Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
81.82% |
18 / 22 |
|
33.33% |
1 / 3 |
CRAP | |
0.00% |
0 / 1 |
| HistoricalUIDGenerator | |
81.82% |
18 / 22 |
|
33.33% |
1 / 3 |
6.22 | |
0.00% |
0 / 1 |
| historicalTimestampedUID88 | |
92.86% |
13 / 14 |
|
0.00% |
0 / 1 |
3.00 | |||
| millisecondsSinceEpochBinary | |
57.14% |
4 / 7 |
|
0.00% |
0 / 1 |
2.31 | |||
| newNodeId | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace Flow\Import; |
| 4 | |
| 5 | use MWCryptRand; |
| 6 | use RuntimeException; |
| 7 | |
| 8 | /** |
| 9 | * Modified version of UIDGenerator generates historical timestamped |
| 10 | * uid's for use when importing older data. |
| 11 | * |
| 12 | * DO NOT USE for normal UID generation, this is likely to run into |
| 13 | * id collisions. |
| 14 | * |
| 15 | * The import process needs to identify collision failures reported by |
| 16 | * the database and re-try importing that item with another generated |
| 17 | * uid. |
| 18 | */ |
| 19 | class HistoricalUIDGenerator { |
| 20 | private const COUNTER_MAX = 1023; // 2^10 - 1 |
| 21 | |
| 22 | public static function historicalTimestampedUID88( $timestamp, $base = 10 ) { |
| 23 | static $counter = false; |
| 24 | if ( $counter === false ) { |
| 25 | $counter = mt_rand( 0, self::COUNTER_MAX ); |
| 26 | } |
| 27 | |
| 28 | // (seconds, milliseconds) |
| 29 | $time = [ |
| 30 | wfTimestamp( TS_UNIX, $timestamp ), |
| 31 | mt_rand( 0, 999 ) |
| 32 | ]; |
| 33 | ++$counter; |
| 34 | |
| 35 | // Take the 46 LSBs of "milliseconds since epoch" |
| 36 | $id_bin = self::millisecondsSinceEpochBinary( $time ); |
| 37 | // Add a 10 bit counter resulting in 56 bits total |
| 38 | $id_bin .= str_pad( decbin( $counter % ( self::COUNTER_MAX + 1 ) ), 10, '0', STR_PAD_LEFT ); |
| 39 | // Add the 32 bit node ID resulting in 88 bits total |
| 40 | $id_bin .= self::newNodeId(); |
| 41 | if ( strlen( $id_bin ) !== 88 ) { |
| 42 | throw new RuntimeException( "Detected overflow for millisecond timestamp." ); |
| 43 | } |
| 44 | |
| 45 | return \Wikimedia\base_convert( $id_bin, 2, $base ); |
| 46 | } |
| 47 | |
| 48 | /** |
| 49 | * @param array $time Array of second and millisecond integers |
| 50 | * @return string 46 LSBs of "milliseconds since epoch" in binary (rolls over in 4201) |
| 51 | * @throws RuntimeException |
| 52 | */ |
| 53 | protected static function millisecondsSinceEpochBinary( array $time ) { |
| 54 | [ $sec, $msec ] = $time; |
| 55 | $ts = 1000 * $sec + $msec; |
| 56 | if ( $ts > 2 ** 52 ) { |
| 57 | throw new RuntimeException( |
| 58 | __METHOD__ . ': sorry, this function doesn\'t work after the year 144680' |
| 59 | ); |
| 60 | } |
| 61 | |
| 62 | return substr( \Wikimedia\base_convert( (string)$ts, 10, 2, 46 ), -46 ); |
| 63 | } |
| 64 | |
| 65 | /** |
| 66 | * Rotate the nodeId to a random one. The stable node is best for |
| 67 | * generating "now" uid's on a cluster of servers, but repeated |
| 68 | * creation of historical uid's with one or a smaller number of |
| 69 | * machines requires use of a random node id. |
| 70 | * |
| 71 | * @return string String of 32 binary digits |
| 72 | */ |
| 73 | protected static function newNodeId() { |
| 74 | // 4 bytes = 32 bits |
| 75 | |
| 76 | return \Wikimedia\base_convert( MWCryptRand::generateHex( 8 ), 16, 2, 32 ); |
| 77 | } |
| 78 | } |