercepting write operations +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +A new :ref:`write_wrapper() ` plugin hook allows plugins to intercept and wrap database write operations. (`#2636 `__) + +Plugins implement the hook as a generator-based context manager: + +.. code-block:: python + + @hookimpl + def write_wrapper(datasette, database, request): + def wrapper(conn): + # Setup code runs before the write + yield + # Cleanup code runs after the write + + return wrapper + +``register_token_handler()`` plugin hook for custom API token backends +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +A new :ref:`register_token_handler() ` plugin hook allows plugins to provide custom token backends for API authentication. (`#2650 `__) + +This includes a **backwards incompatible change**: the ``datasette.create_token()`` internal method is now an ``async`` method. Consult the :ref:`upgrade guide ` for details on how to update your code. + +``render_cell()`` now receives a ``pks`` parameter +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +The :ref:`render_cell() ` plugin hook now receives a ``pks`` parameter containing the list of primary key column names for the table being rendered. This avoids plugins needing to make redundant async calls to look up primary keys. (`#2641 `__) + +Other changes +~~~~~~~~~~~~~ + +- Facets defined in metadata now preserve their configured order, instead of being sorted by result count. Request-based facets added via the ``_facet`` parameter are still sorted by result count and appear after metadata-defined facets. (:issue:`2647`) +- Fixed ``--reload`` incorrectly interpreting the ``serve`` command as a file argument. Thanks, `Daniel Bates `__. (`#2646 `__) + .. _v1_0_a24: 1.0a24 (2026-01-29) diff --git a/docs/contributing.rst b/docs/contributing.rst index 3d41a125..635ca60e 100644 --- a/docs/contributing.rst +++ b/docs/contributing.rst @@ -90,6 +90,7 @@ If you want to change Datasette's Python code you can use the ``--reload`` optio You can also use the ``fixtures.py`` script to recreate the testing version of ``metadata.json`` used by the unit tests. To do that:: uv run python tests/fixtures.py fixtures.db fixtures-metadata.json + Or to output the plugins used by the tests, run this:: uv run python tests/fixtures.py fixtures.db fixtures-metadata.json fixtures-plugins diff --git a/docs/upgrade-1.0a20.md b/docs/upgrade-1.0a20.md index 749d383c..fbc3f4a8 100644 --- a/docs/upgrade-1.0a20.md +++ b/docs/upgrade-1.0a20.md @@ -2,7 +2,6 @@ orphan: true --- -(upgrade_guide_v1_a20)= # Datasette 1.0a20 plugin upgrade guide Datasette 1.0a20 makes some breaking changes to Datasette's permission system. Plugins need to be updated if they use **any of the following**: diff --git a/docs/upgrade_guide.md b/docs/upgrade_guide.md index 861a8795..b67eb054 100644 --- a/docs/upgrade_guide.md +++ b/docs/upgrade_guide.md @@ -111,6 +111,7 @@ Instead, one should use the following methods on a Datasette class: - {ref}`get_resource_metadata() ` - {ref}`get_column_metadata() ` +(upgrade_guide_v1_a20)= ```{include} upgrade-1.0a20.md :heading-offset: 1 ``` From e4ff5e27d356ca5b3c807e821acedf8c71c37e47 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Wed, 25 Feb 2026 16:54:51 -0800 Subject: [PATCH 009/472] Fix RST heading underlin --- docs/changelog.rst | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index c0467793..1e6a8e90 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -35,7 +35,7 @@ A new :ref:`register_token_handler() ` plugi This includes a **backwards incompatible change**: the ``datasette.create_token()`` internal method is now an ``async`` method. Consult the :ref:`upgrade guide ` for details on how to update your code. ``render_cell()`` now receives a ``pks`` parameter -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The :ref:`render_cell() ` plugin hook now receives a ``pks`` parameter containing the list of primary key column names for the table being rendered. This avoids plugins needing to make redundant async calls to look up primary keys. (`#2641 `__) From 8f0d60236f844a6d12bd1439f57b1b3d65fcad36 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Wed, 25 Feb 2026 17:01:03 -0800 Subject: [PATCH 010/472] Bump version for 1.0a25 --- datasette/version.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/datasette/version.py b/datasette/version.py index de7585ca..2907e537 100644 --- a/datasette/version.py +++ b/datasette/version.py @@ -1,2 +1,2 @@ -__version__ = "1.0a24" +__version__ = "1.0a25" __version_info__ = tuple(__version__.split(".")) From 1263380ea6b138ac63683edfd525323c6fe8eef9 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Wed, 25 Feb 2026 20:50:46 -0800 Subject: [PATCH 011/472] Better heading for write_wrapper() --- docs/changelog.rst | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/changelog.rst b/docs/changelog.rst index 1e6a8e90..2c9b7170 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -9,8 +9,8 @@ Changelog 1.0a25 (2026-02-25) ------------------- -``write_wrapper`` plugin hook for intercepting write operations -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +``write_wrapper()`` plugin hook for intercepting write operations +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ A new :ref:`write_wrapper() ` plugin hook allows plugins to intercept and wrap database write operations. (`#2636 `__) From 97201f067c4f64b00ccf7e02f787d65c767f9bc9 Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Fri, 6 Mar 2026 20:16:50 -0800 Subject: [PATCH 012/472] Row pages link to foreign keys from table display, closes #1592 https://gisthost.github.io/?40813f5b3e4d83c0efe1c09135f84290/index.html Also now shows primary key column first and in bold on that page. --- datasette/views/row.py | 64 ++++++++++++++++++++++++++++++++++++++++-- tests/test_html.py | 32 +++++++++++++++++---- 2 files changed, 88 insertions(+), 8 deletions(-) diff --git a/datasette/views/row.py b/datasette/views/row.py index 9c59cd3b..7cc46368 100644 --- a/datasette/views/row.py +++ b/datasette/views/row.py @@ -5,12 +5,14 @@ from datasette.resources import TableResource from .base import DataView, BaseView, _error from datasette.utils import ( await_me_maybe, + CustomRow, make_slot_function, to_css_class, escape_sqlite, ) from datasette.plugins import pm import json +import markupsafe import sqlite_utils from .table import display_columns_and_rows, _get_extras @@ -42,13 +44,62 @@ class RowView(DataView): if not rows: raise NotFound(f"Record not found: {pk_values}") + pks = resolved.pks + async def template_data(): + # Reorder columns so primary keys come first + pk_set = set(pks) + pk_cols = [d for d in results.description if d[0] in pk_set] + non_pk_cols = [d for d in results.description if d[0] not in pk_set] + reordered_description = pk_cols + non_pk_cols + reordered_columns = [d[0] for d in reordered_description] + + # Reorder row data to match + reordered_rows = [] + for row in rows: + new_row = CustomRow(reordered_columns) + for col in reordered_columns: + new_row[col] = row[col] + reordered_rows.append(new_row) + + # Expand foreign key columns into dicts so display_columns_and_rows + # renders them as hyperlinks, matching the table view behavior + expanded_rows = reordered_rows + for fk in await db.foreign_keys_for_table(table): + column = fk["column"] + if column not in reordered_columns: + continue + column_index = reordered_columns.index(column) + values = [row[column_index] for row in expanded_rows] + expanded_labels = await self.ds.expand_foreign_keys( + request.actor, database, table, column, values + ) + if expanded_labels: + new_rows = [] + for row in expanded_rows: + new_row = CustomRow(reordered_columns) + for col in reordered_columns: + value = row[col] + if ( + col == column + and (col, value) in expanded_labels + and value is not None + ): + new_row[col] = { + "value": value, + "label": expanded_labels[(col, value)], + } + else: + new_row[col] = value + new_rows.append(new_row) + expanded_rows = new_rows + display_columns, display_rows = await display_columns_and_rows( self.ds, database, table, - results.description, - rows, + reordered_description, + expanded_rows, link_column=False, truncate_cells=0, request=request, @@ -56,6 +107,14 @@ class RowView(DataView): for column in display_columns: column["sortable"] = False + # Bold primary key cell values + for row in display_rows: + for cell in row: + if cell["column"] in pk_set: + cell["value"] = markupsafe.Markup( + "{}".format(cell["value"]) + ) + row_actions = [] for hook in pm.hook.row_actions( datasette=self.ds, @@ -71,6 +130,7 @@ class RowView(DataView): return { "private": private, + "columns": reordered_columns, "foreign_key_tables": await self.foreign_key_tables( database, table, pk_values ), diff --git a/tests/test_html.py b/tests/test_html.py index 757f3e6e..64ae7b2d 100644 --- a/tests/test_html.py +++ b/tests/test_html.py @@ -347,7 +347,7 @@ async def test_row_html_simple_primary_key(ds_client): assert ["id", "content"] == [th.string.strip() for th in table.select("thead th")] assert [ [ - '1', + '1', 'hello', ] ] == [[str(td) for td in tr.select("td")] for tr in table.select("tbody tr")] @@ -363,7 +363,7 @@ async def test_row_html_no_primary_key(ds_client): ] expected = [ [ - '1', + '1', '1', 'a1', 'b1', @@ -406,6 +406,26 @@ async def test_row_links_from_other_tables( assert link == expected_link +@pytest.mark.asyncio +async def test_row_foreign_key_links(ds_client): + # Row detail page should render foreign key values as hyperlinks + response = await ds_client.get("/fixtures/foreign_key_references/1") + assert response.status_code == 200 + soup = Soup(response.text, "html.parser") + # foreign_key_with_label=1 references simple_primary_key(id=1, content="hello") + td = soup.find("td", {"class": "col-foreign_key_with_label"}) + a = td.find("a") + assert a is not None, "Expected foreign key value to be a hyperlink" + assert a["href"] == "/fixtures/simple_primary_key/1" + assert a.text == "hello" + # Primary key column should be first and bold + table = soup.find("table") + headers = [th.text.strip() for th in table.select("thead th")] + assert headers[0] == "pk" + first_td = table.select("tbody tr td")[0] + assert first_td.find("strong") is not None, "PK value should be bold" + + @pytest.mark.asyncio @pytest.mark.parametrize( "path,expected", @@ -414,8 +434,8 @@ async def test_row_links_from_other_tables( "/fixtures/compound_primary_key/a,b", [ [ - 'a', - 'b', + 'a', + 'b', 'c', ] ], @@ -424,8 +444,8 @@ async def test_row_links_from_other_tables( "/fixtures/compound_primary_key/a~2Fb,~2Ec~2Dd", [ [ - 'a/b', - '.c-d', + 'a/b', + '.c-d', 'c', ] ], From e2c1e81ec9505f02566de840c1dba5ea7b0b121d Mon Sep 17 00:00:00 2001 From: Simon Willison Date: Mon, 9 Mar 2026 17:45:24 -0700 Subject: [PATCH 013/472] UI for selecting and re-ordering columns on the table page (#2662) New Web Component on table/view page with a dialog for selecting and re-ordering columns. Closes #2661 Refs #1298 --- datasette/static/app.css | 19 + datasette/static/column-chooser.js | 698 ++++++++++++++++++++++++++ datasette/static/navigation-search.js | 13 +- datasette/static/table.js | 58 +++ datasette/templates/table.html | 9 + datasette/views/table.py | 6 + tests/test_html.py | 23 +- tests/test_table_html.py | 63 +++ 8 files changed, 882 insertions(+), 7 deletions(-) create mode 100644 datasette/static/column-chooser.js diff --git a/datasette/static/app.css b/datasette/static/app.css index a7fc7fa3..4183b58e 100644 --- a/datasette/static/app.css +++ b/datasette/static/app.css @@ -63,6 +63,14 @@ em { } /* end reset */ +/* Modal CSS variables (shared by web components via Shadow DOM) */ +:root { + --modal-backdrop-bg: rgba(0, 0, 0, 0.5); + --modal-backdrop-blur: blur(4px); + --modal-border-radius: 0.75rem; + --modal-shadow: 0 20px 25px -5px rgba(0, 0, 0, 0.1), 0 10px 10px -5px rgba(0, 0, 0, 0.04); + --modal-animation-duration: 0.2s; +} body { margin: 0; @@ -795,6 +803,17 @@ p.zero-results { .filters input.filter-value { width: 140px; } + button.choose-columns-mobile { + display: inline-block; + padding: 0.5rem 1rem; + margin-bottom: 1em; + font-size: 0.9rem; + font-family: inherit; + background: white; + border: 1px solid #ccc; + border-radius: 5px; + cursor: pointer; + } } svg.dropdown-menu-icon { diff --git a/datasette/static/column-chooser.js b/datasette/static/column-chooser.js new file mode 100644 index 00000000..9680398c --- /dev/null +++ b/datasette/static/column-chooser.js @@ -0,0 +1,698 @@ +class ColumnChooser extends HTMLElement { + constructor() { + super(); + this.attachShadow({ mode: "open" }); + + // State + this._items = []; + this._checked = new Set(); + this._savedItems = null; + this._savedChecked = null; + this._onApply = null; + + // Drag state + this._ghost = null; + this._dragSrcIdx = null; + this._dropTargetIdx = null; + this._dropPosition = null; + this._ghostOffX = 0; + this._ghostOffY = 0; + this._autoScrollRAF = null; + this._lastPointerY = 0; + this._lastPointerX = 0; + this._SCROLL_ZONE = 72; + this._SCROLL_SPEED = 0.4; + + // Bound handlers + this._onMove = this._onMove.bind(this); + this._onUp = this._onUp.bind(this); + + this.shadowRoot.innerHTML = ` + + + + +
+ + +
+
+
+
+
    +
    + +
    + `; + + // DOM refs + this._dialog = this.shadowRoot.querySelector("dialog"); + this._listWrap = this.shadowRoot.getElementById("listWrap"); + this._dragList = this.shadowRoot.getElementById("dragList"); + this._pulseTop = this.shadowRoot.getElementById("pulseTop"); + this._pulseBot = this.shadowRoot.getElementById("pulseBot"); + this._selectAllBtn = this.shadowRoot.getElementById("selectAllBtn"); + this._deselectAllBtn = this.shadowRoot.getElementById("deselectAllBtn"); + this._cancelBtn = this.shadowRoot.getElementById("cancelBtn"); + this._applyBtn = this.shadowRoot.getElementById("applyBtn"); + this._countEl = this.shadowRoot.getElementById("selectedCount"); + this._footerEl = this.shadowRoot.getElementById("footerInfo"); + + // Event listeners + this._selectAllBtn.addEventListener("click", () => this._selectAll()); + this._deselectAllBtn.addEventListener("click", () => this._deselectAll()); + this._cancelBtn.addEventListener("click", () => this._close()); + this._applyBtn.addEventListener("click", () => this._apply()); + this._dialog.addEventListener("click", (e) => { + if (e.target === this._dialog) this._close(); + }); + this._dialog.addEventListener("cancel", (e) => { + e.preventDefault(); + this._close(); + }); + } + + /** + * Open the column chooser dialog. + * @param {Object} opts + * @param {string[]} opts.columns - All available column names, in display order. + * @param {string[]} opts.selected - Column names that should be pre-checked. + * @param {function(string[]): void} opts.onApply - Called with the selected columns in order when Apply is clicked. + */ + open({ columns, selected = [], onApply }) { + this._items = [...columns]; + this._checked = new Set(selected); + this._onApply = onApply || null; + + // Save state for cancel/restore + this._savedItems = [...this._items]; + this._savedChecked = new Set(this._checked); + + this._render(); + this._dialog.showModal(); + } + + // ── Internal methods ── + + _close() { + this._items = this._savedItems ? [...this._savedItems] : this._items; + this._checked = this._savedChecked + ? new Set(this._savedChecked) + : this._checked; + this._dialog.close(); + } + + _selectAll() { + this._items.forEach((col) => this._checked.add(col)); + this._dragList.querySelectorAll('input[type="checkbox"]').forEach((cb) => { + cb.checked = true; + }); + this._updateCounts(); + } + + _deselectAll() { + this._checked.clear(); + this._dragList.querySelectorAll('input[type="checkbox"]').forEach((cb) => { + cb.checked = false; + }); + this._updateCounts(); + } + + _apply() { + const selected = this._items.filter((col) => this._checked.has(col)); + this._dialog.close(); + if (this._onApply) { + this._onApply(selected); + } + } + + _render() { + this._dragList.innerHTML = ""; + this._items.forEach((col, i) => { + const li = document.createElement("li"); + li.className = "drag-item"; + li.dataset.idx = i; + li.innerHTML = ` + + + + + + + + + + + +
    + `; + + li.querySelector("input").addEventListener("change", (e) => { + e.target.checked ? this._checked.add(col) : this._checked.delete(col); + this._updateCounts(); + }); + + li.querySelector(".drag-handle").addEventListener("pointerdown", (e) => + this._startDrag(e, i), + ); + this._dragList.appendChild(li); + }); + + this._updateCounts(); + } + + _updateCounts() { + const n = this._checked.size; + this._countEl.textContent = `${n} of ${this._items.length} selected`; + this._footerEl.textContent = `${this._items.length} columns`; + } + + // ── Drag engine ── + + _startDrag(e, idx) { + e.preventDefault(); + this._dragSrcIdx = idx; + + const srcEl = this._dragList.children[idx]; + const rect = srcEl.getBoundingClientRect(); + + this._ghostOffX = e.clientX - rect.left; + this._ghostOffY = e.clientY - rect.top; + + // Build ghost inside shadow DOM + this._ghost = document.createElement("div"); + this._ghost.className = "drag-ghost"; + this._ghost.style.width = rect.width + "px"; + this._ghost.style.height = rect.height + "px"; + this._ghost.innerHTML = srcEl.innerHTML; + this._ghost.querySelector(".drop-indicator")?.remove(); + const h = this._ghost.querySelector(".drag-handle"); + if (h) h.style.color = "var(--accent)"; + this.shadowRoot.appendChild(this._ghost); + + srcEl.classList.add("is-dragging"); + this._positionGhost(e.clientX, e.clientY); + + document.addEventListener("pointermove", this._onMove); + document.addEventListener("pointerup", this._onUp); + document.addEventListener("pointercancel", this._onUp); + } + + _positionGhost(cx, cy) { + this._ghost.style.left = cx - this._ghostOffX + "px"; + this._ghost.style.top = cy - this._ghostOffY + "px"; + } + + _onMove(e) { + this._lastPointerX = e.clientX; + this._lastPointerY = e.clientY; + this._positionGhost(e.clientX, e.clientY); + this._updateDropTarget(e.clientY); + this._updateAutoScroll(e.clientY); + } + + _onUp() { + document.removeEventListener("pointermove", this._onMove); + document.removeEventListener("pointerup", this._onUp); + document.removeEventListener("pointercancel", this._onUp); + + this._stopAutoScroll(); + + const noMove = + this._dropTargetIdx === null || this._dropTargetIdx === this._dragSrcIdx; + this._clearDropIndicators(); + + let dest = null; + if (!noMove) { + const moved = this._items.splice(this._dragSrcIdx, 1)[0]; + dest = this._dropTargetIdx; + if (this._dropPosition === "after") dest++; + if (dest > this._dragSrcIdx) dest--; + this._items.splice(dest, 0, moved); + } + + this._dragSrcIdx = null; + this._dropTargetIdx = null; + this._dropPosition = null; + + const g = this._ghost; + this._ghost = null; + + if (noMove) { + if (g) g.remove(); + this._render(); + return; + } + + this._render(); + + if (g && dest !== null) { + const landedEl = this._dragList.children[dest]; + if (landedEl) { + landedEl.style.opacity = "0"; + const r = landedEl.getBoundingClientRect(); + g.getBoundingClientRect(); + g.style.transition = + "left 0.15s cubic-bezier(0.22, 1, 0.36, 1), top 0.15s cubic-bezier(0.22, 1, 0.36, 1), box-shadow 0.15s, opacity 0.1s 0.1s"; + g.style.left = r.left + "px"; + g.style.top = r.top + "px"; + g.style.boxShadow = "0 1px 4px rgba(0,0,0,0.08)"; + g.style.opacity = "0"; + setTimeout(() => { + g.remove(); + if (landedEl) landedEl.style.opacity = ""; + }, 160); + } else { + g.remove(); + } + } else if (g) { + g.remove(); + } + } + + _updateDropTarget(clientY) { + this._clearDropIndicators(); + const listItems = [ + ...this._dragList.querySelectorAll(".drag-item:not(.is-dragging)"), + ]; + if (!listItems.length) return; + + let best = null, + bestDist = Infinity; + listItems.forEach((li) => { + const r = li.getBoundingClientRect(); + const mid = r.top + r.height / 2; + const dist = Math.abs(clientY - mid); + if (dist < bestDist) { + bestDist = dist; + best = li; + } + }); + + if (!best) return; + const r = best.getBoundingClientRect(); + const mid = r.top + r.height / 2; + const above = clientY < mid; + const indic = best.querySelector(".drop-indicator"); + + this._dropTargetIdx = parseInt(best.dataset.idx); + this._dropPosition = above ? "before" : "after"; + + if (indic) { + indic.className = "drop-indicator " + (above ? "top" : "bottom"); + } + } + + _clearDropIndicators() { + this._dragList.querySelectorAll(".drop-indicator").forEach((el) => { + el.className = "drop-indicator"; + }); + } + + _updateAutoScroll(clientY) { + const rect = this._listWrap.getBoundingClientRect(); + const relY = clientY - rect.top; + const distTop = relY; + const distBot = rect.height - relY; + + const inTop = distTop < this._SCROLL_ZONE && distTop >= 0; + const inBot = distBot < this._SCROLL_ZONE && distBot >= 0; + + this._pulseTop.classList.toggle("active", inTop); + this._pulseBot.classList.toggle("active", inBot); + + if ((inTop || inBot) && !this._autoScrollRAF) { + let lastTime = null; + const loop = (ts) => { + if (!this._ghost) { + this._stopAutoScroll(); + return; + } + if (lastTime !== null) { + const dt = ts - lastTime; + const rect2 = this._listWrap.getBoundingClientRect(); + const relY2 = this._lastPointerY - rect2.top; + const dTop = relY2; + const dBot = rect2.height - relY2; + + if (dTop < this._SCROLL_ZONE && dTop >= 0) { + const factor = 1 - dTop / this._SCROLL_ZONE; + this._listWrap.scrollTop -= this._SCROLL_SPEED * dt * factor * 2.5; + } else if (dBot < this._SCROLL_ZONE && dBot >= 0) { + const factor = 1 - dBot / this._SCROLL_ZONE; + this._listWrap.scrollTop += this._SCROLL_SPEED * dt * factor * 2.5; + } else { + this._stopAutoScroll(); + return; + } + this._updateDropTarget(this._lastPointerY); + } + lastTime = ts; + this._autoScrollRAF = requestAnimationFrame(loop); + }; + this._autoScrollRAF = requestAnimationFrame(loop); + } + + if (!inTop && !inBot) this._stopAutoScroll(); + } + + _stopAutoScroll() { + if (this._autoScrollRAF) { + cancelAnimationFrame(this._autoScrollRAF); + this._autoScrollRAF = null; + } + this._pulseTop.classList.remove("active"); + this._pulseBot.classList.remove("active"); + } +} + +customElements.define("column-chooser", ColumnChooser); diff --git a/datasette/static/navigation-search.js b/datasette/static/navigation-search.js index 48de5c4f..95e7dfc5 100644 --- a/datasette/static/navigation-search.js +++ b/datasette/static/navigation-search.js @@ -19,19 +19,20 @@ class NavigationSearch extends HTMLElement { dialog { border: none; - border-radius: 0.75rem; + border-radius: var(--modal-border-radius, 0.75rem); padding: 0; max-width: 90vw; width: 600px; max-height: 80vh; - box-shadow: 0 20px 25px -5px rgba(0, 0, 0, 0.1), 0 10px 10px -5px rgba(0, 0, 0, 0.04); - animation: slideIn 0.2s ease-out; + box-shadow: var(--modal-shadow, 0 20px 25px -5px rgba(0, 0, 0, 0.1), 0 10px 10px -5px rgba(0, 0, 0, 0.04)); + animation: slideIn var(--modal-animation-duration, 0.2s) ease-out; } dialog::backdrop { - background: rgba(0, 0, 0, 0.5); - backdrop-filter: blur(4px); - animation: fadeIn 0.2s ease-out; + background: var(--modal-backdrop-bg, rgba(0, 0, 0, 0.5)); + backdrop-filter: var(--modal-backdrop-blur, blur(4px)); + -webkit-backdrop-filter: var(--modal-backdrop-blur, blur(4px)); + animation: fadeIn var(--modal-animation-duration, 0.2s) ease-out; } @keyframes slideIn { diff --git a/datasette/static/table.js b/datasette/static/table.js index 0caeeb91..c26dda5a 100644 --- a/datasette/static/table.js +++ b/datasette/static/table.js @@ -4,6 +4,7 @@ var DROPDOWN_HTML = `