|
| 1 | +<?php |
| 2 | + |
| 3 | +/* |
| 4 | + * This file is part of Psy Shell. |
| 5 | + * |
| 6 | + * (c) 2012-2018 Justin Hileman |
| 7 | + * |
| 8 | + * For the full copyright and license information, please view the LICENSE |
| 9 | + * file that was distributed with this source code. |
| 10 | + */ |
| 11 | + |
| 12 | +namespace Psy; |
| 13 | + |
| 14 | +use PhpParser\NodeTraverser; |
| 15 | +use PhpParser\Parser; |
| 16 | +use PhpParser\PrettyPrinter\Standard as Printer; |
| 17 | +use Psy\CodeCleaner\AbstractClassPass; |
| 18 | +use Psy\CodeCleaner\AssignThisVariablePass; |
| 19 | +use Psy\CodeCleaner\CalledClassPass; |
| 20 | +use Psy\CodeCleaner\CallTimePassByReferencePass; |
| 21 | +use Psy\CodeCleaner\ExitPass; |
| 22 | +use Psy\CodeCleaner\FinalClassPass; |
| 23 | +use Psy\CodeCleaner\FunctionContextPass; |
| 24 | +use Psy\CodeCleaner\FunctionReturnInWriteContextPass; |
| 25 | +use Psy\CodeCleaner\ImplicitReturnPass; |
| 26 | +use Psy\CodeCleaner\InstanceOfPass; |
| 27 | +use Psy\CodeCleaner\LeavePsyshAlonePass; |
| 28 | +use Psy\CodeCleaner\LegacyEmptyPass; |
| 29 | +use Psy\CodeCleaner\ListPass; |
| 30 | +use Psy\CodeCleaner\LoopContextPass; |
| 31 | +use Psy\CodeCleaner\MagicConstantsPass; |
| 32 | +use Psy\CodeCleaner\NamespacePass; |
| 33 | +use Psy\CodeCleaner\PassableByReferencePass; |
| 34 | +use Psy\CodeCleaner\RequirePass; |
| 35 | +use Psy\CodeCleaner\StrictTypesPass; |
| 36 | +use Psy\CodeCleaner\UseStatementPass; |
| 37 | +use Psy\CodeCleaner\ValidClassNamePass; |
| 38 | +use Psy\CodeCleaner\ValidConstantPass; |
| 39 | +use Psy\CodeCleaner\ValidConstructorPass; |
| 40 | +use Psy\CodeCleaner\ValidFunctionNamePass; |
| 41 | +use Psy\Exception\ParseErrorException; |
| 42 | + |
| 43 | +/** |
| 44 | + * A service to clean up user input, detect parse errors before they happen, |
| 45 | + * and generally work around issues with the PHP code evaluation experience. |
| 46 | + */ |
| 47 | +class CodeCleaner |
| 48 | +{ |
| 49 | + private $parser; |
| 50 | + private $printer; |
| 51 | + private $traverser; |
| 52 | + private $namespace; |
| 53 | + |
| 54 | + /** |
| 55 | + * CodeCleaner constructor. |
| 56 | + * |
| 57 | + * @param Parser $parser A PhpParser Parser instance. One will be created if not explicitly supplied |
| 58 | + * @param Printer $printer A PhpParser Printer instance. One will be created if not explicitly supplied |
| 59 | + * @param NodeTraverser $traverser A PhpParser NodeTraverser instance. One will be created if not explicitly supplied |
| 60 | + */ |
| 61 | + public function __construct(?Parser $parser = null, ?Printer $printer = null, ?NodeTraverser $traverser = null) |
| 62 | + { |
| 63 | + if ($parser === null) { |
| 64 | + $parserFactory = new ParserFactory(); |
| 65 | + $parser = $parserFactory->createParser(); |
| 66 | + } |
| 67 | + |
| 68 | + $this->parser = $parser; |
| 69 | + $this->printer = $printer ?: new Printer(); |
| 70 | + $this->traverser = $traverser ?: new NodeTraverser(); |
| 71 | + |
| 72 | + foreach ($this->getDefaultPasses() as $pass) { |
| 73 | + $this->traverser->addVisitor($pass); |
| 74 | + } |
| 75 | + } |
| 76 | + |
| 77 | + /** |
| 78 | + * Get default CodeCleaner passes. |
| 79 | + * |
| 80 | + * @return array |
| 81 | + */ |
| 82 | + private function getDefaultPasses() |
| 83 | + { |
| 84 | + $useStatementPass = new UseStatementPass(); |
| 85 | + $namespacePass = new NamespacePass($this); |
| 86 | + |
| 87 | + // Try to add implicit `use` statements and an implicit namespace, |
| 88 | + // based on the file in which the `debug` call was made. |
| 89 | + $this->addImplicitDebugContext([$useStatementPass, $namespacePass]); |
| 90 | + |
| 91 | + return [ |
| 92 | + // Validation passes |
| 93 | + new AbstractClassPass(), |
| 94 | + new AssignThisVariablePass(), |
| 95 | + new CalledClassPass(), |
| 96 | + new CallTimePassByReferencePass(), |
| 97 | + new FinalClassPass(), |
| 98 | + new FunctionContextPass(), |
| 99 | + new FunctionReturnInWriteContextPass(), |
| 100 | + new InstanceOfPass(), |
| 101 | + new LeavePsyshAlonePass(), |
| 102 | + new LegacyEmptyPass(), |
| 103 | + new ListPass(), |
| 104 | + new LoopContextPass(), |
| 105 | + new PassableByReferencePass(), |
| 106 | + new ValidConstructorPass(), |
| 107 | + |
| 108 | + // Rewriting shenanigans |
| 109 | + $useStatementPass, // must run before the namespace pass |
| 110 | + new ExitPass(), |
| 111 | + new ImplicitReturnPass(), |
| 112 | + new MagicConstantsPass(), |
| 113 | + $namespacePass, // must run after the implicit return pass |
| 114 | + new RequirePass(), |
| 115 | + new StrictTypesPass(), |
| 116 | + |
| 117 | + // Namespace-aware validation (which depends on aforementioned shenanigans) |
| 118 | + new ValidClassNamePass(), |
| 119 | + new ValidConstantPass(), |
| 120 | + new ValidFunctionNamePass(), |
| 121 | + ]; |
| 122 | + } |
| 123 | + |
| 124 | + /** |
| 125 | + * "Warm up" code cleaner passes when we're coming from a debug call. |
| 126 | + * |
| 127 | + * This is useful, for example, for `UseStatementPass` and `NamespacePass` |
| 128 | + * which keep track of state between calls, to maintain the current |
| 129 | + * namespace and a map of use statements. |
| 130 | + * |
| 131 | + * @param array $passes |
| 132 | + */ |
| 133 | + private function addImplicitDebugContext(array $passes) |
| 134 | + { |
| 135 | + $file = $this->getDebugFile(); |
| 136 | + if ($file === null) { |
| 137 | + return; |
| 138 | + } |
| 139 | + |
| 140 | + try { |
| 141 | + $code = @\file_get_contents($file); |
| 142 | + if (!$code) { |
| 143 | + return; |
| 144 | + } |
| 145 | + |
| 146 | + $stmts = $this->parse($code, true); |
| 147 | + if ($stmts === false) { |
| 148 | + return; |
| 149 | + } |
| 150 | + |
| 151 | + // Set up a clean traverser for just these code cleaner passes |
| 152 | + $traverser = new NodeTraverser(); |
| 153 | + foreach ($passes as $pass) { |
| 154 | + $traverser->addVisitor($pass); |
| 155 | + } |
| 156 | + |
| 157 | + $traverser->traverse($stmts); |
| 158 | + } catch (\Throwable $e) { |
| 159 | + // Don't care. |
| 160 | + } catch (\Exception $e) { |
| 161 | + // Still don't care. |
| 162 | + } |
| 163 | + } |
| 164 | + |
| 165 | + /** |
| 166 | + * Search the stack trace for a file in which the user called Psy\debug. |
| 167 | + * |
| 168 | + * @return string|null |
| 169 | + */ |
| 170 | + private static function getDebugFile() |
| 171 | + { |
| 172 | + $trace = \debug_backtrace(DEBUG_BACKTRACE_IGNORE_ARGS); |
| 173 | + |
| 174 | + foreach (\array_reverse($trace) as $stackFrame) { |
| 175 | + if (!self::isDebugCall($stackFrame)) { |
| 176 | + continue; |
| 177 | + } |
| 178 | + |
| 179 | + if (\preg_match('/eval\(/', $stackFrame['file'])) { |
| 180 | + \preg_match_all('/([^\(]+)\((\d+)/', $stackFrame['file'], $matches); |
| 181 | + |
| 182 | + return $matches[1][0]; |
| 183 | + } |
| 184 | + |
| 185 | + return $stackFrame['file']; |
| 186 | + } |
| 187 | + } |
| 188 | + |
| 189 | + /** |
| 190 | + * Check whether a given backtrace frame is a call to Psy\debug. |
| 191 | + * |
| 192 | + * @param array $stackFrame |
| 193 | + * |
| 194 | + * @return bool |
| 195 | + */ |
| 196 | + private static function isDebugCall(array $stackFrame) |
| 197 | + { |
| 198 | + $class = isset($stackFrame['class']) ? $stackFrame['class'] : null; |
| 199 | + $function = isset($stackFrame['function']) ? $stackFrame['function'] : null; |
| 200 | + |
| 201 | + return ($class === null && $function === 'Psy\debug') || |
| 202 | + ($class === 'Psy\Shell' && $function === 'debug'); |
| 203 | + } |
| 204 | + |
| 205 | + /** |
| 206 | + * Clean the given array of code. |
| 207 | + * |
| 208 | + * @throws ParseErrorException if the code is invalid PHP, and cannot be coerced into valid PHP |
| 209 | + * |
| 210 | + * @param array $codeLines |
| 211 | + * @param bool $requireSemicolons |
| 212 | + * |
| 213 | + * @return string|false Cleaned PHP code, False if the input is incomplete |
| 214 | + */ |
| 215 | + public function clean(array $codeLines, $requireSemicolons = false) |
| 216 | + { |
| 217 | + $stmts = $this->parse('<?php ' . \implode(PHP_EOL, $codeLines) . PHP_EOL, $requireSemicolons); |
| 218 | + if ($stmts === false) { |
| 219 | + return false; |
| 220 | + } |
| 221 | + |
| 222 | + // Catch fatal errors before they happen |
| 223 | + $stmts = $this->traverser->traverse($stmts); |
| 224 | + |
| 225 | + // Work around https://github.com/nikic/PHP-Parser/issues/399 |
| 226 | + $oldLocale = \setlocale(LC_NUMERIC, 0); |
| 227 | + \setlocale(LC_NUMERIC, 'C'); |
| 228 | + |
| 229 | + $code = $this->printer->prettyPrint($stmts); |
| 230 | + |
| 231 | + // Now put the locale back |
| 232 | + \setlocale(LC_NUMERIC, $oldLocale); |
| 233 | + |
| 234 | + return $code; |
| 235 | + } |
| 236 | + |
| 237 | + /** |
| 238 | + * Set the current local namespace. |
| 239 | + * |
| 240 | + * @param null|array $namespace (default: null) |
| 241 | + * |
| 242 | + * @return null|array |
| 243 | + */ |
| 244 | + public function setNamespace(?array $namespace = null) |
| 245 | + { |
| 246 | + $this->namespace = $namespace; |
| 247 | + } |
| 248 | + |
| 249 | + /** |
| 250 | + * Get the current local namespace. |
| 251 | + * |
| 252 | + * @return null|array |
| 253 | + */ |
| 254 | + public function getNamespace() |
| 255 | + { |
| 256 | + return $this->namespace; |
| 257 | + } |
| 258 | + |
| 259 | + /** |
| 260 | + * Lex and parse a block of code. |
| 261 | + * |
| 262 | + * @see Parser::parse |
| 263 | + * |
| 264 | + * @throws ParseErrorException for parse errors that can't be resolved by |
| 265 | + * waiting a line to see what comes next |
| 266 | + * |
| 267 | + * @param string $code |
| 268 | + * @param bool $requireSemicolons |
| 269 | + * |
| 270 | + * @return array|false A set of statements, or false if incomplete |
| 271 | + */ |
| 272 | + protected function parse($code, $requireSemicolons = false) |
| 273 | + { |
| 274 | + try { |
| 275 | + return $this->parser->parse($code); |
| 276 | + } catch (\PhpParser\Error $e) { |
| 277 | + if ($this->parseErrorIsUnclosedString($e, $code)) { |
| 278 | + return false; |
| 279 | + } |
| 280 | + |
| 281 | + if ($this->parseErrorIsUnterminatedComment($e, $code)) { |
| 282 | + return false; |
| 283 | + } |
| 284 | + |
| 285 | + if ($this->parseErrorIsTrailingComma($e, $code)) { |
| 286 | + return false; |
| 287 | + } |
| 288 | + |
| 289 | + if (!$this->parseErrorIsEOF($e)) { |
| 290 | + throw ParseErrorException::fromParseError($e); |
| 291 | + } |
| 292 | + |
| 293 | + if ($requireSemicolons) { |
| 294 | + return false; |
| 295 | + } |
| 296 | + |
| 297 | + try { |
| 298 | + // Unexpected EOF, try again with an implicit semicolon |
| 299 | + return $this->parser->parse($code . ';'); |
| 300 | + } catch (\PhpParser\Error $e) { |
| 301 | + return false; |
| 302 | + } |
| 303 | + } |
| 304 | + } |
| 305 | + |
| 306 | + private function parseErrorIsEOF(\PhpParser\Error $e) |
| 307 | + { |
| 308 | + $msg = $e->getRawMessage(); |
| 309 | + |
| 310 | + return ($msg === 'Unexpected token EOF') || (\strpos($msg, 'Syntax error, unexpected EOF') !== false); |
| 311 | + } |
| 312 | + |
| 313 | + /** |
| 314 | + * A special test for unclosed single-quoted strings. |
| 315 | + * |
| 316 | + * Unlike (all?) other unclosed statements, single quoted strings have |
| 317 | + * their own special beautiful snowflake syntax error just for |
| 318 | + * themselves. |
| 319 | + * |
| 320 | + * @param \PhpParser\Error $e |
| 321 | + * @param string $code |
| 322 | + * |
| 323 | + * @return bool |
| 324 | + */ |
| 325 | + private function parseErrorIsUnclosedString(\PhpParser\Error $e, $code) |
| 326 | + { |
| 327 | + if ($e->getRawMessage() !== 'Syntax error, unexpected T_ENCAPSED_AND_WHITESPACE') { |
| 328 | + return false; |
| 329 | + } |
| 330 | + |
| 331 | + try { |
| 332 | + $this->parser->parse($code . "';"); |
| 333 | + } catch (\Exception $e) { |
| 334 | + return false; |
| 335 | + } |
| 336 | + |
| 337 | + return true; |
| 338 | + } |
| 339 | + |
| 340 | + private function parseErrorIsUnterminatedComment(\PhpParser\Error $e, $code) |
| 341 | + { |
| 342 | + return $e->getRawMessage() === 'Unterminated comment'; |
| 343 | + } |
| 344 | + |
| 345 | + private function parseErrorIsTrailingComma(\PhpParser\Error $e, $code) |
| 346 | + { |
| 347 | + return ($e->getRawMessage() === 'A trailing comma is not allowed here') && (\substr(\rtrim($code), -1) === ','); |
| 348 | + } |
| 349 | +} |
0 commit comments