<?php
/**
* PHP lexer which can handle fragments of source code, e.g. for syntax
* highlighting of inline snippets. This is largely based on Pygments, but
* alters some rules to better annotate symbols from source fragments.
*
* This lexer is not suitable for parser construction; it always lexes any
* input stream, even if the input is not PHP.
*/
final class PhutilPHPFragmentLexer extends PhutilLexer {
protected function getRawRules() {
$keywords = array(
'abstract',
'and',
'array',
'as',
'break',
'case',
'catch',
'callable',
'class',
'clone',
'const',
'continue',
'default',
'declare',
'do',
'else',
'elseif',
'enddeclare',
'endfor',
'endforeach',
'endif',
'endswitch',
'endwhile',
'extends',
'final',
'for',
'foreach',
'function',
'global',
'goto',
'if',
'implements',
'include',
'include_once',
'insteadof',
'instanceof',
'interface',
'list',
'namespace',
'new',
'not',
'or',
'parent',
'private',
'protected',
'public',
'require',
'require_once',
'return',
'self',
'static',
'switch',
'throw',
'try',
'use',
'var',
'while',
'xor',
);
$constants = array(
'__CLASS__',
'__COMPILER_HALT_OFFSET__',
'__DIR__',
'__FILE__',
'__FUNCTION__',
'__LINE__',
'__METHOD__',
'__NAMESPACE__',
'__TRAIT__',
'true',
'false',
'null',
);
$nonsemantic_rules = array(
array('\\s+', null),
array('(#|//)', 'c', 'line_comment'),
// This prevents the doc comment rule from over-consuming when we see
// an empty coment.
array('/\\*\\*/', 'cm'),
array('/\\*\\*.*?\\*/', 'sd'),
array('/\\*.*?\\*/', 'cm'),
// Unterminated multi-line comment.
array('/\\*.*$', 'cm'),
);
$identifier_pattern = "[a-zA-Z_\x7f-\xff][a-zA-Z0-9_\x7f-\xff]*";
$identifier_ns_pattern = "[\\\\a-zA-Z_\x7f-\xff][\\\\a-zA-Z0-9_\x7f-\xff]*";
return array(
'start' => array(
array('<\\?(?i:php)?', 'cp', 'php'),
array('[^<]+', null),
array('<', null),
),
'php' => array_merge(array(
array('\\?>', 'cp', '!pop'),
array(
'<<<([\'"]?)('.$identifier_pattern.')\\1\\n.*?\\n\\2\\;?\\n',
's',
),
), $nonsemantic_rules, array(
array('(?i:__halt_compiler)\\b', 'cp', 'halt_compiler'),
array('(->|::)', 'o', 'attr'),
array('[~!%^&*+=|:.<>/?@-]+', 'o'),
array('[\\[\\]{}();,]', 'o'),
// After 'new', try to match an unadorned symbol.
array('(?i:new|instanceof)\\b', 'k', 'possible_classname'),
array('(?i:function)\\b', 'k', 'function_definition'),
// After 'extends' or 'implements', match a list of classes/interfaces.
array('(?i:extends|implements)\\b', 'k', 'class_list'),
array('(?i:catch)\\b', 'k', 'catch'),
array('(?i:'.implode('|', $keywords).')\\b', 'k'),
array('(?i:'.implode('|', $constants).')\\b', 'kc'),
array('\\$+'.$identifier_pattern, 'nv'),
// Match "f(" as a function and "C::" as a class. These won't work
// if you put a comment between the symbol and the operator, but
// that's a bizarre usage.
array($identifier_ns_pattern.'(?=\s*[\\(])', 'nf'),
array(
$identifier_ns_pattern.'(?=\s*::)',
'nc',
'context_attr',
array(
'context' => 'push',
),
),
array($identifier_ns_pattern, 'no'),
array('(\\d+\\.\\d*|\\d*\\.\\d+)([eE][+-]?[0-9]+)?', 'mf'),
array('\\d+[eE][+-]?[0-9]+', 'mf'),
array('0[0-7]+', 'mo'),
array('0[xX][a-fA-F0-9]+', 'mh'),
array('0[bB][0-1]+', 'm'),
array('\d+', 'mi'),
array("'", 's1', 'string1'),
array('`', 'sb', 'stringb'),
array('"', 's2', 'string2'),
array('.', null),
)),
// We've just matched a class name, with a "::" lookahead. The name of
// the class is on the top of the context stack. We want to try to match
// the attribute or method (e.g., "X::C" or "X::f()").
'context_attr' => array_merge($nonsemantic_rules, array(
array('::', 'o'),
array(
$identifier_pattern.'(?=\s*[\\(])',
'nf',
'!pop',
array(
'context' => 'pop',
),
),
array(
$identifier_pattern,
'na',
'!pop',
array(
'context' => 'pop',
),
),
array(
'',
null,
'!pop',
array(
'context' => 'discard',
),
),
)),
// After '->' or '::', a symbol is an attribute name. Note that we end
// up in 'context_attr' instead of here in some cases.
'attr' => array_merge($nonsemantic_rules, array(
array($identifier_pattern, 'na', '!pop'),
array('', null, '!pop'),
)),
// After 'new', a symbol is a class name.
'possible_classname' => array_merge($nonsemantic_rules, array(
array($identifier_ns_pattern, 'nc', '!pop'),
array('', null, '!pop'),
)),
'string1' => array(
array('[^\'\\\\]+', 's1'),
array("'", 's1', '!pop'),
array('\\\\.', 'k'),
array('\\\\$', 'k'),
),
'stringb' => array(
array('[^`\\\\]+', 'sb'),
array('`', 'sb', '!pop'),
array('\\\\.', 'k'),
array('\\\\$', 'k'),
),
'string2' => array(
array('[^"\\\\]+', 's2'),
array('"', 's2', '!pop'),
array('\\\\.', 'k'),
array('\\\\$', 'k'),
),
// In a function definition (after "function"), we don't link the name
// as a "nf" (name.function) since it is its own definition.
'function_definition' => array_merge($nonsemantic_rules, array(
array('&', 'o'),
array('\\(', 'o', '!pop'),
array($identifier_pattern, 'no', '!pop'),
array('', null, '!pop'),
)),
// For "//" and "#" comments, we need to break out if we see "?" followed
// by ">".
'line_comment' => array(
array('[^?\\n]+', 'c'),
array('\\n', null, '!pop'),
array('(?=\\?>)', null, '!pop'),
array('\\?', 'c'),
),
// We've seen __halt_compiler. Grab the '();' afterward and then eat
// the rest of the file as raw data.
'halt_compiler' => array_merge($nonsemantic_rules, array(
array('[()]', 'o'),
array(';', 'o', 'compiler_halted'),
array('\\?>', 'o', 'compiler_halted'),
// Just halt on anything else.
array('', null, 'compiler_halted'),
)),
// __halt_compiler has taken effect.
'compiler_halted' => array(
array('.+', null),
),
'class_list' => array_merge($nonsemantic_rules, array(
array(',', 'o'),
array('(?i:implements)', 'k'),
array($identifier_ns_pattern, 'nc'),
array('', null, '!pop'),
)),
'catch' => array_merge($nonsemantic_rules, array(
array('\\(', 'o'),
array($identifier_ns_pattern, 'nc'),
array('', null, '!pop'),
)),
);
}
}