tests/phpunit/includes/utils/StringUtilsTest.php

   1 <?php
   2
   3 class StringUtilsTest extends MediaWikiTestCase {
   4
   5         /**
   6          * This tests StringUtils::isUtf8 whenever we have the mbstring extension
   7          * loaded.
   8          *
   9          * @covers StringUtils::isUtf8
  10          * @dataProvider provideStringsForIsUtf8Check
  11          */
  12         public function testIsUtf8WithMbstring( $expected, $string ) {
  13                 if ( !function_exists( 'mb_check_encoding' ) ) {
  14                         $this->markTestSkipped( 'Test requires the mbstring PHP extension' );
  15                 }
  16                 $this->assertEquals( $expected,
  17                         StringUtils::isUtf8( $string ),
  18                         'Testing string "' . $this->escaped( $string ) . '" with mb_check_encoding'
  19                 );
  20         }
  21
  22         /**
  23          * This tests StringUtils::isUtf8 making sure we use the pure PHP
  24          * implementation used as a fallback when mb_check_encoding() is
  25          * not available.
  26          *
  27          * @covers StringUtils::isUtf8
  28          * @dataProvider provideStringsForIsUtf8Check
  29          */
  30         public function testIsUtf8WithPhpFallbackImplementation( $expected, $string ) {
  31                 $this->assertEquals( $expected,
  32                         StringUtils::isUtf8( $string, /** disable mbstring: */true ),
  33                         'Testing string "' . $this->escaped( $string ) . '" with pure PHP implementation'
  34                 );
  35         }
  36
  37         /**
  38          * Print high range characters as an hexadecimal
  39          */
  40         function escaped( $string ) {
  41                 $escaped = '';
  42                 $length = strlen( $string );
  43                 for ( $i = 0; $i < $length; $i++ ) {
  44                         $char = $string[$i];
  45                         $val = ord( $char );
  46                         if ( $val > 127 ) {
  47                                 $escaped .= '\x' . dechex( $val );
  48                         } else {
  49                                 $escaped .= $char;
  50                         }
  51                 }
  52
  53                 return $escaped;
  54         }
  55
  56         /**
  57          * See also "UTF-8 decoder capability and stress test" by
  58          * Markus Kuhn:
  59          * http://www.cl.cam.ac.uk/~mgk25/ucs/examples/UTF-8-test.txt
  60          */
  61         public static function provideStringsForIsUtf8Check() {
  62                 // Expected return values for StringUtils::isUtf8()
  63                 $PASS = true;
  64                 $FAIL = false;
  65
  66                 return array(
  67                         'some ASCII' => array( $PASS, 'Some ASCII' ),
  68                         'euro sign' => array( $PASS, "Euro sign €" ),
  69
  70                         'first possible sequence 1 byte' => array( $PASS, "\x00" ),
  71                         'first possible sequence 2 bytes' => array( $PASS, "\xc2\x80" ),
  72                         'first possible sequence 3 bytes' => array( $PASS, "\xe0\xa0\x80" ),
  73                         'first possible sequence 4 bytes' => array( $PASS, "\xf0\x90\x80\x80" ),
  74                         'first possible sequence 5 bytes' => array( $FAIL, "\xf8\x88\x80\x80\x80" ),
  75                         'first possible sequence 6 bytes' => array( $FAIL, "\xfc\x84\x80\x80\x80\x80" ),
  76
  77                         'last possible sequence 1 byte' => array( $PASS, "\x7f" ),
  78                         'last possible sequence 2 bytes' => array( $PASS, "\xdf\xbf" ),
  79                         'last possible sequence 3 bytes' => array( $PASS, "\xef\xbf\xbf" ),
  80                         'last possible sequence 4 bytes (U+1FFFFF)' => array( $FAIL, "\xf7\xbf\xbf\xbf" ),
  81                         'last possible sequence 5 bytes' => array( $FAIL, "\xfb\xbf\xbf\xbf\xbf" ),
  82                         'last possible sequence 6 bytes' => array( $FAIL, "\xfd\xbf\xbf\xbf\xbf\xbf" ),
  83
  84                         'boundary 1' => array( $PASS, "\xed\x9f\xbf" ),
  85                         'boundary 2' => array( $PASS, "\xee\x80\x80" ),
  86                         'boundary 3' => array( $PASS, "\xef\xbf\xbd" ),
  87                         'boundary 4' => array( $PASS, "\xf2\x80\x80\x80" ),
  88                         'boundary 5 (U+FFFFF)' => array( $PASS, "\xf3\xbf\xbf\xbf" ),
  89                         'boundary 6 (U+100000)' => array( $PASS, "\xf4\x80\x80\x80" ),
  90                         'boundary 7 (U+10FFFF)' => array( $PASS, "\xf4\x8f\xbf\xbf" ),
  91                         'boundary 8 (U+110000)' => array( $FAIL, "\xf4\x90\x80\x80" ),
  92
  93                         'malformed 1' => array( $FAIL, "\x80" ),
  94                         'malformed 2' => array( $FAIL, "\xbf" ),
  95                         'malformed 3' => array( $FAIL, "\x80\xbf" ),
  96                         'malformed 4' => array( $FAIL, "\x80\xbf\x80" ),
  97                         'malformed 5' => array( $FAIL, "\x80\xbf\x80\xbf" ),
  98                         'malformed 6' => array( $FAIL, "\x80\xbf\x80\xbf\x80" ),
  99                         'malformed 7' => array( $FAIL, "\x80\xbf\x80\xbf\x80\xbf" ),
 100                         'malformed 8' => array( $FAIL, "\x80\xbf\x80\xbf\x80\xbf\x80" ),
 101
 102                         'last byte missing 1' => array( $FAIL, "\xc0" ),
 103                         'last byte missing 2' => array( $FAIL, "\xe0\x80" ),
 104                         'last byte missing 3' => array( $FAIL, "\xf0\x80\x80" ),
 105                         'last byte missing 4' => array( $FAIL, "\xf8\x80\x80\x80" ),
 106                         'last byte missing 5' => array( $FAIL, "\xfc\x80\x80\x80\x80" ),
 107                         'last byte missing 6' => array( $FAIL, "\xdf" ),
 108                         'last byte missing 7' => array( $FAIL, "\xef\xbf" ),
 109                         'last byte missing 8' => array( $FAIL, "\xf7\xbf\xbf" ),
 110                         'last byte missing 9' => array( $FAIL, "\xfb\xbf\xbf\xbf" ),
 111                         'last byte missing 10' => array( $FAIL, "\xfd\xbf\xbf\xbf\xbf" ),
 112
 113                         'extra continuation byte 1' => array( $FAIL, "e\xaf" ),
 114                         'extra continuation byte 2' => array( $FAIL, "\xc3\x89\xaf" ),
 115                         'extra continuation byte 3' => array( $FAIL, "\xef\xbc\xa5\xaf" ),
 116                         'extra continuation byte 4' => array( $FAIL, "\xf0\x9d\x99\xb4\xaf" ),
 117
 118                         'impossible bytes 1' => array( $FAIL, "\xfe" ),
 119                         'impossible bytes 2' => array( $FAIL, "\xff" ),
 120                         'impossible bytes 3' => array( $FAIL, "\xfe\xfe\xff\xff" ),
 121
 122                         'overlong sequences 1' => array( $FAIL, "\xc0\xaf" ),
 123                         'overlong sequences 2' => array( $FAIL, "\xc1\xaf" ),
 124                         'overlong sequences 3' => array( $FAIL, "\xe0\x80\xaf" ),
 125                         'overlong sequences 4' => array( $FAIL, "\xf0\x80\x80\xaf" ),
 126                         'overlong sequences 5' => array( $FAIL, "\xf8\x80\x80\x80\xaf" ),
 127                         'overlong sequences 6' => array( $FAIL, "\xfc\x80\x80\x80\x80\xaf" ),
 128
 129                         'maximum overlong sequences 1' => array( $FAIL, "\xc1\xbf" ),
 130                         'maximum overlong sequences 2' => array( $FAIL, "\xe0\x9f\xbf" ),
 131                         'maximum overlong sequences 3' => array( $FAIL, "\xf0\x8f\xbf\xbf" ),
 132                         'maximum overlong sequences 4' => array( $FAIL, "\xf8\x87\xbf\xbf" ),
 133                         'maximum overlong sequences 5' => array( $FAIL, "\xfc\x83\xbf\xbf\xbf\xbf" ),
 134
 135                         'surrogates 1 (U+D799)' => array( $PASS, "\xed\x9f\xbf" ),
 136                         'surrogates 2 (U+E000)' => array( $PASS, "\xee\x80\x80" ),
 137                         'surrogates 3 (U+D800)' => array( $FAIL, "\xed\xa0\x80" ),
 138                         'surrogates 4 (U+DBFF)' => array( $FAIL, "\xed\xaf\xbf" ),
 139                         'surrogates 5 (U+DC00)' => array( $FAIL, "\xed\xb0\x80" ),
 140                         'surrogates 6 (U+DFFF)' => array( $FAIL, "\xed\xbf\xbf" ),
 141                         'surrogates 7 (U+D800 U+DC00)' => array( $FAIL, "\xed\xa0\x80\xed\xb0\x80" ),
 142
 143                         'noncharacters 1' => array( $PASS, "\xef\xbf\xbe" ),
 144                         'noncharacters 2' => array( $PASS, "\xef\xbf\xbf" ),
 145                 );
 146         }
 147 }