Protect api.php and rest.php entry points from crawlers

[Copilot]

Both api.php and rest.php were entirely unprotected — a crawler blocked on index.php?action=history could trivially fetch identical data via api.php?action=query&prop=revisions or /page/{title}/history. This adds hook-based protection for both entry points with the same bypass semantics (registered users, allowed IPs) as the existing index.php protection.

New configuration

Setting	Default	Description
$wgCrawlerProtectedApiModules	[]	Action API module names to block (case-insensitive; matches top-level actions and query sub-modules)
$wgCrawlerProtectedRestPaths	[]	REST path glob patterns to block via fnmatch()

Both default to empty — operators must opt in. This avoids silently breaking existing anonymous API consumers on upgrade.

Example configuration protecting the most crawler-attractive endpoints:

$wgCrawlerProtectedApiModules = [ 'compare', 'parse', 'revisions', 'recentchanges', 'backlinks' ];
$wgCrawlerProtectedRestPaths  = [ '/page/*/history', '/revision/*/compare/*' ];

Implementation

• ApiCheckCanExecute hook — fires for every API module before execution; denies with a message key when the module name is in the protected list.
• RestCheckCanExecute hook — fires on MW 1.42+; denies with an HttpException(403) when the request path matches a configured glob pattern. Silently inactive on MW 1.39–1.41 (hook predates those versions).
• Both hooks get the user from their respective hook parameters ($user for API, $handler->getAuthority()->getUser() for REST), then delegate the allow/deny decision to CrawlerProtectionService.
• Hooks.php is restructured to bracketted namespace blocks so it can conditionally define no-op stub interfaces for ApiCheckCanExecuteHook and RestCheckCanExecuteHook on MW versions that don't ship them — allowing the class to unconditionally implement both without fatal errors.
• CrawlerProtectionService gains checkApiModule(), isProtectedApiModule(), checkRestPath(), and isProtectedRestPath().

Coverage table (added to README)

Entry point	Protected
index.php page views / actions / diffs	✅ Always
index.php Special pages	✅ Always
api.php Action API modules	✅ Configurable via $wgCrawlerProtectedApiModules
rest.php REST API paths	✅ Configurable via $wgCrawlerProtectedRestPaths (MW 1.42+)

• Fixes api.php and rest.php expose the same expensive data with no protection (functional) #48

[Copilot]

Pull request overview

Adds crawler-protection coverage to non-index.php entry points by wiring Action API and REST API pre-execution hooks into the existing CrawlerProtectionService, with opt-in configuration to avoid breaking existing anonymous consumers.

Changes:

• Register and implement ApiCheckCanExecute and RestCheckCanExecute hook handlers that delegate allow/deny decisions to CrawlerProtectionService.
• Add new config settings for protected API modules and REST path patterns, plus README documentation for entry-point coverage and configuration.
• Extend unit tests and test stubs to cover the new hook handlers and service methods.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
includes/Hooks.php	Adds API/REST hook handlers and compatibility stubs; delegates to service for allow/deny.
includes/CrawlerProtectionService.php	Adds API module and REST path protection logic (check* + isProtected*).
extension.json	Registers new hooks and exposes new config keys.
README.md	Documents entry-point coverage and new configuration options.
tests/phpunit/unit/HooksTest.php	Adds delegation tests for the new hook handlers.
tests/phpunit/unit/CrawlerProtectionServiceTest.php	Adds unit tests for API/REST protection logic and configuration.
tests/phpunit/namespaced-stubs.php	Adds stub hook interfaces and a REST HttpException stub for unit tests.

[jeffw16]

@copilot Please address failing pipelines for REL1_39

[Copilot]

@copilot Please address failing pipelines for REL1_39

Fixed in b07351d — all CI jobs are now green (run #150).

Three failures, all rooted in the compat stub interfaces added to Hooks.php:

• Unit Tests (REL1_39): onRestCheckCanExecute called wfMessage()->plain(), which triggers "Premature access to service container" in the unit-test suite. Now uses LocalizedHttpException + MessageValue, which needs no services.
• Code Style (REL1_39): Generic.Classes.DuplicateClassName between Hooks.php and tests/phpunit/namespaced-stubs.php.
• Static Analysis (all versions): PhanRedefineClass / PhanRedefinedInheritedInterface for ApiCheckCanExecuteHook.

Hooks.php no longer redeclares core hook interfaces and no longer uses bracketted namespace blocks — it's back to the original single-namespace layout with the two new handler methods appended. HookContainer dispatches handlers by method name, so implementing the interfaces isn't required, and skipping them is what keeps the class loadable on MW versions that don't ship RestCheckCanExecuteHook.

[jeffw16]

@copilot Pull the latest changes from the main branch and then review the code

[jeffw16]

@copilot Please fix these pipeline errors:

FILE: ...erProtection/extensions/CrawlerProtection/tests/phpunit/unit/HooksTest.php
--------------------------------------------------------------------------------
FOUND 4 ERRORS AFFECTING 4 LINES
--------------------------------------------------------------------------------
 371 | ERROR | Missing class property doc comment
     |       | (MediaWiki.Commenting.PropertyDocumentation.MissingDocumentationPrivate)
 383 | ERROR | Missing class property doc comment
     |       | (MediaWiki.Commenting.PropertyDocumentation.MissingDocumentationPrivate)
 414 | ERROR | Missing class property doc comment
     |       | (MediaWiki.Commenting.PropertyDocumentation.MissingDocumentationPrivate)
 426 | ERROR | Missing class property doc comment
     |       | (MediaWiki.Commenting.PropertyDocumentation.MissingDocumentationPrivate)
--------------------------------------------------------------------------------