Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
100.00% |
285 / 285 |
|
100.00% |
17 / 17 |
CRAP | |
100.00% |
1 / 1 |
| SyntaxChecker | |
100.00% |
285 / 285 |
|
100.00% |
17 / 17 |
89 | |
100.00% |
1 / 1 |
| __construct | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
1 | |||
| start | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
4 | |||
| desugar | |
100.00% |
62 / 62 |
|
100.00% |
1 / 1 |
23 | |||
| desugarAndOr | |
100.00% |
47 / 47 |
|
100.00% |
1 / 1 |
3 | |||
| newNode | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| newNodeReplaceType | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| newNodeMapAll | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| newNodeMapExceptFirst | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
2 | |||
| newNodeBinop | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
1 | |||
| newNodeNamedBinop | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
1 | |||
| check | |
100.00% |
81 / 81 |
|
100.00% |
1 / 1 |
28 | |||
| mapUnion | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
4 | |||
| mapIntersect | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
3 | |||
| assignVar | |
100.00% |
9 / 9 |
|
100.00% |
1 / 1 |
2 | |||
| lookupVar | |
100.00% |
23 / 23 |
|
100.00% |
1 / 1 |
5 | |||
| checkArgCount | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
5 | |||
| isReservedIdentifier | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
3 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace MediaWiki\Extension\AbuseFilter\Parser; |
| 4 | |
| 5 | use InvalidArgumentException; |
| 6 | use LogicException; |
| 7 | use MediaWiki\Extension\AbuseFilter\KeywordsManager; |
| 8 | use MediaWiki\Extension\AbuseFilter\Parser\Exception\InternalException; |
| 9 | use MediaWiki\Extension\AbuseFilter\Parser\Exception\UserVisibleException; |
| 10 | use Message; |
| 11 | |
| 12 | /** |
| 13 | * SyntaxChecker statically analyzes the code without actually running it. |
| 14 | * Currently, it only checks for |
| 15 | * |
| 16 | * - unbound variables |
| 17 | * - unused variables: note that a := 1; a := 1; a |
| 18 | * is considered OK even though the first `a` seems unused |
| 19 | * because the pattern "a := null; if ... then (a := ...) end; ..." |
| 20 | * should not count first `a` as unused. |
| 21 | * - assignment to built-in identifiers |
| 22 | * - invalid function call (arity mismatch, non-valid function) |
| 23 | * - first-order information of `set_var` and `set` |
| 24 | * |
| 25 | * Because it doesn't cover all checks that the current Check Syntax does, |
| 26 | * it is currently complementary to the current Check Syntax. |
| 27 | * In the future, it could subsume the current Check Syntax, and could be |
| 28 | * extended to perform type checking or type inference. |
| 29 | */ |
| 30 | class SyntaxChecker { |
| 31 | /** |
| 32 | * @var AFPTreeNode|null Root of the AST to check |
| 33 | */ |
| 34 | private $treeRoot; |
| 35 | |
| 36 | /** @var KeywordsManager */ |
| 37 | private $keywordsManager; |
| 38 | |
| 39 | public const MCONSERVATIVE = 'MODE_CONSERVATIVE'; |
| 40 | public const MLIBERAL = 'MODE_LIBERAL'; |
| 41 | public const DUMMYPOS = 0; |
| 42 | public const CACHE_VERSION = 1; |
| 43 | |
| 44 | /** |
| 45 | * @var string The mode of checking. The value should be either |
| 46 | * |
| 47 | * - MLIBERAL: which guarantees that all user-defined variables |
| 48 | * will be bound, but incompatible with what the evaluator currently |
| 49 | * permits. E.g., |
| 50 | * |
| 51 | * if true then (a := 1) else null end; a |
| 52 | * |
| 53 | * is rejected in this mode, even though `a` is in fact always bound. |
| 54 | * |
| 55 | * - MCONSERVATIVE which is compatible with what the evaluator |
| 56 | * currently permits, but could allow undefined variables to occur. |
| 57 | * E.g., |
| 58 | * |
| 59 | * if false then (a := 1) else null end; a |
| 60 | * |
| 61 | * is accepted in this mode, even though `a` is in fact always unbound. |
| 62 | */ |
| 63 | private $mode; |
| 64 | |
| 65 | /** |
| 66 | * @var bool Whether we want to check for unused variables |
| 67 | */ |
| 68 | private $checkUnusedVars; |
| 69 | |
| 70 | /** |
| 71 | * @param AFPSyntaxTree $tree |
| 72 | * @param KeywordsManager $keywordsManager |
| 73 | * @param string $mode |
| 74 | * @param bool $checkUnusedVars |
| 75 | */ |
| 76 | public function __construct( |
| 77 | AFPSyntaxTree $tree, |
| 78 | KeywordsManager $keywordsManager, |
| 79 | string $mode = self::MCONSERVATIVE, |
| 80 | bool $checkUnusedVars = false |
| 81 | ) { |
| 82 | $this->treeRoot = $tree->getRoot(); |
| 83 | $this->keywordsManager = $keywordsManager; |
| 84 | $this->mode = $mode; |
| 85 | $this->checkUnusedVars = $checkUnusedVars; |
| 86 | } |
| 87 | |
| 88 | /** |
| 89 | * Start the static analysis |
| 90 | * |
| 91 | * @throws UserVisibleException |
| 92 | */ |
| 93 | public function start(): void { |
| 94 | if ( !$this->treeRoot ) { |
| 95 | return; |
| 96 | } |
| 97 | $bound = $this->check( $this->desugar( $this->treeRoot ), [] ); |
| 98 | $unused = array_keys( array_filter( $bound, static function ( $v ) { |
| 99 | return !$v; |
| 100 | } ) ); |
| 101 | if ( $this->checkUnusedVars && $unused ) { |
| 102 | throw new UserVisibleException( |
| 103 | 'unusedvars', |
| 104 | self::DUMMYPOS, |
| 105 | [ Message::listParam( $unused, 'comma' ) ] |
| 106 | ); |
| 107 | } |
| 108 | } |
| 109 | |
| 110 | /** |
| 111 | * Remove syntactic sugar so that we don't need to deal with |
| 112 | * too many cases. |
| 113 | * |
| 114 | * This could benefit the evaluator as well, but for now, this is |
| 115 | * only used for static analysis. |
| 116 | * |
| 117 | * Postcondition: |
| 118 | * - The tree will not contain nodes of |
| 119 | * type ASSIGNMENT, LOGIC, COMPARE, SUM_REL, MUL_REL, POW, |
| 120 | * KEYWORD_OPERATOR, and ARRAY_INDEX |
| 121 | * - The tree may additionally contain a node of type BINOP. |
| 122 | * - The tree should not have set_var function application. |
| 123 | * - Conditionals will have both branches. |
| 124 | * |
| 125 | * @param AFPTreeNode $node |
| 126 | * @return AFPTreeNode |
| 127 | * @throws InternalException |
| 128 | */ |
| 129 | private function desugar( AFPTreeNode $node ): AFPTreeNode { |
| 130 | switch ( $node->type ) { |
| 131 | case AFPTreeNode::ATOM: |
| 132 | return $node; |
| 133 | |
| 134 | case AFPTreeNode::FUNCTION_CALL: |
| 135 | if ( $node->children[0] === 'set_var' ) { |
| 136 | $node->children[0] = 'set'; |
| 137 | } |
| 138 | return $this->newNodeMapExceptFirst( $node ); |
| 139 | |
| 140 | case AFPTreeNode::ARRAY_INDEX: |
| 141 | return $this->newNodeNamedBinop( $node, '[]' ); |
| 142 | |
| 143 | case AFPTreeNode::POW: |
| 144 | return $this->newNodeNamedBinop( $node, '**' ); |
| 145 | |