| 1 |
<?php |
| 2 |
|
| 3 |
/** |
| 4 |
* A recursive descent parser. |
| 5 |
* |
| 6 |
* This is a dynamic recursive descent parser that can parse LL grammars. |
| 7 |
* |
| 8 |
* @TODO: Add a detailed description and list the properties that a grammar must |
| 9 |
* satisfy in order to be supported by this parser (e.g., no left recursion). |
| 10 |
*/ |
| 11 |
class WP_Parser { |
| 12 |
protected $grammar; |
| 13 |
protected $tokens; |
| 14 |
protected $position; |
| 15 |
|
| 16 |
public function __construct( WP_Parser_Grammar $grammar, array $tokens ) { |
| 17 |
$this->grammar = $grammar; |
| 18 |
$this->tokens = $tokens; |
| 19 |
$this->position = 0; |
| 20 |
} |
| 21 |
|
| 22 |
public function parse() { |
| 23 |
// @TODO: Make the starting rule lookup non-grammar-specific. |
| 24 |
$query_rule_id = $this->grammar->get_rule_id( 'query' ); |
| 25 |
$ast = $this->parse_recursive( $query_rule_id ); |
| 26 |
return false === $ast ? null : $ast; |
| 27 |
} |
| 28 |
|
| 29 |
private function parse_recursive( $rule_id ) { |
| 30 |
$is_terminal = $rule_id <= $this->grammar->highest_terminal_id; |
| 31 |
if ( $is_terminal ) { |
| 32 |
if ( $this->position >= count( $this->tokens ) ) { |
| 33 |
return false; |
| 34 |
} |
| 35 |
|
| 36 |
if ( WP_Parser_Grammar::EMPTY_RULE_ID === $rule_id ) { |
| 37 |
return true; |
| 38 |
} |
| 39 |
|
| 40 |
if ( $this->tokens[ $this->position ]->id === $rule_id ) { |
| 41 |
++$this->position; |
| 42 |
return $this->tokens[ $this->position - 1 ]; |
| 43 |
} |
| 44 |
return false; |
| 45 |
} |
| 46 |
|
| 47 |
$branches = $this->grammar->rules[ $rule_id ]; |
| 48 |
if ( ! count( $branches ) ) { |
| 49 |
return false; |
| 50 |
} |
| 51 |
|
| 52 |
// Bale out from processing the current branch if none of its rules can |
| 53 |
// possibly match the current token. |
| 54 |
if ( isset( $this->grammar->lookahead_is_match_possible[ $rule_id ] ) ) { |
| 55 |
$token_id = $this->tokens[ $this->position ]->id; |
| 56 |
if ( |
| 57 |
! isset( $this->grammar->lookahead_is_match_possible[ $rule_id ][ $token_id ] ) && |
| 58 |
! isset( $this->grammar->lookahead_is_match_possible[ $rule_id ][ WP_Parser_Grammar::EMPTY_RULE_ID ] ) |
| 59 |
) { |
| 60 |
return false; |
| 61 |
} |
| 62 |
} |
| 63 |
|
| 64 |
$rule_name = $this->grammar->rule_names[ $rule_id ]; |
| 65 |
$starting_position = $this->position; |
| 66 |
foreach ( $branches as $branch ) { |
| 67 |
$this->position = $starting_position; |
| 68 |
$node = new WP_Parser_Node( $rule_id, $rule_name ); |
| 69 |
$branch_matches = true; |
| 70 |
foreach ( $branch as $subrule_id ) { |
| 71 |
$subnode = $this->parse_recursive( $subrule_id ); |
| 72 |
if ( false === $subnode ) { |
| 73 |
$branch_matches = false; |
| 74 |
break; |
| 75 |
} elseif ( true === $subnode ) { |
| 76 |
/* |
| 77 |
* The subrule was matched without actually matching a token. |
| 78 |
* This means a special empty "ε" (epsilon) rule was matched. |
| 79 |
* An "ε" rule in a grammar matches an empty input of 0 bytes. |
| 80 |
* It is used to represent optional grammar productions. |
| 81 |
*/ |
| 82 |
continue; |
| 83 |
} elseif ( is_array( $subnode ) && 0 === count( $subnode ) ) { |
| 84 |
continue; |
| 85 |
} |
| 86 |
if ( is_array( $subnode ) && ! count( $subnode ) ) { |
| 87 |
continue; |
| 88 |
} |
| 89 |
if ( isset( $this->grammar->fragment_ids[ $subrule_id ] ) ) { |
| 90 |
$node->merge_fragment( $subnode ); |
| 91 |
} else { |
| 92 |
$node->append_child( $subnode ); |
| 93 |
} |
| 94 |
} |
| 95 |
|
| 96 |
// Negative lookahead for INTO after a valid SELECT statement. |
| 97 |
// If we match a SELECT statement, but there is an INTO keyword after it, |
| 98 |
// we're in the wrong branch and need to leave matching to a later rule. |
| 99 |
// @TODO: Extract this to the "WP_MySQL_Parser" class, or add support |
| 100 |
// for right-associative rules, which could solve this. |
| 101 |
// See: https://github.com/mysql/mysql-workbench/blob/8.0.38/library/parsers/grammars/MySQLParser.g4#L994 |
| 102 |
// See: https://github.com/antlr/antlr4/issues/488 |
| 103 |
$la = $this->tokens[ $this->position ] ?? null; |
| 104 |
if ( $la && 'selectStatement' === $rule_name && WP_MySQL_Lexer::INTO_SYMBOL === $la->id ) { |
| 105 |
$branch_matches = false; |
| 106 |
} |
| 107 |
|
| 108 |
if ( true === $branch_matches ) { |
| 109 |
break; |
| 110 |
} |
| 111 |
} |
| 112 |
|
| 113 |
if ( ! $branch_matches ) { |
| 114 |
$this->position = $starting_position; |
| 115 |
return false; |
| 116 |
} |
| 117 |
|
| 118 |
if ( ! $node->has_child() ) { |
| 119 |
return true; |
| 120 |
} |
| 121 |
|
| 122 |
return $node; |
| 123 |
} |
| 124 |
} |
| 125 |
|
