001/** 002 * Copyright The Apache Software Foundation 003 * 004 * Licensed to the Apache Software Foundation (ASF) under one 005 * or more contributor license agreements. See the NOTICE file 006 * distributed with this work for additional information 007 * regarding copyright ownership. The ASF licenses this file 008 * to you under the Apache License, Version 2.0 (the 009 * "License"); you may not use this file except in compliance 010 * with the License. You may obtain a copy of the License at 011 * 012 * http://www.apache.org/licenses/LICENSE-2.0 013 * 014 * Unless required by applicable law or agreed to in writing, software 015 * distributed under the License is distributed on an "AS IS" BASIS, 016 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 017 * See the License for the specific language governing permissions and 018 * limitations under the License. 019 */ 020package org.apache.hadoop.hbase; 021 022import static org.apache.hadoop.hbase.util.Bytes.len; 023 024import java.io.DataInput; 025import java.io.DataOutput; 026import java.io.IOException; 027import java.io.OutputStream; 028import java.nio.ByteBuffer; 029import java.util.ArrayList; 030import java.util.Arrays; 031import java.util.HashMap; 032import java.util.Iterator; 033import java.util.List; 034import java.util.Map; 035import org.apache.hadoop.hbase.util.ByteBufferUtils; 036import org.apache.hadoop.hbase.util.Bytes; 037import org.apache.hadoop.hbase.util.ClassSize; 038import org.apache.hadoop.io.RawComparator; 039import org.apache.yetus.audience.InterfaceAudience; 040import org.slf4j.Logger; 041import org.slf4j.LoggerFactory; 042 043import org.apache.hbase.thirdparty.com.google.common.annotations.VisibleForTesting; 044 045/** 046 * An HBase Key/Value. This is the fundamental HBase Type. 047 * <p> 048 * HBase applications and users should use the Cell interface and avoid directly using KeyValue and 049 * member functions not defined in Cell. 050 * <p> 051 * If being used client-side, the primary methods to access individual fields are 052 * {@link #getRowArray()}, {@link #getFamilyArray()}, {@link #getQualifierArray()}, 053 * {@link #getTimestamp()}, and {@link #getValueArray()}. These methods allocate new byte arrays 054 * and return copies. Avoid their use server-side. 055 * <p> 056 * Instances of this class are immutable. They do not implement Comparable but Comparators are 057 * provided. Comparators change with context, whether user table or a catalog table comparison. Its 058 * critical you use the appropriate comparator. There are Comparators for normal HFiles, Meta's 059 * Hfiles, and bloom filter keys. 060 * <p> 061 * KeyValue wraps a byte array and takes offsets and lengths into passed array at where to start 062 * interpreting the content as KeyValue. The KeyValue format inside a byte array is: 063 * <code><keylength> <valuelength> <key> <value></code> Key is further 064 * decomposed as: <code><rowlength> <row> <columnfamilylength> 065 * <columnfamily> <columnqualifier> 066 * <timestamp> <keytype></code> The <code>rowlength</code> maximum is 067 * <code>Short.MAX_SIZE</code>, column family length maximum is <code>Byte.MAX_SIZE</code>, and 068 * column qualifier + key length must be < <code>Integer.MAX_SIZE</code>. The column does not 069 * contain the family/qualifier delimiter, {@link #COLUMN_FAMILY_DELIMITER}<br> 070 * KeyValue can optionally contain Tags. When it contains tags, it is added in the byte array after 071 * the value part. The format for this part is: <code><tagslength><tagsbytes></code>. 072 * <code>tagslength</code> maximum is <code>Short.MAX_SIZE</code>. The <code>tagsbytes</code> 073 * contain one or more tags where as each tag is of the form 074 * <code><taglength><tagtype><tagbytes></code>. <code>tagtype</code> is one byte 075 * and <code>taglength</code> maximum is <code>Short.MAX_SIZE</code> and it includes 1 byte type 076 * length and actual tag bytes length. 077 */ 078@InterfaceAudience.Private 079public class KeyValue implements ExtendedCell, Cloneable { 080 private static final ArrayList<Tag> EMPTY_ARRAY_LIST = new ArrayList<>(); 081 082 private static final Logger LOG = LoggerFactory.getLogger(KeyValue.class); 083 084 public static final int FIXED_OVERHEAD = ClassSize.OBJECT + // the KeyValue object itself 085 ClassSize.REFERENCE + // pointer to "bytes" 086 2 * Bytes.SIZEOF_INT + // offset, length 087 Bytes.SIZEOF_LONG;// memstoreTS 088 089 /** 090 * Colon character in UTF-8 091 */ 092 public static final char COLUMN_FAMILY_DELIMITER = ':'; 093 094 public static final byte[] COLUMN_FAMILY_DELIM_ARRAY = 095 new byte[]{COLUMN_FAMILY_DELIMITER}; 096 097 /** 098 * Comparator for plain key/values; i.e. non-catalog table key/values. Works on Key portion 099 * of KeyValue only. 100 * @deprecated Use {@link CellComparator#getInstance()} instead. Deprecated for hbase 2.0, remove for hbase 3.0. 101 */ 102 @Deprecated 103 public static final KVComparator COMPARATOR = new KVComparator(); 104 /** 105 * A {@link KVComparator} for <code>hbase:meta</code> catalog table 106 * {@link KeyValue}s. 107 * @deprecated Use {@link CellComparatorImpl#META_COMPARATOR} instead. Deprecated for hbase 2.0, remove for hbase 3.0. 108 */ 109 @Deprecated 110 public static final KVComparator META_COMPARATOR = new MetaComparator(); 111 112 /** Size of the key length field in bytes*/ 113 public static final int KEY_LENGTH_SIZE = Bytes.SIZEOF_INT; 114 115 /** Size of the key type field in bytes */ 116 public static final int TYPE_SIZE = Bytes.SIZEOF_BYTE; 117 118 /** Size of the row length field in bytes */ 119 public static final int ROW_LENGTH_SIZE = Bytes.SIZEOF_SHORT; 120 121 /** Size of the family length field in bytes */ 122 public static final int FAMILY_LENGTH_SIZE = Bytes.SIZEOF_BYTE; 123 124 /** Size of the timestamp field in bytes */ 125 public static final int TIMESTAMP_SIZE = Bytes.SIZEOF_LONG; 126 127 // Size of the timestamp and type byte on end of a key -- a long + a byte. 128 public static final int TIMESTAMP_TYPE_SIZE = TIMESTAMP_SIZE + TYPE_SIZE; 129 130 // Size of the length shorts and bytes in key. 131 public static final int KEY_INFRASTRUCTURE_SIZE = ROW_LENGTH_SIZE 132 + FAMILY_LENGTH_SIZE + TIMESTAMP_TYPE_SIZE; 133 134 // How far into the key the row starts at. First thing to read is the short 135 // that says how long the row is. 136 public static final int ROW_OFFSET = 137 Bytes.SIZEOF_INT /*keylength*/ + 138 Bytes.SIZEOF_INT /*valuelength*/; 139 140 public static final int ROW_KEY_OFFSET = ROW_OFFSET + ROW_LENGTH_SIZE; 141 142 // Size of the length ints in a KeyValue datastructure. 143 public static final int KEYVALUE_INFRASTRUCTURE_SIZE = ROW_OFFSET; 144 145 /** Size of the tags length field in bytes */ 146 public static final int TAGS_LENGTH_SIZE = Bytes.SIZEOF_SHORT; 147 148 public static final int KEYVALUE_WITH_TAGS_INFRASTRUCTURE_SIZE = ROW_OFFSET + TAGS_LENGTH_SIZE; 149 150 /** 151 * Computes the number of bytes that a <code>KeyValue</code> instance with the provided 152 * characteristics would take up for its underlying data structure. 153 * 154 * @param rlength row length 155 * @param flength family length 156 * @param qlength qualifier length 157 * @param vlength value length 158 * 159 * @return the <code>KeyValue</code> data structure length 160 */ 161 public static long getKeyValueDataStructureSize(int rlength, 162 int flength, int qlength, int vlength) { 163 return KeyValue.KEYVALUE_INFRASTRUCTURE_SIZE 164 + getKeyDataStructureSize(rlength, flength, qlength) + vlength; 165 } 166 167 /** 168 * Computes the number of bytes that a <code>KeyValue</code> instance with the provided 169 * characteristics would take up for its underlying data structure. 170 * 171 * @param rlength row length 172 * @param flength family length 173 * @param qlength qualifier length 174 * @param vlength value length 175 * @param tagsLength total length of the tags 176 * 177 * @return the <code>KeyValue</code> data structure length 178 */ 179 public static long getKeyValueDataStructureSize(int rlength, int flength, int qlength, 180 int vlength, int tagsLength) { 181 if (tagsLength == 0) { 182 return getKeyValueDataStructureSize(rlength, flength, qlength, vlength); 183 } 184 return KeyValue.KEYVALUE_WITH_TAGS_INFRASTRUCTURE_SIZE 185 + getKeyDataStructureSize(rlength, flength, qlength) + vlength + tagsLength; 186 } 187 188 /** 189 * Computes the number of bytes that a <code>KeyValue</code> instance with the provided 190 * characteristics would take up for its underlying data structure. 191 * 192 * @param klength key length 193 * @param vlength value length 194 * @param tagsLength total length of the tags 195 * 196 * @return the <code>KeyValue</code> data structure length 197 */ 198 public static long getKeyValueDataStructureSize(int klength, int vlength, int tagsLength) { 199 if (tagsLength == 0) { 200 return (long) KeyValue.KEYVALUE_INFRASTRUCTURE_SIZE + klength + vlength; 201 } 202 return (long) KeyValue.KEYVALUE_WITH_TAGS_INFRASTRUCTURE_SIZE + klength + vlength + tagsLength; 203 } 204 205 /** 206 * Computes the number of bytes that a <code>KeyValue</code> instance with the provided 207 * characteristics would take up in its underlying data structure for the key. 208 * 209 * @param rlength row length 210 * @param flength family length 211 * @param qlength qualifier length 212 * 213 * @return the key data structure length 214 */ 215 public static long getKeyDataStructureSize(int rlength, int flength, int qlength) { 216 return (long) KeyValue.KEY_INFRASTRUCTURE_SIZE + rlength + flength + qlength; 217 } 218 219 /** 220 * Key type. 221 * Has space for other key types to be added later. Cannot rely on 222 * enum ordinals . They change if item is removed or moved. Do our own codes. 223 */ 224 public static enum Type { 225 Minimum((byte)0), 226 Put((byte)4), 227 228 Delete((byte)8), 229 DeleteFamilyVersion((byte)10), 230 DeleteColumn((byte)12), 231 DeleteFamily((byte)14), 232 233 // Maximum is used when searching; you look from maximum on down. 234 Maximum((byte)255); 235 236 private final byte code; 237 238 Type(final byte c) { 239 this.code = c; 240 } 241 242 public byte getCode() { 243 return this.code; 244 } 245 246 private static Type[] codeArray = new Type[256]; 247 248 static { 249 for (Type t : Type.values()) { 250 codeArray[t.code & 0xff] = t; 251 } 252 } 253 254 /** 255 * True to indicate that the byte b is a valid type. 256 * @param b byte to check 257 * @return true or false 258 */ 259 static boolean isValidType(byte b) { 260 return codeArray[b & 0xff] != null; 261 } 262 263 /** 264 * Cannot rely on enum ordinals . They change if item is removed or moved. 265 * Do our own codes. 266 * @param b 267 * @return Type associated with passed code. 268 */ 269 public static Type codeToType(final byte b) { 270 Type t = codeArray[b & 0xff]; 271 if (t != null) { 272 return t; 273 } 274 throw new RuntimeException("Unknown code " + b); 275 } 276 } 277 278 /** 279 * Lowest possible key. 280 * Makes a Key with highest possible Timestamp, empty row and column. No 281 * key can be equal or lower than this one in memstore or in store file. 282 */ 283 public static final KeyValue LOWESTKEY = 284 new KeyValue(HConstants.EMPTY_BYTE_ARRAY, HConstants.LATEST_TIMESTAMP); 285 286 //// 287 // KeyValue core instance fields. 288 protected byte [] bytes = null; // an immutable byte array that contains the KV 289 protected int offset = 0; // offset into bytes buffer KV starts at 290 protected int length = 0; // length of the KV starting from offset. 291 292 /** Here be dragons **/ 293 294 /** 295 * used to achieve atomic operations in the memstore. 296 */ 297 @Override 298 public long getSequenceId() { 299 return seqId; 300 } 301 302 @Override 303 public void setSequenceId(long seqId) { 304 this.seqId = seqId; 305 } 306 307 // multi-version concurrency control version. default value is 0, aka do not care. 308 private long seqId = 0; 309 310 /** Dragon time over, return to normal business */ 311 312 313 /** Writable Constructor -- DO NOT USE */ 314 public KeyValue() {} 315 316 /** 317 * Creates a KeyValue from the start of the specified byte array. 318 * Presumes <code>bytes</code> content is formatted as a KeyValue blob. 319 * @param bytes byte array 320 */ 321 public KeyValue(final byte [] bytes) { 322 this(bytes, 0); 323 } 324 325 /** 326 * Creates a KeyValue from the specified byte array and offset. 327 * Presumes <code>bytes</code> content starting at <code>offset</code> is 328 * formatted as a KeyValue blob. 329 * @param bytes byte array 330 * @param offset offset to start of KeyValue 331 */ 332 public KeyValue(final byte [] bytes, final int offset) { 333 this(bytes, offset, getLength(bytes, offset)); 334 } 335 336 /** 337 * Creates a KeyValue from the specified byte array, starting at offset, and 338 * for length <code>length</code>. 339 * @param bytes byte array 340 * @param offset offset to start of the KeyValue 341 * @param length length of the KeyValue 342 */ 343 public KeyValue(final byte[] bytes, final int offset, final int length) { 344 KeyValueUtil.checkKeyValueBytes(bytes, offset, length, true); 345 this.bytes = bytes; 346 this.offset = offset; 347 this.length = length; 348 } 349 350 /** 351 * Creates a KeyValue from the specified byte array, starting at offset, and 352 * for length <code>length</code>. 353 * 354 * @param bytes byte array 355 * @param offset offset to start of the KeyValue 356 * @param length length of the KeyValue 357 * @param ts 358 */ 359 public KeyValue(final byte[] bytes, final int offset, final int length, long ts) { 360 this(bytes, offset, length, null, 0, 0, null, 0, 0, ts, Type.Maximum, null, 0, 0, null); 361 } 362 363 /** Constructors that build a new backing byte array from fields */ 364 365 /** 366 * Constructs KeyValue structure filled with null value. 367 * Sets type to {@link KeyValue.Type#Maximum} 368 * @param row - row key (arbitrary byte array) 369 * @param timestamp 370 */ 371 public KeyValue(final byte [] row, final long timestamp) { 372 this(row, null, null, timestamp, Type.Maximum, null); 373 } 374 375 /** 376 * Constructs KeyValue structure filled with null value. 377 * @param row - row key (arbitrary byte array) 378 * @param timestamp 379 */ 380 public KeyValue(final byte [] row, final long timestamp, Type type) { 381 this(row, null, null, timestamp, type, null); 382 } 383 384 /** 385 * Constructs KeyValue structure filled with null value. 386 * Sets type to {@link KeyValue.Type#Maximum} 387 * @param row - row key (arbitrary byte array) 388 * @param family family name 389 * @param qualifier column qualifier 390 */ 391 public KeyValue(final byte [] row, final byte [] family, 392 final byte [] qualifier) { 393 this(row, family, qualifier, HConstants.LATEST_TIMESTAMP, Type.Maximum); 394 } 395 396 /** 397 * Constructs KeyValue structure as a put filled with specified values and 398 * LATEST_TIMESTAMP. 399 * @param row - row key (arbitrary byte array) 400 * @param family family name 401 * @param qualifier column qualifier 402 */ 403 public KeyValue(final byte [] row, final byte [] family, 404 final byte [] qualifier, final byte [] value) { 405 this(row, family, qualifier, HConstants.LATEST_TIMESTAMP, Type.Put, value); 406 } 407 408 /** 409 * Constructs KeyValue structure filled with specified values. 410 * @param row row key 411 * @param family family name 412 * @param qualifier column qualifier 413 * @param timestamp version timestamp 414 * @param type key type 415 * @throws IllegalArgumentException 416 */ 417 public KeyValue(final byte[] row, final byte[] family, 418 final byte[] qualifier, final long timestamp, Type type) { 419 this(row, family, qualifier, timestamp, type, null); 420 } 421 422 /** 423 * Constructs KeyValue structure filled with specified values. 424 * @param row row key 425 * @param family family name 426 * @param qualifier column qualifier 427 * @param timestamp version timestamp 428 * @param value column value 429 * @throws IllegalArgumentException 430 */ 431 public KeyValue(final byte[] row, final byte[] family, 432 final byte[] qualifier, final long timestamp, final byte[] value) { 433 this(row, family, qualifier, timestamp, Type.Put, value); 434 } 435 436 /** 437 * Constructs KeyValue structure filled with specified values. 438 * @param row row key 439 * @param family family name 440 * @param qualifier column qualifier 441 * @param timestamp version timestamp 442 * @param value column value 443 * @param tags tags 444 * @throws IllegalArgumentException 445 */ 446 public KeyValue(final byte[] row, final byte[] family, 447 final byte[] qualifier, final long timestamp, final byte[] value, 448 final Tag[] tags) { 449 this(row, family, qualifier, timestamp, value, tags != null ? Arrays.asList(tags) : null); 450 } 451 452 /** 453 * Constructs KeyValue structure filled with specified values. 454 * @param row row key 455 * @param family family name 456 * @param qualifier column qualifier 457 * @param timestamp version timestamp 458 * @param value column value 459 * @param tags tags non-empty list of tags or null 460 * @throws IllegalArgumentException 461 */ 462 public KeyValue(final byte[] row, final byte[] family, 463 final byte[] qualifier, final long timestamp, final byte[] value, 464 final List<Tag> tags) { 465 this(row, 0, row==null ? 0 : row.length, 466 family, 0, family==null ? 0 : family.length, 467 qualifier, 0, qualifier==null ? 0 : qualifier.length, 468 timestamp, Type.Put, 469 value, 0, value==null ? 0 : value.length, tags); 470 } 471 472 /** 473 * Constructs KeyValue structure filled with specified values. 474 * @param row row key 475 * @param family family name 476 * @param qualifier column qualifier 477 * @param timestamp version timestamp 478 * @param type key type 479 * @param value column value 480 * @throws IllegalArgumentException 481 */ 482 public KeyValue(final byte[] row, final byte[] family, 483 final byte[] qualifier, final long timestamp, Type type, 484 final byte[] value) { 485 this(row, 0, len(row), family, 0, len(family), qualifier, 0, len(qualifier), 486 timestamp, type, value, 0, len(value)); 487 } 488 489 /** 490 * Constructs KeyValue structure filled with specified values. 491 * <p> 492 * Column is split into two fields, family and qualifier. 493 * @param row row key 494 * @param family family name 495 * @param qualifier column qualifier 496 * @param timestamp version timestamp 497 * @param type key type 498 * @param value column value 499 * @throws IllegalArgumentException 500 */ 501 public KeyValue(final byte[] row, final byte[] family, 502 final byte[] qualifier, final long timestamp, Type type, 503 final byte[] value, final List<Tag> tags) { 504 this(row, family, qualifier, 0, qualifier==null ? 0 : qualifier.length, 505 timestamp, type, value, 0, value==null ? 0 : value.length, tags); 506 } 507 508 /** 509 * Constructs KeyValue structure filled with specified values. 510 * @param row row key 511 * @param family family name 512 * @param qualifier column qualifier 513 * @param timestamp version timestamp 514 * @param type key type 515 * @param value column value 516 * @throws IllegalArgumentException 517 */ 518 public KeyValue(final byte[] row, final byte[] family, 519 final byte[] qualifier, final long timestamp, Type type, 520 final byte[] value, final byte[] tags) { 521 this(row, family, qualifier, 0, qualifier==null ? 0 : qualifier.length, 522 timestamp, type, value, 0, value==null ? 0 : value.length, tags); 523 } 524 525 /** 526 * Constructs KeyValue structure filled with specified values. 527 * @param row row key 528 * @param family family name 529 * @param qualifier column qualifier 530 * @param qoffset qualifier offset 531 * @param qlength qualifier length 532 * @param timestamp version timestamp 533 * @param type key type 534 * @param value column value 535 * @param voffset value offset 536 * @param vlength value length 537 * @throws IllegalArgumentException 538 */ 539 public KeyValue(byte [] row, byte [] family, 540 byte [] qualifier, int qoffset, int qlength, long timestamp, Type type, 541 byte [] value, int voffset, int vlength, List<Tag> tags) { 542 this(row, 0, row==null ? 0 : row.length, 543 family, 0, family==null ? 0 : family.length, 544 qualifier, qoffset, qlength, timestamp, type, 545 value, voffset, vlength, tags); 546 } 547 548 /** 549 * @param row 550 * @param family 551 * @param qualifier 552 * @param qoffset 553 * @param qlength 554 * @param timestamp 555 * @param type 556 * @param value 557 * @param voffset 558 * @param vlength 559 * @param tags 560 */ 561 public KeyValue(byte [] row, byte [] family, 562 byte [] qualifier, int qoffset, int qlength, long timestamp, Type type, 563 byte [] value, int voffset, int vlength, byte[] tags) { 564 this(row, 0, row==null ? 0 : row.length, 565 family, 0, family==null ? 0 : family.length, 566 qualifier, qoffset, qlength, timestamp, type, 567 value, voffset, vlength, tags, 0, tags==null ? 0 : tags.length); 568 } 569 570 /** 571 * Constructs KeyValue structure filled with specified values. 572 * <p> 573 * Column is split into two fields, family and qualifier. 574 * @param row row key 575 * @throws IllegalArgumentException 576 */ 577 public KeyValue(final byte [] row, final int roffset, final int rlength, 578 final byte [] family, final int foffset, final int flength, 579 final byte [] qualifier, final int qoffset, final int qlength, 580 final long timestamp, final Type type, 581 final byte [] value, final int voffset, final int vlength) { 582 this(row, roffset, rlength, family, foffset, flength, qualifier, qoffset, 583 qlength, timestamp, type, value, voffset, vlength, null); 584 } 585 586 /** 587 * Constructs KeyValue structure filled with specified values. Uses the provided buffer as the 588 * data buffer. 589 * <p> 590 * Column is split into two fields, family and qualifier. 591 * 592 * @param buffer the bytes buffer to use 593 * @param boffset buffer offset 594 * @param row row key 595 * @param roffset row offset 596 * @param rlength row length 597 * @param family family name 598 * @param foffset family offset 599 * @param flength family length 600 * @param qualifier column qualifier 601 * @param qoffset qualifier offset 602 * @param qlength qualifier length 603 * @param timestamp version timestamp 604 * @param type key type 605 * @param value column value 606 * @param voffset value offset 607 * @param vlength value length 608 * @param tags non-empty list of tags or null 609 * @throws IllegalArgumentException an illegal value was passed or there is insufficient space 610 * remaining in the buffer 611 */ 612 public KeyValue(byte [] buffer, final int boffset, 613 final byte [] row, final int roffset, final int rlength, 614 final byte [] family, final int foffset, final int flength, 615 final byte [] qualifier, final int qoffset, final int qlength, 616 final long timestamp, final Type type, 617 final byte [] value, final int voffset, final int vlength, 618 final Tag[] tags) { 619 this.bytes = buffer; 620 this.length = writeByteArray(buffer, boffset, 621 row, roffset, rlength, 622 family, foffset, flength, qualifier, qoffset, qlength, 623 timestamp, type, value, voffset, vlength, tags); 624 this.offset = boffset; 625 } 626 627 /** 628 * Constructs KeyValue structure filled with specified values. 629 * <p> 630 * Column is split into two fields, family and qualifier. 631 * @param row row key 632 * @param roffset row offset 633 * @param rlength row length 634 * @param family family name 635 * @param foffset family offset 636 * @param flength family length 637 * @param qualifier column qualifier 638 * @param qoffset qualifier offset 639 * @param qlength qualifier length 640 * @param timestamp version timestamp 641 * @param type key type 642 * @param value column value 643 * @param voffset value offset 644 * @param vlength value length 645 * @param tags tags 646 * @throws IllegalArgumentException 647 */ 648 public KeyValue(final byte [] row, final int roffset, final int rlength, 649 final byte [] family, final int foffset, final int flength, 650 final byte [] qualifier, final int qoffset, final int qlength, 651 final long timestamp, final Type type, 652 final byte [] value, final int voffset, final int vlength, 653 final List<Tag> tags) { 654 this.bytes = createByteArray(row, roffset, rlength, 655 family, foffset, flength, qualifier, qoffset, qlength, 656 timestamp, type, value, voffset, vlength, tags); 657 this.length = bytes.length; 658 this.offset = 0; 659 } 660 661 /** 662 * @param row 663 * @param roffset 664 * @param rlength 665 * @param family 666 * @param foffset 667 * @param flength 668 * @param qualifier 669 * @param qoffset 670 * @param qlength 671 * @param timestamp 672 * @param type 673 * @param value 674 * @param voffset 675 * @param vlength 676 * @param tags 677 */ 678 public KeyValue(final byte [] row, final int roffset, final int rlength, 679 final byte [] family, final int foffset, final int flength, 680 final byte [] qualifier, final int qoffset, final int qlength, 681 final long timestamp, final Type type, 682 final byte [] value, final int voffset, final int vlength, 683 final byte[] tags, final int tagsOffset, final int tagsLength) { 684 this.bytes = createByteArray(row, roffset, rlength, 685 family, foffset, flength, qualifier, qoffset, qlength, 686 timestamp, type, value, voffset, vlength, tags, tagsOffset, tagsLength); 687 this.length = bytes.length; 688 this.offset = 0; 689 } 690 691 /** 692 * Constructs an empty KeyValue structure, with specified sizes. 693 * This can be used to partially fill up KeyValues. 694 * <p> 695 * Column is split into two fields, family and qualifier. 696 * @param rlength row length 697 * @param flength family length 698 * @param qlength qualifier length 699 * @param timestamp version timestamp 700 * @param type key type 701 * @param vlength value length 702 * @throws IllegalArgumentException 703 */ 704 public KeyValue(final int rlength, 705 final int flength, 706 final int qlength, 707 final long timestamp, final Type type, 708 final int vlength) { 709 this(rlength, flength, qlength, timestamp, type, vlength, 0); 710 } 711 712 /** 713 * Constructs an empty KeyValue structure, with specified sizes. 714 * This can be used to partially fill up KeyValues. 715 * <p> 716 * Column is split into two fields, family and qualifier. 717 * @param rlength row length 718 * @param flength family length 719 * @param qlength qualifier length 720 * @param timestamp version timestamp 721 * @param type key type 722 * @param vlength value length 723 * @param tagsLength 724 * @throws IllegalArgumentException 725 */ 726 public KeyValue(final int rlength, 727 final int flength, 728 final int qlength, 729 final long timestamp, final Type type, 730 final int vlength, final int tagsLength) { 731 this.bytes = createEmptyByteArray(rlength, flength, qlength, timestamp, type, vlength, 732 tagsLength); 733 this.length = bytes.length; 734 this.offset = 0; 735 } 736 737 738 public KeyValue(byte[] row, int roffset, int rlength, 739 byte[] family, int foffset, int flength, 740 ByteBuffer qualifier, long ts, Type type, ByteBuffer value, List<Tag> tags) { 741 this.bytes = createByteArray(row, roffset, rlength, family, foffset, flength, 742 qualifier, 0, qualifier == null ? 0 : qualifier.remaining(), ts, type, 743 value, 0, value == null ? 0 : value.remaining(), tags); 744 this.length = bytes.length; 745 this.offset = 0; 746 } 747 748 public KeyValue(Cell c) { 749 this(c.getRowArray(), c.getRowOffset(), c.getRowLength(), 750 c.getFamilyArray(), c.getFamilyOffset(), c.getFamilyLength(), 751 c.getQualifierArray(), c.getQualifierOffset(), c.getQualifierLength(), 752 c.getTimestamp(), Type.codeToType(c.getTypeByte()), c.getValueArray(), c.getValueOffset(), 753 c.getValueLength(), c.getTagsArray(), c.getTagsOffset(), c.getTagsLength()); 754 this.seqId = c.getSequenceId(); 755 } 756 757 /** 758 * Create an empty byte[] representing a KeyValue 759 * All lengths are preset and can be filled in later. 760 * @param rlength 761 * @param flength 762 * @param qlength 763 * @param timestamp 764 * @param type 765 * @param vlength 766 * @return The newly created byte array. 767 */ 768 private static byte[] createEmptyByteArray(final int rlength, int flength, 769 int qlength, final long timestamp, final Type type, int vlength, int tagsLength) { 770 if (rlength > Short.MAX_VALUE) { 771 throw new IllegalArgumentException("Row > " + Short.MAX_VALUE); 772 } 773 if (flength > Byte.MAX_VALUE) { 774 throw new IllegalArgumentException("Family > " + Byte.MAX_VALUE); 775 } 776 // Qualifier length 777 if (qlength > Integer.MAX_VALUE - rlength - flength) { 778 throw new IllegalArgumentException("Qualifier > " + Integer.MAX_VALUE); 779 } 780 RawCell.checkForTagsLength(tagsLength); 781 // Key length 782 long longkeylength = getKeyDataStructureSize(rlength, flength, qlength); 783 if (longkeylength > Integer.MAX_VALUE) { 784 throw new IllegalArgumentException("keylength " + longkeylength + " > " + 785 Integer.MAX_VALUE); 786 } 787 int keylength = (int)longkeylength; 788 // Value length 789 if (vlength > HConstants.MAXIMUM_VALUE_LENGTH) { // FindBugs INT_VACUOUS_COMPARISON 790 throw new IllegalArgumentException("Valuer > " + 791 HConstants.MAXIMUM_VALUE_LENGTH); 792 } 793 794 // Allocate right-sized byte array. 795 byte[] bytes= new byte[(int) getKeyValueDataStructureSize(rlength, flength, qlength, vlength, 796 tagsLength)]; 797 // Write the correct size markers 798 int pos = 0; 799 pos = Bytes.putInt(bytes, pos, keylength); 800 pos = Bytes.putInt(bytes, pos, vlength); 801 pos = Bytes.putShort(bytes, pos, (short)(rlength & 0x0000ffff)); 802 pos += rlength; 803 pos = Bytes.putByte(bytes, pos, (byte)(flength & 0x0000ff)); 804 pos += flength + qlength; 805 pos = Bytes.putLong(bytes, pos, timestamp); 806 pos = Bytes.putByte(bytes, pos, type.getCode()); 807 pos += vlength; 808 if (tagsLength > 0) { 809 pos = Bytes.putAsShort(bytes, pos, tagsLength); 810 } 811 return bytes; 812 } 813 814 /** 815 * Checks the parameters passed to a constructor. 816 * 817 * @param row row key 818 * @param rlength row length 819 * @param family family name 820 * @param flength family length 821 * @param qlength qualifier length 822 * @param vlength value length 823 * 824 * @throws IllegalArgumentException an illegal value was passed 825 */ 826 static void checkParameters(final byte [] row, final int rlength, 827 final byte [] family, int flength, int qlength, int vlength) 828 throws IllegalArgumentException { 829 if (rlength > Short.MAX_VALUE) { 830 throw new IllegalArgumentException("Row > " + Short.MAX_VALUE); 831 } 832 if (row == null) { 833 throw new IllegalArgumentException("Row is null"); 834 } 835 // Family length 836 flength = family == null ? 0 : flength; 837 if (flength > Byte.MAX_VALUE) { 838 throw new IllegalArgumentException("Family > " + Byte.MAX_VALUE); 839 } 840 // Qualifier length 841 if (qlength > Integer.MAX_VALUE - rlength - flength) { 842 throw new IllegalArgumentException("Qualifier > " + Integer.MAX_VALUE); 843 } 844 // Key length 845 long longKeyLength = getKeyDataStructureSize(rlength, flength, qlength); 846 if (longKeyLength > Integer.MAX_VALUE) { 847 throw new IllegalArgumentException("keylength " + longKeyLength + " > " + 848 Integer.MAX_VALUE); 849 } 850 // Value length 851 if (vlength > HConstants.MAXIMUM_VALUE_LENGTH) { // FindBugs INT_VACUOUS_COMPARISON 852 throw new IllegalArgumentException("Value length " + vlength + " > " + 853 HConstants.MAXIMUM_VALUE_LENGTH); 854 } 855 } 856 857 /** 858 * Write KeyValue format into the provided byte array. 859 * 860 * @param buffer the bytes buffer to use 861 * @param boffset buffer offset 862 * @param row row key 863 * @param roffset row offset 864 * @param rlength row length 865 * @param family family name 866 * @param foffset family offset 867 * @param flength family length 868 * @param qualifier column qualifier 869 * @param qoffset qualifier offset 870 * @param qlength qualifier length 871 * @param timestamp version timestamp 872 * @param type key type 873 * @param value column value 874 * @param voffset value offset 875 * @param vlength value length 876 * 877 * @return The number of useful bytes in the buffer. 878 * 879 * @throws IllegalArgumentException an illegal value was passed or there is insufficient space 880 * remaining in the buffer 881 */ 882 public static int writeByteArray(byte [] buffer, final int boffset, 883 final byte [] row, final int roffset, final int rlength, 884 final byte [] family, final int foffset, int flength, 885 final byte [] qualifier, final int qoffset, int qlength, 886 final long timestamp, final Type type, 887 final byte [] value, final int voffset, int vlength, Tag[] tags) { 888 889 checkParameters(row, rlength, family, flength, qlength, vlength); 890 891 // Calculate length of tags area 892 int tagsLength = 0; 893 if (tags != null && tags.length > 0) { 894 for (Tag t: tags) { 895 tagsLength += t.getValueLength() + Tag.INFRASTRUCTURE_SIZE; 896 } 897 } 898 RawCell.checkForTagsLength(tagsLength); 899 int keyLength = (int) getKeyDataStructureSize(rlength, flength, qlength); 900 int keyValueLength = (int) getKeyValueDataStructureSize(rlength, flength, qlength, vlength, 901 tagsLength); 902 if (keyValueLength > buffer.length - boffset) { 903 throw new IllegalArgumentException("Buffer size " + (buffer.length - boffset) + " < " + 904 keyValueLength); 905 } 906 907 // Write key, value and key row length. 908 int pos = boffset; 909 pos = Bytes.putInt(buffer, pos, keyLength); 910 pos = Bytes.putInt(buffer, pos, vlength); 911 pos = Bytes.putShort(buffer, pos, (short)(rlength & 0x0000ffff)); 912 pos = Bytes.putBytes(buffer, pos, row, roffset, rlength); 913 pos = Bytes.putByte(buffer, pos, (byte) (flength & 0x0000ff)); 914 if (flength != 0) { 915 pos = Bytes.putBytes(buffer, pos, family, foffset, flength); 916 } 917 if (qlength != 0) { 918 pos = Bytes.putBytes(buffer, pos, qualifier, qoffset, qlength); 919 } 920 pos = Bytes.putLong(buffer, pos, timestamp); 921 pos = Bytes.putByte(buffer, pos, type.getCode()); 922 if (value != null && value.length > 0) { 923 pos = Bytes.putBytes(buffer, pos, value, voffset, vlength); 924 } 925 // Write the number of tags. If it is 0 then it means there are no tags. 926 if (tagsLength > 0) { 927 pos = Bytes.putAsShort(buffer, pos, tagsLength); 928 for (Tag t : tags) { 929 int tlen = t.getValueLength(); 930 pos = Bytes.putAsShort(buffer, pos, tlen + Tag.TYPE_LENGTH_SIZE); 931 pos = Bytes.putByte(buffer, pos, t.getType()); 932 Tag.copyValueTo(t, buffer, pos); 933 pos += tlen; 934 } 935 } 936 return keyValueLength; 937 } 938 939 /** 940 * Write KeyValue format into a byte array. 941 * @param row row key 942 * @param roffset row offset 943 * @param rlength row length 944 * @param family family name 945 * @param foffset family offset 946 * @param flength family length 947 * @param qualifier column qualifier 948 * @param qoffset qualifier offset 949 * @param qlength qualifier length 950 * @param timestamp version timestamp 951 * @param type key type 952 * @param value column value 953 * @param voffset value offset 954 * @param vlength value length 955 * @return The newly created byte array. 956 */ 957 private static byte [] createByteArray(final byte [] row, final int roffset, 958 final int rlength, final byte [] family, final int foffset, int flength, 959 final byte [] qualifier, final int qoffset, int qlength, 960 final long timestamp, final Type type, 961 final byte [] value, final int voffset, 962 int vlength, byte[] tags, int tagsOffset, int tagsLength) { 963 964 checkParameters(row, rlength, family, flength, qlength, vlength); 965 RawCell.checkForTagsLength(tagsLength); 966 // Allocate right-sized byte array. 967 int keyLength = (int) getKeyDataStructureSize(rlength, flength, qlength); 968 byte[] bytes = new byte[(int) getKeyValueDataStructureSize(rlength, flength, qlength, vlength, 969 tagsLength)]; 970 // Write key, value and key row length. 971 int pos = 0; 972 pos = Bytes.putInt(bytes, pos, keyLength); 973 pos = Bytes.putInt(bytes, pos, vlength); 974 pos = Bytes.putShort(bytes, pos, (short)(rlength & 0x0000ffff)); 975 pos = Bytes.putBytes(bytes, pos, row, roffset, rlength); 976 pos = Bytes.putByte(bytes, pos, (byte)(flength & 0x0000ff)); 977 if(flength != 0) { 978 pos = Bytes.putBytes(bytes, pos, family, foffset, flength); 979 } 980 if(qlength != 0) { 981 pos = Bytes.putBytes(bytes, pos, qualifier, qoffset, qlength); 982 } 983 pos = Bytes.putLong(bytes, pos, timestamp); 984 pos = Bytes.putByte(bytes, pos, type.getCode()); 985 if (value != null && value.length > 0) { 986 pos = Bytes.putBytes(bytes, pos, value, voffset, vlength); 987 } 988 // Add the tags after the value part 989 if (tagsLength > 0) { 990 pos = Bytes.putAsShort(bytes, pos, tagsLength); 991 pos = Bytes.putBytes(bytes, pos, tags, tagsOffset, tagsLength); 992 } 993 return bytes; 994 } 995 996 /** 997 * @param qualifier can be a ByteBuffer or a byte[], or null. 998 * @param value can be a ByteBuffer or a byte[], or null. 999 */ 1000 private static byte [] createByteArray(final byte [] row, final int roffset, 1001 final int rlength, final byte [] family, final int foffset, int flength, 1002 final Object qualifier, final int qoffset, int qlength, 1003 final long timestamp, final Type type, 1004 final Object value, final int voffset, int vlength, List<Tag> tags) { 1005 1006 checkParameters(row, rlength, family, flength, qlength, vlength); 1007 1008 // Calculate length of tags area 1009 int tagsLength = 0; 1010 if (tags != null && !tags.isEmpty()) { 1011 for (Tag t : tags) { 1012 tagsLength += t.getValueLength() + Tag.INFRASTRUCTURE_SIZE; 1013 } 1014 } 1015 RawCell.checkForTagsLength(tagsLength); 1016 // Allocate right-sized byte array. 1017 int keyLength = (int) getKeyDataStructureSize(rlength, flength, qlength); 1018 byte[] bytes = new byte[(int) getKeyValueDataStructureSize(rlength, flength, qlength, vlength, 1019 tagsLength)]; 1020 1021 // Write key, value and key row length. 1022 int pos = 0; 1023 pos = Bytes.putInt(bytes, pos, keyLength); 1024 1025 pos = Bytes.putInt(bytes, pos, vlength); 1026 pos = Bytes.putShort(bytes, pos, (short)(rlength & 0x0000ffff)); 1027 pos = Bytes.putBytes(bytes, pos, row, roffset, rlength); 1028 pos = Bytes.putByte(bytes, pos, (byte)(flength & 0x0000ff)); 1029 if(flength != 0) { 1030 pos = Bytes.putBytes(bytes, pos, family, foffset, flength); 1031 } 1032 if (qlength > 0) { 1033 if (qualifier instanceof ByteBuffer) { 1034 pos = Bytes.putByteBuffer(bytes, pos, (ByteBuffer) qualifier); 1035 } else { 1036 pos = Bytes.putBytes(bytes, pos, (byte[]) qualifier, qoffset, qlength); 1037 } 1038 } 1039 pos = Bytes.putLong(bytes, pos, timestamp); 1040 pos = Bytes.putByte(bytes, pos, type.getCode()); 1041 if (vlength > 0) { 1042 if (value instanceof ByteBuffer) { 1043 pos = Bytes.putByteBuffer(bytes, pos, (ByteBuffer) value); 1044 } else { 1045 pos = Bytes.putBytes(bytes, pos, (byte[]) value, voffset, vlength); 1046 } 1047 } 1048 // Add the tags after the value part 1049 if (tagsLength > 0) { 1050 pos = Bytes.putAsShort(bytes, pos, tagsLength); 1051 for (Tag t : tags) { 1052 int tlen = t.getValueLength(); 1053 pos = Bytes.putAsShort(bytes, pos, tlen + Tag.TYPE_LENGTH_SIZE); 1054 pos = Bytes.putByte(bytes, pos, t.getType()); 1055 Tag.copyValueTo(t, bytes, pos); 1056 pos += tlen; 1057 } 1058 } 1059 return bytes; 1060 } 1061 1062 /** 1063 * Needed doing 'contains' on List. Only compares the key portion, not the value. 1064 */ 1065 @Override 1066 public boolean equals(Object other) { 1067 if (!(other instanceof Cell)) { 1068 return false; 1069 } 1070 return CellUtil.equals(this, (Cell)other); 1071 } 1072 1073 /** 1074 * In line with {@link #equals(Object)}, only uses the key portion, not the value. 1075 */ 1076 @Override 1077 public int hashCode() { 1078 return calculateHashForKey(this); 1079 } 1080 1081 private int calculateHashForKey(Cell cell) { 1082 // pre-calculate the 3 hashes made of byte ranges 1083 int rowHash = Bytes.hashCode(cell.getRowArray(), cell.getRowOffset(), cell.getRowLength()); 1084 int familyHash = Bytes.hashCode(cell.getFamilyArray(), cell.getFamilyOffset(), 1085 cell.getFamilyLength()); 1086 int qualifierHash = Bytes.hashCode(cell.getQualifierArray(), cell.getQualifierOffset(), 1087 cell.getQualifierLength()); 1088 1089 // combine the 6 sub-hashes 1090 int hash = 31 * rowHash + familyHash; 1091 hash = 31 * hash + qualifierHash; 1092 hash = 31 * hash + (int) cell.getTimestamp(); 1093 hash = 31 * hash + cell.getTypeByte(); 1094 return hash; 1095 } 1096 1097 //--------------------------------------------------------------------------- 1098 // 1099 // KeyValue cloning 1100 // 1101 //--------------------------------------------------------------------------- 1102 1103 /** 1104 * Clones a KeyValue. This creates a copy, re-allocating the buffer. 1105 * @return Fully copied clone of this KeyValue 1106 * @throws CloneNotSupportedException 1107 */ 1108 @Override 1109 public KeyValue clone() throws CloneNotSupportedException { 1110 super.clone(); 1111 byte [] b = new byte[this.length]; 1112 System.arraycopy(this.bytes, this.offset, b, 0, this.length); 1113 KeyValue ret = new KeyValue(b, 0, b.length); 1114 // Important to clone the memstoreTS as well - otherwise memstore's 1115 // update-in-place methods (eg increment) will end up creating 1116 // new entries 1117 ret.setSequenceId(seqId); 1118 return ret; 1119 } 1120 1121 /** 1122 * Creates a shallow copy of this KeyValue, reusing the data byte buffer. 1123 * http://en.wikipedia.org/wiki/Object_copy 1124 * @return Shallow copy of this KeyValue 1125 */ 1126 public KeyValue shallowCopy() { 1127 KeyValue shallowCopy = new KeyValue(this.bytes, this.offset, this.length); 1128 shallowCopy.setSequenceId(this.seqId); 1129 return shallowCopy; 1130 } 1131 1132 //--------------------------------------------------------------------------- 1133 // 1134 // String representation 1135 // 1136 //--------------------------------------------------------------------------- 1137 1138 @Override 1139 public String toString() { 1140 if (this.bytes == null || this.bytes.length == 0) { 1141 return "empty"; 1142 } 1143 return keyToString(this.bytes, this.offset + ROW_OFFSET, getKeyLength()) + "/vlen=" 1144 + getValueLength() + "/seqid=" + seqId; 1145 } 1146 1147 /** 1148 * @param k Key portion of a KeyValue. 1149 * @return Key as a String, empty string if k is null. 1150 */ 1151 public static String keyToString(final byte [] k) { 1152 if (k == null) { 1153 return ""; 1154 } 1155 return keyToString(k, 0, k.length); 1156 } 1157 1158 /** 1159 * Produces a string map for this key/value pair. Useful for programmatic use 1160 * and manipulation of the data stored in an WALKey, for example, printing 1161 * as JSON. Values are left out due to their tendency to be large. If needed, 1162 * they can be added manually. 1163 * 1164 * @return the Map<String,?> containing data from this key 1165 */ 1166 public Map<String, Object> toStringMap() { 1167 Map<String, Object> stringMap = new HashMap<>(); 1168 stringMap.put("row", Bytes.toStringBinary(getRowArray(), getRowOffset(), getRowLength())); 1169 stringMap.put("family", 1170 Bytes.toStringBinary(getFamilyArray(), getFamilyOffset(), getFamilyLength())); 1171 stringMap.put("qualifier", 1172 Bytes.toStringBinary(getQualifierArray(), getQualifierOffset(), getQualifierLength())); 1173 stringMap.put("timestamp", getTimestamp()); 1174 stringMap.put("vlen", getValueLength()); 1175 Iterator<Tag> tags = getTags(); 1176 if (tags != null) { 1177 List<String> tagsString = new ArrayList<String>(); 1178 while (tags.hasNext()) { 1179 tagsString.add(tags.next().toString()); 1180 } 1181 stringMap.put("tag", tagsString); 1182 } 1183 return stringMap; 1184 } 1185 1186 /** 1187 * Use for logging. 1188 * @param b Key portion of a KeyValue. 1189 * @param o Offset to start of key 1190 * @param l Length of key. 1191 * @return Key as a String. 1192 */ 1193 public static String keyToString(final byte [] b, final int o, final int l) { 1194 if (b == null) return ""; 1195 int rowlength = Bytes.toShort(b, o); 1196 String row = Bytes.toStringBinary(b, o + Bytes.SIZEOF_SHORT, rowlength); 1197 int columnoffset = o + Bytes.SIZEOF_SHORT + 1 + rowlength; 1198 int familylength = b[columnoffset - 1]; 1199 int columnlength = l - ((columnoffset - o) + TIMESTAMP_TYPE_SIZE); 1200 String family = familylength == 0? "": 1201 Bytes.toStringBinary(b, columnoffset, familylength); 1202 String qualifier = columnlength == 0? "": 1203 Bytes.toStringBinary(b, columnoffset + familylength, 1204 columnlength - familylength); 1205 long timestamp = Bytes.toLong(b, o + (l - TIMESTAMP_TYPE_SIZE)); 1206 String timestampStr = humanReadableTimestamp(timestamp); 1207 byte type = b[o + l - 1]; 1208 return row + "/" + family + 1209 (family != null && family.length() > 0? ":" :"") + 1210 qualifier + "/" + timestampStr + "/" + Type.codeToType(type); 1211 } 1212 1213 public static String humanReadableTimestamp(final long timestamp) { 1214 if (timestamp == HConstants.LATEST_TIMESTAMP) { 1215 return "LATEST_TIMESTAMP"; 1216 } 1217 if (timestamp == HConstants.OLDEST_TIMESTAMP) { 1218 return "OLDEST_TIMESTAMP"; 1219 } 1220 return String.valueOf(timestamp); 1221 } 1222 1223 //--------------------------------------------------------------------------- 1224 // 1225 // Public Member Accessors 1226 // 1227 //--------------------------------------------------------------------------- 1228 1229 /** 1230 * To be used only in tests where the Cells are clearly assumed to be of type KeyValue 1231 * and that we need access to the backing array to do some test case related assertions. 1232 * @return The byte array backing this KeyValue. 1233 */ 1234 @VisibleForTesting 1235 public byte [] getBuffer() { 1236 return this.bytes; 1237 } 1238 1239 /** 1240 * @return Offset into {@link #getBuffer()} at which this KeyValue starts. 1241 */ 1242 public int getOffset() { 1243 return this.offset; 1244 } 1245 1246 /** 1247 * @return Length of bytes this KeyValue occupies in {@link #getBuffer()}. 1248 */ 1249 public int getLength() { 1250 return length; 1251 } 1252 1253 //--------------------------------------------------------------------------- 1254 // 1255 // Length and Offset Calculators 1256 // 1257 //--------------------------------------------------------------------------- 1258 1259 /** 1260 * Determines the total length of the KeyValue stored in the specified 1261 * byte array and offset. Includes all headers. 1262 * @param bytes byte array 1263 * @param offset offset to start of the KeyValue 1264 * @return length of entire KeyValue, in bytes 1265 */ 1266 private static int getLength(byte [] bytes, int offset) { 1267 int klength = ROW_OFFSET + Bytes.toInt(bytes, offset); 1268 int vlength = Bytes.toInt(bytes, offset + Bytes.SIZEOF_INT); 1269 return klength + vlength; 1270 } 1271 1272 /** 1273 * @return Key offset in backing buffer.. 1274 */ 1275 public int getKeyOffset() { 1276 return this.offset + ROW_OFFSET; 1277 } 1278 1279 public String getKeyString() { 1280 return Bytes.toStringBinary(getBuffer(), getKeyOffset(), getKeyLength()); 1281 } 1282 1283 /** 1284 * @return Length of key portion. 1285 */ 1286 public int getKeyLength() { 1287 return Bytes.toInt(this.bytes, this.offset); 1288 } 1289 1290 /** 1291 * @return the backing array of the entire KeyValue (all KeyValue fields are in a single array) 1292 */ 1293 @Override 1294 public byte[] getValueArray() { 1295 return bytes; 1296 } 1297 1298 /** 1299 * @return the value offset 1300 */ 1301 @Override 1302 public int getValueOffset() { 1303 int voffset = getKeyOffset() + getKeyLength(); 1304 return voffset; 1305 } 1306 1307 /** 1308 * @return Value length 1309 */ 1310 @Override 1311 public int getValueLength() { 1312 int vlength = Bytes.toInt(this.bytes, this.offset + Bytes.SIZEOF_INT); 1313 return vlength; 1314 } 1315 1316 /** 1317 * @return the backing array of the entire KeyValue (all KeyValue fields are in a single array) 1318 */ 1319 @Override 1320 public byte[] getRowArray() { 1321 return bytes; 1322 } 1323 1324 /** 1325 * @return Row offset 1326 */ 1327 @Override 1328 public int getRowOffset() { 1329 return this.offset + ROW_KEY_OFFSET; 1330 } 1331 1332 /** 1333 * @return Row length 1334 */ 1335 @Override 1336 public short getRowLength() { 1337 return Bytes.toShort(this.bytes, getKeyOffset()); 1338 } 1339 1340 /** 1341 * @return the backing array of the entire KeyValue (all KeyValue fields are in a single array) 1342 */ 1343 @Override 1344 public byte[] getFamilyArray() { 1345 return bytes; 1346 } 1347 1348 /** 1349 * @return Family offset 1350 */ 1351 @Override 1352 public int getFamilyOffset() { 1353 return getFamilyOffset(getRowLength()); 1354 } 1355 1356 /** 1357 * @return Family offset 1358 */ 1359 private int getFamilyOffset(int rlength) { 1360 return this.offset + ROW_KEY_OFFSET + rlength + Bytes.SIZEOF_BYTE; 1361 } 1362 1363 /** 1364 * @return Family length 1365 */ 1366 @Override 1367 public byte getFamilyLength() { 1368 return getFamilyLength(getFamilyOffset()); 1369 } 1370 1371 /** 1372 * @return Family length 1373 */ 1374 public byte getFamilyLength(int foffset) { 1375 return this.bytes[foffset-1]; 1376 } 1377 1378 /** 1379 * @return the backing array of the entire KeyValue (all KeyValue fields are in a single array) 1380 */ 1381 @Override 1382 public byte[] getQualifierArray() { 1383 return bytes; 1384 } 1385 1386 /** 1387 * @return Qualifier offset 1388 */ 1389 @Override 1390 public int getQualifierOffset() { 1391 return getQualifierOffset(getFamilyOffset()); 1392 } 1393 1394 /** 1395 * @return Qualifier offset 1396 */ 1397 private int getQualifierOffset(int foffset) { 1398 return foffset + getFamilyLength(foffset); 1399 } 1400 1401 /** 1402 * @return Qualifier length 1403 */ 1404 @Override 1405 public int getQualifierLength() { 1406 return getQualifierLength(getRowLength(),getFamilyLength()); 1407 } 1408 1409 /** 1410 * @return Qualifier length 1411 */ 1412 private int getQualifierLength(int rlength, int flength) { 1413 return getKeyLength() - (int) getKeyDataStructureSize(rlength, flength, 0); 1414 } 1415 1416 /** 1417 * @return Timestamp offset 1418 */ 1419 public int getTimestampOffset() { 1420 return getTimestampOffset(getKeyLength()); 1421 } 1422 1423 /** 1424 * @param keylength Pass if you have it to save on a int creation. 1425 * @return Timestamp offset 1426 */ 1427 private int getTimestampOffset(final int keylength) { 1428 return getKeyOffset() + keylength - TIMESTAMP_TYPE_SIZE; 1429 } 1430 1431 /** 1432 * @return True if this KeyValue has a LATEST_TIMESTAMP timestamp. 1433 */ 1434 public boolean isLatestTimestamp() { 1435 return Bytes.equals(getBuffer(), getTimestampOffset(), Bytes.SIZEOF_LONG, 1436 HConstants.LATEST_TIMESTAMP_BYTES, 0, Bytes.SIZEOF_LONG); 1437 } 1438 1439 /** 1440 * @param now Time to set into <code>this</code> IFF timestamp == 1441 * {@link HConstants#LATEST_TIMESTAMP} (else, its a noop). 1442 * @return True is we modified this. 1443 */ 1444 public boolean updateLatestStamp(final byte [] now) { 1445 if (this.isLatestTimestamp()) { 1446 int tsOffset = getTimestampOffset(); 1447 System.arraycopy(now, 0, this.bytes, tsOffset, Bytes.SIZEOF_LONG); 1448 // clear cache or else getTimestamp() possibly returns an old value 1449 return true; 1450 } 1451 return false; 1452 } 1453 1454 @Override 1455 public void setTimestamp(long ts) { 1456 Bytes.putBytes(this.bytes, this.getTimestampOffset(), Bytes.toBytes(ts), 0, Bytes.SIZEOF_LONG); 1457 } 1458 1459 @Override 1460 public void setTimestamp(byte[] ts) { 1461 Bytes.putBytes(this.bytes, this.getTimestampOffset(), ts, 0, Bytes.SIZEOF_LONG); 1462 } 1463 1464 //--------------------------------------------------------------------------- 1465 // 1466 // Methods that return copies of fields 1467 // 1468 //--------------------------------------------------------------------------- 1469 1470 /** 1471 * Do not use unless you have to. Used internally for compacting and testing. Use 1472 * {@link #getRowArray()}, {@link #getFamilyArray()}, {@link #getQualifierArray()}, and 1473 * {@link #getValueArray()} if accessing a KeyValue client-side. 1474 * @return Copy of the key portion only. 1475 */ 1476 public byte [] getKey() { 1477 int keylength = getKeyLength(); 1478 byte [] key = new byte[keylength]; 1479 System.arraycopy(getBuffer(), getKeyOffset(), key, 0, keylength); 1480 return key; 1481 } 1482 1483 /** 1484 * 1485 * @return Timestamp 1486 */ 1487 @Override 1488 public long getTimestamp() { 1489 return getTimestamp(getKeyLength()); 1490 } 1491 1492 /** 1493 * @param keylength Pass if you have it to save on a int creation. 1494 * @return Timestamp 1495 */ 1496 long getTimestamp(final int keylength) { 1497 int tsOffset = getTimestampOffset(keylength); 1498 return Bytes.toLong(this.bytes, tsOffset); 1499 } 1500 1501 /** 1502 * @return KeyValue.TYPE byte representation 1503 */ 1504 @Override 1505 public byte getTypeByte() { 1506 return this.bytes[this.offset + getKeyLength() - 1 + ROW_OFFSET]; 1507 } 1508 1509 /** 1510 * This returns the offset where the tag actually starts. 1511 */ 1512 @Override 1513 public int getTagsOffset() { 1514 int tagsLen = getTagsLength(); 1515 if (tagsLen == 0) { 1516 return this.offset + this.length; 1517 } 1518 return this.offset + this.length - tagsLen; 1519 } 1520 1521 /** 1522 * This returns the total length of the tag bytes 1523 */ 1524 @Override 1525 public int getTagsLength() { 1526 int tagsLen = this.length - (getKeyLength() + getValueLength() + KEYVALUE_INFRASTRUCTURE_SIZE); 1527 if (tagsLen > 0) { 1528 // There are some Tag bytes in the byte[]. So reduce 2 bytes which is added to denote the tags 1529 // length 1530 tagsLen -= TAGS_LENGTH_SIZE; 1531 } 1532 return tagsLen; 1533 } 1534 1535 /** 1536 * @return the backing array of the entire KeyValue (all KeyValue fields are in a single array) 1537 */ 1538 @Override 1539 public byte[] getTagsArray() { 1540 return bytes; 1541 } 1542 1543 /** 1544 * Creates a new KeyValue that only contains the key portion (the value is 1545 * set to be null). 1546 * 1547 * TODO only used by KeyOnlyFilter -- move there. 1548 * @param lenAsVal replace value with the actual value length (false=empty) 1549 */ 1550 public KeyValue createKeyOnly(boolean lenAsVal) { 1551 // KV format: <keylen:4><valuelen:4><key:keylen><value:valuelen> 1552 // Rebuild as: <keylen:4><0:4><key:keylen> 1553 int dataLen = lenAsVal? Bytes.SIZEOF_INT : 0; 1554 byte [] newBuffer = new byte[getKeyLength() + ROW_OFFSET + dataLen]; 1555 System.arraycopy(this.bytes, this.offset, newBuffer, 0, 1556 Math.min(newBuffer.length,this.length)); 1557 Bytes.putInt(newBuffer, Bytes.SIZEOF_INT, dataLen); 1558 if (lenAsVal) { 1559 Bytes.putInt(newBuffer, newBuffer.length - dataLen, this.getValueLength()); 1560 } 1561 return new KeyValue(newBuffer); 1562 } 1563 1564 /** 1565 * @param b 1566 * @param delimiter 1567 * @return Index of delimiter having started from start of <code>b</code> 1568 * moving rightward. 1569 */ 1570 public static int getDelimiter(final byte [] b, int offset, final int length, 1571 final int delimiter) { 1572 if (b == null) { 1573 throw new IllegalArgumentException("Passed buffer is null"); 1574 } 1575 int result = -1; 1576 for (int i = offset; i < length + offset; i++) { 1577 if (b[i] == delimiter) { 1578 result = i; 1579 break; 1580 } 1581 } 1582 return result; 1583 } 1584 1585 /** 1586 * Find index of passed delimiter walking from end of buffer backwards. 1587 * @param b 1588 * @param delimiter 1589 * @return Index of delimiter 1590 */ 1591 public static int getDelimiterInReverse(final byte [] b, final int offset, 1592 final int length, final int delimiter) { 1593 if (b == null) { 1594 throw new IllegalArgumentException("Passed buffer is null"); 1595 } 1596 int result = -1; 1597 for (int i = (offset + length) - 1; i >= offset; i--) { 1598 if (b[i] == delimiter) { 1599 result = i; 1600 break; 1601 } 1602 } 1603 return result; 1604 } 1605 1606 /** 1607 * A {@link KVComparator} for <code>hbase:meta</code> catalog table 1608 * {@link KeyValue}s. 1609 * @deprecated : {@link CellComparatorImpl#META_COMPARATOR} to be used. Deprecated for hbase 2.0, remove for hbase 3.0. 1610 */ 1611 @Deprecated 1612 public static class MetaComparator extends KVComparator { 1613 /** 1614 * Compare key portion of a {@link KeyValue} for keys in <code>hbase:meta</code> 1615 * table. 1616 */ 1617 @Override 1618 public int compare(final Cell left, final Cell right) { 1619 return PrivateCellUtil.compareKeyIgnoresMvcc(CellComparatorImpl.META_COMPARATOR, left, right); 1620 } 1621 1622 @Override 1623 public int compareOnlyKeyPortion(Cell left, Cell right) { 1624 return compare(left, right); 1625 } 1626 1627 @Override 1628 public int compareRows(byte [] left, int loffset, int llength, 1629 byte [] right, int roffset, int rlength) { 1630 int leftDelimiter = getDelimiter(left, loffset, llength, 1631 HConstants.DELIMITER); 1632 int rightDelimiter = getDelimiter(right, roffset, rlength, 1633 HConstants.DELIMITER); 1634 // Compare up to the delimiter 1635 int lpart = (leftDelimiter < 0 ? llength :leftDelimiter - loffset); 1636 int rpart = (rightDelimiter < 0 ? rlength :rightDelimiter - roffset); 1637 int result = Bytes.compareTo(left, loffset, lpart, right, roffset, rpart); 1638 if (result != 0) { 1639 return result; 1640 } else { 1641 if (leftDelimiter < 0 && rightDelimiter >= 0) { 1642 return -1; 1643 } else if (rightDelimiter < 0 && leftDelimiter >= 0) { 1644 return 1; 1645 } else if (leftDelimiter < 0 && rightDelimiter < 0) { 1646 return 0; 1647 } 1648 } 1649 // Compare middle bit of the row. 1650 // Move past delimiter 1651 leftDelimiter++; 1652 rightDelimiter++; 1653 int leftFarDelimiter = getDelimiterInReverse(left, leftDelimiter, 1654 llength - (leftDelimiter - loffset), HConstants.DELIMITER); 1655 int rightFarDelimiter = getDelimiterInReverse(right, 1656 rightDelimiter, rlength - (rightDelimiter - roffset), 1657 HConstants.DELIMITER); 1658 // Now compare middlesection of row. 1659 lpart = (leftFarDelimiter < 0 ? llength + loffset: leftFarDelimiter) - leftDelimiter; 1660 rpart = (rightFarDelimiter < 0 ? rlength + roffset: rightFarDelimiter)- rightDelimiter; 1661 result = super.compareRows(left, leftDelimiter, lpart, right, rightDelimiter, rpart); 1662 if (result != 0) { 1663 return result; 1664 } else { 1665 if (leftDelimiter < 0 && rightDelimiter >= 0) { 1666 return -1; 1667 } else if (rightDelimiter < 0 && leftDelimiter >= 0) { 1668 return 1; 1669 } else if (leftDelimiter < 0 && rightDelimiter < 0) { 1670 return 0; 1671 } 1672 } 1673 // Compare last part of row, the rowid. 1674 leftFarDelimiter++; 1675 rightFarDelimiter++; 1676 result = Bytes.compareTo(left, leftFarDelimiter, llength - (leftFarDelimiter - loffset), 1677 right, rightFarDelimiter, rlength - (rightFarDelimiter - roffset)); 1678 return result; 1679 } 1680 1681 /** 1682 * Don't do any fancy Block Index splitting tricks. 1683 */ 1684 @Override 1685 public byte[] getShortMidpointKey(final byte[] leftKey, final byte[] rightKey) { 1686 return Arrays.copyOf(rightKey, rightKey.length); 1687 } 1688 1689 /** 1690 * The HFileV2 file format's trailer contains this class name. We reinterpret this and 1691 * instantiate the appropriate comparator. 1692 * TODO: With V3 consider removing this. 1693 * @return legacy class name for FileFileTrailer#comparatorClassName 1694 */ 1695 @Override 1696 public String getLegacyKeyComparatorName() { 1697 return "org.apache.hadoop.hbase.KeyValue$MetaKeyComparator"; 1698 } 1699 1700 @Override 1701 protected Object clone() throws CloneNotSupportedException { 1702 return new MetaComparator(); 1703 } 1704 1705 /** 1706 * Override the row key comparison to parse and compare the meta row key parts. 1707 */ 1708 @Override 1709 protected int compareRowKey(final Cell l, final Cell r) { 1710 byte[] left = l.getRowArray(); 1711 int loffset = l.getRowOffset(); 1712 int llength = l.getRowLength(); 1713 byte[] right = r.getRowArray(); 1714 int roffset = r.getRowOffset(); 1715 int rlength = r.getRowLength(); 1716 return compareRows(left, loffset, llength, right, roffset, rlength); 1717 } 1718 } 1719 1720 /** 1721 * Compare KeyValues. When we compare KeyValues, we only compare the Key 1722 * portion. This means two KeyValues with same Key but different Values are 1723 * considered the same as far as this Comparator is concerned. 1724 * @deprecated : Use {@link CellComparatorImpl}. Deprecated for hbase 2.0, remove for hbase 3.0. 1725 */ 1726 @Deprecated 1727 public static class KVComparator implements RawComparator<Cell>, SamePrefixComparator<byte[]> { 1728 1729 /** 1730 * The HFileV2 file format's trailer contains this class name. We reinterpret this and 1731 * instantiate the appropriate comparator. 1732 * TODO: With V3 consider removing this. 1733 * @return legacy class name for FileFileTrailer#comparatorClassName 1734 */ 1735 public String getLegacyKeyComparatorName() { 1736 return "org.apache.hadoop.hbase.KeyValue$KeyComparator"; 1737 } 1738 1739 @Override // RawComparator 1740 public int compare(byte[] l, int loff, int llen, byte[] r, int roff, int rlen) { 1741 return compareFlatKey(l,loff,llen, r,roff,rlen); 1742 } 1743 1744 1745 /** 1746 * Compares the only the user specified portion of a Key. This is overridden by MetaComparator. 1747 * @param left 1748 * @param right 1749 * @return 0 if equal, <0 if left smaller, >0 if right smaller 1750 */ 1751 protected int compareRowKey(final Cell left, final Cell right) { 1752 return CellComparatorImpl.COMPARATOR.compareRows(left, right); 1753 } 1754 1755 /** 1756 * Compares left to right assuming that left,loffset,llength and right,roffset,rlength are 1757 * full KVs laid out in a flat byte[]s. 1758 * @param left 1759 * @param loffset 1760 * @param llength 1761 * @param right 1762 * @param roffset 1763 * @param rlength 1764 * @return 0 if equal, <0 if left smaller, >0 if right smaller 1765 */ 1766 public int compareFlatKey(byte[] left, int loffset, int llength, 1767 byte[] right, int roffset, int rlength) { 1768 // Compare row 1769 short lrowlength = Bytes.toShort(left, loffset); 1770 short rrowlength = Bytes.toShort(right, roffset); 1771 int compare = compareRows(left, loffset + Bytes.SIZEOF_SHORT, 1772 lrowlength, right, roffset + Bytes.SIZEOF_SHORT, rrowlength); 1773 if (compare != 0) { 1774 return compare; 1775 } 1776 1777 // Compare the rest of the two KVs without making any assumptions about 1778 // the common prefix. This function will not compare rows anyway, so we 1779 // don't need to tell it that the common prefix includes the row. 1780 return compareWithoutRow(0, left, loffset, llength, right, roffset, 1781 rlength, rrowlength); 1782 } 1783 1784 public int compareFlatKey(byte[] left, byte[] right) { 1785 return compareFlatKey(left, 0, left.length, right, 0, right.length); 1786 } 1787 1788 // compare a key against row/fam/qual/ts/type 1789 public int compareKey(Cell cell, 1790 byte[] row, int roff, int rlen, 1791 byte[] fam, int foff, int flen, 1792 byte[] col, int coff, int clen, 1793 long ts, byte type) { 1794 1795 int compare = compareRows( 1796 cell.getRowArray(), cell.getRowOffset(), cell.getRowLength(), 1797 row, roff, rlen); 1798 if (compare != 0) { 1799 return compare; 1800 } 1801 // If the column is not specified, the "minimum" key type appears the 1802 // latest in the sorted order, regardless of the timestamp. This is used 1803 // for specifying the last key/value in a given row, because there is no 1804 // "lexicographically last column" (it would be infinitely long). The 1805 // "maximum" key type does not need this behavior. 1806 if (cell.getFamilyLength() + cell.getQualifierLength() == 0 1807 && cell.getTypeByte() == Type.Minimum.getCode()) { 1808 // left is "bigger", i.e. it appears later in the sorted order 1809 return 1; 1810 } 1811 if (flen+clen == 0 && type == Type.Minimum.getCode()) { 1812 return -1; 1813 } 1814 1815 compare = compareFamilies( 1816 cell.getFamilyArray(), cell.getFamilyOffset(), cell.getFamilyLength(), 1817 fam, foff, flen); 1818 if (compare != 0) { 1819 return compare; 1820 } 1821 compare = compareColumns( 1822 cell.getQualifierArray(), cell.getQualifierOffset(), cell.getQualifierLength(), 1823 col, coff, clen); 1824 if (compare != 0) { 1825 return compare; 1826 } 1827 // Next compare timestamps. 1828 compare = compareTimestamps(cell.getTimestamp(), ts); 1829 if (compare != 0) { 1830 return compare; 1831 } 1832 1833 // Compare types. Let the delete types sort ahead of puts; i.e. types 1834 // of higher numbers sort before those of lesser numbers. Maximum (255) 1835 // appears ahead of everything, and minimum (0) appears after 1836 // everything. 1837 return (0xff & type) - (0xff & cell.getTypeByte()); 1838 } 1839 1840 public int compareOnlyKeyPortion(Cell left, Cell right) { 1841 return PrivateCellUtil.compareKeyIgnoresMvcc(CellComparatorImpl.COMPARATOR, left, right); 1842 } 1843 1844 /** 1845 * Compares the Key of a cell -- with fields being more significant in this order: 1846 * rowkey, colfam/qual, timestamp, type, mvcc 1847 */ 1848 @Override 1849 public int compare(final Cell left, final Cell right) { 1850 int compare = CellComparatorImpl.COMPARATOR.compare(left, right); 1851 return compare; 1852 } 1853 1854 public int compareTimestamps(final Cell left, final Cell right) { 1855 return CellComparatorImpl.COMPARATOR.compareTimestamps(left, right); 1856 } 1857 1858 /** 1859 * @param left 1860 * @param right 1861 * @return Result comparing rows. 1862 */ 1863 public int compareRows(final Cell left, final Cell right) { 1864 return compareRows(left.getRowArray(),left.getRowOffset(), left.getRowLength(), 1865 right.getRowArray(), right.getRowOffset(), right.getRowLength()); 1866 } 1867 1868 /** 1869 * Get the b[],o,l for left and right rowkey portions and compare. 1870 * @param left 1871 * @param loffset 1872 * @param llength 1873 * @param right 1874 * @param roffset 1875 * @param rlength 1876 * @return 0 if equal, <0 if left smaller, >0 if right smaller 1877 */ 1878 public int compareRows(byte [] left, int loffset, int llength, 1879 byte [] right, int roffset, int rlength) { 1880 return Bytes.compareTo(left, loffset, llength, right, roffset, rlength); 1881 } 1882 1883 int compareColumns(final Cell left, final short lrowlength, final Cell right, 1884 final short rrowlength) { 1885 return CellComparatorImpl.COMPARATOR.compareColumns(left, right); 1886 } 1887 1888 protected int compareColumns( 1889 byte [] left, int loffset, int llength, final int lfamilylength, 1890 byte [] right, int roffset, int rlength, final int rfamilylength) { 1891 // Compare family portion first. 1892 int diff = Bytes.compareTo(left, loffset, lfamilylength, 1893 right, roffset, rfamilylength); 1894 if (diff != 0) { 1895 return diff; 1896 } 1897 // Compare qualifier portion 1898 return Bytes.compareTo(left, loffset + lfamilylength, 1899 llength - lfamilylength, 1900 right, roffset + rfamilylength, rlength - rfamilylength); 1901 } 1902 1903 static int compareTimestamps(final long ltimestamp, final long rtimestamp) { 1904 // The below older timestamps sorting ahead of newer timestamps looks 1905 // wrong but it is intentional. This way, newer timestamps are first 1906 // found when we iterate over a memstore and newer versions are the 1907 // first we trip over when reading from a store file. 1908 if (ltimestamp < rtimestamp) { 1909 return 1; 1910 } else if (ltimestamp > rtimestamp) { 1911 return -1; 1912 } 1913 return 0; 1914 } 1915 1916 /** 1917 * Overridden 1918 * @param commonPrefix 1919 * @param left 1920 * @param loffset 1921 * @param llength 1922 * @param right 1923 * @param roffset 1924 * @param rlength 1925 * @return 0 if equal, <0 if left smaller, >0 if right smaller 1926 */ 1927 @Override // SamePrefixComparator 1928 public int compareIgnoringPrefix(int commonPrefix, byte[] left, 1929 int loffset, int llength, byte[] right, int roffset, int rlength) { 1930 // Compare row 1931 short lrowlength = Bytes.toShort(left, loffset); 1932 short rrowlength; 1933 1934 int comparisonResult = 0; 1935 if (commonPrefix < ROW_LENGTH_SIZE) { 1936 // almost nothing in common 1937 rrowlength = Bytes.toShort(right, roffset); 1938 comparisonResult = compareRows(left, loffset + ROW_LENGTH_SIZE, 1939 lrowlength, right, roffset + ROW_LENGTH_SIZE, rrowlength); 1940 } else { // the row length is the same 1941 rrowlength = lrowlength; 1942 if (commonPrefix < ROW_LENGTH_SIZE + rrowlength) { 1943 // The rows are not the same. Exclude the common prefix and compare 1944 // the rest of the two rows. 1945 int common = commonPrefix - ROW_LENGTH_SIZE; 1946 comparisonResult = compareRows( 1947 left, loffset + common + ROW_LENGTH_SIZE, lrowlength - common, 1948 right, roffset + common + ROW_LENGTH_SIZE, rrowlength - common); 1949 } 1950 } 1951 if (comparisonResult != 0) { 1952 return comparisonResult; 1953 } 1954 1955 assert lrowlength == rrowlength; 1956 return compareWithoutRow(commonPrefix, left, loffset, llength, right, 1957 roffset, rlength, lrowlength); 1958 } 1959 1960 /** 1961 * Compare columnFamily, qualifier, timestamp, and key type (everything 1962 * except the row). This method is used both in the normal comparator and 1963 * the "same-prefix" comparator. Note that we are assuming that row portions 1964 * of both KVs have already been parsed and found identical, and we don't 1965 * validate that assumption here. 1966 * @param commonPrefix 1967 * the length of the common prefix of the two key-values being 1968 * compared, including row length and row 1969 */ 1970 private int compareWithoutRow(int commonPrefix, byte[] left, int loffset, 1971 int llength, byte[] right, int roffset, int rlength, short rowlength) { 1972 /*** 1973 * KeyValue Format and commonLength: 1974 * |_keyLen_|_valLen_|_rowLen_|_rowKey_|_famiLen_|_fami_|_Quali_|.... 1975 * ------------------|-------commonLength--------|-------------- 1976 */ 1977 int commonLength = ROW_LENGTH_SIZE + FAMILY_LENGTH_SIZE + rowlength; 1978 1979 // commonLength + TIMESTAMP_TYPE_SIZE 1980 int commonLengthWithTSAndType = TIMESTAMP_TYPE_SIZE + commonLength; 1981 // ColumnFamily + Qualifier length. 1982 int lcolumnlength = llength - commonLengthWithTSAndType; 1983 int rcolumnlength = rlength - commonLengthWithTSAndType; 1984 1985 byte ltype = left[loffset + (llength - 1)]; 1986 byte rtype = right[roffset + (rlength - 1)]; 1987 1988 // If the column is not specified, the "minimum" key type appears the 1989 // latest in the sorted order, regardless of the timestamp. This is used 1990 // for specifying the last key/value in a given row, because there is no 1991 // "lexicographically last column" (it would be infinitely long). The 1992 // "maximum" key type does not need this behavior. 1993 if (lcolumnlength == 0 && ltype == Type.Minimum.getCode()) { 1994 // left is "bigger", i.e. it appears later in the sorted order 1995 return 1; 1996 } 1997 if (rcolumnlength == 0 && rtype == Type.Minimum.getCode()) { 1998 return -1; 1999 } 2000 2001 int lfamilyoffset = commonLength + loffset; 2002 int rfamilyoffset = commonLength + roffset; 2003 2004 // Column family length. 2005 int lfamilylength = left[lfamilyoffset - 1]; 2006 int rfamilylength = right[rfamilyoffset - 1]; 2007 // If left family size is not equal to right family size, we need not 2008 // compare the qualifiers. 2009 boolean sameFamilySize = (lfamilylength == rfamilylength); 2010 int common = 0; 2011 if (commonPrefix > 0) { 2012 common = Math.max(0, commonPrefix - commonLength); 2013 if (!sameFamilySize) { 2014 // Common should not be larger than Math.min(lfamilylength, 2015 // rfamilylength). 2016 common = Math.min(common, Math.min(lfamilylength, rfamilylength)); 2017 } else { 2018 common = Math.min(common, Math.min(lcolumnlength, rcolumnlength)); 2019 } 2020 } 2021 if (!sameFamilySize) { 2022 // comparing column family is enough. 2023 return Bytes.compareTo(left, lfamilyoffset + common, lfamilylength 2024 - common, right, rfamilyoffset + common, rfamilylength - common); 2025 } 2026 // Compare family & qualifier together. 2027 final int comparison = Bytes.compareTo(left, lfamilyoffset + common, 2028 lcolumnlength - common, right, rfamilyoffset + common, 2029 rcolumnlength - common); 2030 if (comparison != 0) { 2031 return comparison; 2032 } 2033 2034 //// 2035 // Next compare timestamps. 2036 long ltimestamp = Bytes.toLong(left, 2037 loffset + (llength - TIMESTAMP_TYPE_SIZE)); 2038 long rtimestamp = Bytes.toLong(right, 2039 roffset + (rlength - TIMESTAMP_TYPE_SIZE)); 2040 int compare = compareTimestamps(ltimestamp, rtimestamp); 2041 if (compare != 0) { 2042 return compare; 2043 } 2044 2045 // Compare types. Let the delete types sort ahead of puts; i.e. types 2046 // of higher numbers sort before those of lesser numbers. Maximum (255) 2047 // appears ahead of everything, and minimum (0) appears after 2048 // everything. 2049 return (0xff & rtype) - (0xff & ltype); 2050 } 2051 2052 protected int compareFamilies(final byte[] left, final int loffset, final int lfamilylength, 2053 final byte[] right, final int roffset, final int rfamilylength) { 2054 int diff = Bytes.compareTo(left, loffset, lfamilylength, right, roffset, rfamilylength); 2055 return diff; 2056 } 2057 2058 protected int compareColumns(final byte[] left, final int loffset, final int lquallength, 2059 final byte[] right, final int roffset, final int rquallength) { 2060 int diff = Bytes.compareTo(left, loffset, lquallength, right, roffset, rquallength); 2061 return diff; 2062 } 2063 /** 2064 * Compares the row and column of two keyvalues for equality 2065 * @param left 2066 * @param right 2067 * @return True if same row and column. 2068 */ 2069 public boolean matchingRowColumn(final Cell left, 2070 final Cell right) { 2071 short lrowlength = left.getRowLength(); 2072 short rrowlength = right.getRowLength(); 2073 2074 // TsOffset = end of column data. just comparing Row+CF length of each 2075 if ((left.getRowLength() + left.getFamilyLength() + left.getQualifierLength()) != (right 2076 .getRowLength() + right.getFamilyLength() + right.getQualifierLength())) { 2077 return false; 2078 } 2079 2080 if (!matchingRows(left, lrowlength, right, rrowlength)) { 2081 return false; 2082 } 2083 2084 int lfoffset = left.getFamilyOffset(); 2085 int rfoffset = right.getFamilyOffset(); 2086 int lclength = left.getQualifierLength(); 2087 int rclength = right.getQualifierLength(); 2088 int lfamilylength = left.getFamilyLength(); 2089 int rfamilylength = right.getFamilyLength(); 2090 int diff = compareFamilies(left.getFamilyArray(), lfoffset, lfamilylength, 2091 right.getFamilyArray(), rfoffset, rfamilylength); 2092 if (diff != 0) { 2093 return false; 2094 } else { 2095 diff = compareColumns(left.getQualifierArray(), left.getQualifierOffset(), lclength, 2096 right.getQualifierArray(), right.getQualifierOffset(), rclength); 2097 return diff == 0; 2098 } 2099 } 2100 2101 /** 2102 * Compares the row of two keyvalues for equality 2103 * @param left 2104 * @param right 2105 * @return True if rows match. 2106 */ 2107 public boolean matchingRows(final Cell left, final Cell right) { 2108 short lrowlength = left.getRowLength(); 2109 short rrowlength = right.getRowLength(); 2110 return matchingRows(left, lrowlength, right, rrowlength); 2111 } 2112 2113 /** 2114 * @param left 2115 * @param lrowlength 2116 * @param right 2117 * @param rrowlength 2118 * @return True if rows match. 2119 */ 2120 private boolean matchingRows(final Cell left, final short lrowlength, 2121 final Cell right, final short rrowlength) { 2122 return lrowlength == rrowlength && 2123 matchingRows(left.getRowArray(), left.getRowOffset(), lrowlength, 2124 right.getRowArray(), right.getRowOffset(), rrowlength); 2125 } 2126 2127 /** 2128 * Compare rows. Just calls Bytes.equals, but it's good to have this encapsulated. 2129 * @param left Left row array. 2130 * @param loffset Left row offset. 2131 * @param llength Left row length. 2132 * @param right Right row array. 2133 * @param roffset Right row offset. 2134 * @param rlength Right row length. 2135 * @return Whether rows are the same row. 2136 */ 2137 public boolean matchingRows(final byte [] left, final int loffset, final int llength, 2138 final byte [] right, final int roffset, final int rlength) { 2139 return Bytes.equals(left, loffset, llength, right, roffset, rlength); 2140 } 2141 2142 public byte[] calcIndexKey(byte[] lastKeyOfPreviousBlock, byte[] firstKeyInBlock) { 2143 byte[] fakeKey = getShortMidpointKey(lastKeyOfPreviousBlock, firstKeyInBlock); 2144 if (compareFlatKey(fakeKey, firstKeyInBlock) > 0) { 2145 LOG.error("Unexpected getShortMidpointKey result, fakeKey:" 2146 + Bytes.toStringBinary(fakeKey) + ", firstKeyInBlock:" 2147 + Bytes.toStringBinary(firstKeyInBlock)); 2148 return firstKeyInBlock; 2149 } 2150 if (lastKeyOfPreviousBlock != null && compareFlatKey(lastKeyOfPreviousBlock, fakeKey) >= 0) { 2151 LOG.error("Unexpected getShortMidpointKey result, lastKeyOfPreviousBlock:" + 2152 Bytes.toStringBinary(lastKeyOfPreviousBlock) + ", fakeKey:" + 2153 Bytes.toStringBinary(fakeKey)); 2154 return firstKeyInBlock; 2155 } 2156 return fakeKey; 2157 } 2158 2159 /** 2160 * This is a HFile block index key optimization. 2161 * @param leftKey 2162 * @param rightKey 2163 * @return 0 if equal, <0 if left smaller, >0 if right smaller 2164 * @deprecated Since 0.99.2; 2165 */ 2166 @Deprecated 2167 public byte[] getShortMidpointKey(final byte[] leftKey, final byte[] rightKey) { 2168 if (rightKey == null) { 2169 throw new IllegalArgumentException("rightKey can not be null"); 2170 } 2171 if (leftKey == null) { 2172 return Arrays.copyOf(rightKey, rightKey.length); 2173 } 2174 if (compareFlatKey(leftKey, rightKey) >= 0) { 2175 throw new IllegalArgumentException("Unexpected input, leftKey:" + Bytes.toString(leftKey) 2176 + ", rightKey:" + Bytes.toString(rightKey)); 2177 } 2178 2179 short leftRowLength = Bytes.toShort(leftKey, 0); 2180 short rightRowLength = Bytes.toShort(rightKey, 0); 2181 int leftCommonLength = ROW_LENGTH_SIZE + FAMILY_LENGTH_SIZE + leftRowLength; 2182 int rightCommonLength = ROW_LENGTH_SIZE + FAMILY_LENGTH_SIZE + rightRowLength; 2183 int leftCommonLengthWithTSAndType = TIMESTAMP_TYPE_SIZE + leftCommonLength; 2184 int rightCommonLengthWithTSAndType = TIMESTAMP_TYPE_SIZE + rightCommonLength; 2185 int leftColumnLength = leftKey.length - leftCommonLengthWithTSAndType; 2186 int rightColumnLength = rightKey.length - rightCommonLengthWithTSAndType; 2187 // rows are equal 2188 if (leftRowLength == rightRowLength && compareRows(leftKey, ROW_LENGTH_SIZE, leftRowLength, 2189 rightKey, ROW_LENGTH_SIZE, rightRowLength) == 0) { 2190 // Compare family & qualifier together. 2191 int comparison = Bytes.compareTo(leftKey, leftCommonLength, leftColumnLength, rightKey, 2192 rightCommonLength, rightColumnLength); 2193 // same with "row + family + qualifier", return rightKey directly 2194 if (comparison == 0) { 2195 return Arrays.copyOf(rightKey, rightKey.length); 2196 } 2197 // "family + qualifier" are different, generate a faked key per rightKey 2198 byte[] newKey = Arrays.copyOf(rightKey, rightKey.length); 2199 Bytes.putLong(newKey, rightKey.length - TIMESTAMP_TYPE_SIZE, HConstants.LATEST_TIMESTAMP); 2200 Bytes.putByte(newKey, rightKey.length - TYPE_SIZE, Type.Maximum.getCode()); 2201 return newKey; 2202 } 2203 // rows are different 2204 short minLength = leftRowLength < rightRowLength ? leftRowLength : rightRowLength; 2205 short diffIdx = 0; 2206 while (diffIdx < minLength 2207 && leftKey[ROW_LENGTH_SIZE + diffIdx] == rightKey[ROW_LENGTH_SIZE + diffIdx]) { 2208 diffIdx++; 2209 } 2210 byte[] newRowKey = null; 2211 if (diffIdx >= minLength) { 2212 // leftKey's row is prefix of rightKey's. 2213 newRowKey = new byte[diffIdx + 1]; 2214 System.arraycopy(rightKey, ROW_LENGTH_SIZE, newRowKey, 0, diffIdx + 1); 2215 } else { 2216 int diffByte = leftKey[ROW_LENGTH_SIZE + diffIdx]; 2217 if ((0xff & diffByte) < 0xff && (diffByte + 1) < 2218 (rightKey[ROW_LENGTH_SIZE + diffIdx] & 0xff)) { 2219 newRowKey = new byte[diffIdx + 1]; 2220 System.arraycopy(leftKey, ROW_LENGTH_SIZE, newRowKey, 0, diffIdx); 2221 newRowKey[diffIdx] = (byte) (diffByte + 1); 2222 } else { 2223 newRowKey = new byte[diffIdx + 1]; 2224 System.arraycopy(rightKey, ROW_LENGTH_SIZE, newRowKey, 0, diffIdx + 1); 2225 } 2226 } 2227 return new KeyValue(newRowKey, null, null, HConstants.LATEST_TIMESTAMP, 2228 Type.Maximum).getKey(); 2229 } 2230 2231 @Override 2232 protected Object clone() throws CloneNotSupportedException { 2233 super.clone(); 2234 return new KVComparator(); 2235 } 2236 2237 } 2238 2239 /** 2240 * @param in Where to read bytes from. Creates a byte array to hold the KeyValue 2241 * backing bytes copied from the steam. 2242 * @return KeyValue created by deserializing from <code>in</code> OR if we find a length 2243 * of zero, we will return null which can be useful marking a stream as done. 2244 * @throws IOException 2245 */ 2246 public static KeyValue create(final DataInput in) throws IOException { 2247 return create(in.readInt(), in); 2248 } 2249 2250 /** 2251 * Create a KeyValue reading <code>length</code> from <code>in</code> 2252 * @param length 2253 * @param in 2254 * @return Created KeyValue OR if we find a length of zero, we will return null which 2255 * can be useful marking a stream as done. 2256 * @throws IOException 2257 */ 2258 public static KeyValue create(int length, final DataInput in) throws IOException { 2259 2260 if (length <= 0) { 2261 if (length == 0) return null; 2262 throw new IOException("Failed read " + length + " bytes, stream corrupt?"); 2263 } 2264 2265 // This is how the old Writables.readFrom used to deserialize. Didn't even vint. 2266 byte [] bytes = new byte[length]; 2267 in.readFully(bytes); 2268 return new KeyValue(bytes, 0, length); 2269 } 2270 2271 /** 2272 * Write out a KeyValue in the manner in which we used to when KeyValue was a Writable. 2273 * @param kv 2274 * @param out 2275 * @return Length written on stream 2276 * @throws IOException 2277 * @see #create(DataInput) for the inverse function 2278 */ 2279 public static long write(final KeyValue kv, final DataOutput out) throws IOException { 2280 // This is how the old Writables write used to serialize KVs. Need to figure way to make it 2281 // work for all implementations. 2282 int length = kv.getLength(); 2283 out.writeInt(length); 2284 out.write(kv.getBuffer(), kv.getOffset(), length); 2285 return (long) length + Bytes.SIZEOF_INT; 2286 } 2287 2288 /** 2289 * Write out a KeyValue in the manner in which we used to when KeyValue was a Writable but do 2290 * not require a {@link DataOutput}, just take plain {@link OutputStream} 2291 * Named <code>oswrite</code> so does not clash with {@link #write(KeyValue, DataOutput)} 2292 * @param kv 2293 * @param out 2294 * @param withTags 2295 * @return Length written on stream 2296 * @throws IOException 2297 * @see #create(DataInput) for the inverse function 2298 * @see #write(KeyValue, DataOutput) 2299 * @see KeyValueUtil#oswrite(Cell, OutputStream, boolean) 2300 * @deprecated As of release 2.0.0, this will be removed in HBase 3.0.0. 2301 * Instead use {@link #write(OutputStream, boolean)} 2302 */ 2303 @Deprecated 2304 public static long oswrite(final KeyValue kv, final OutputStream out, final boolean withTags) 2305 throws IOException { 2306 ByteBufferUtils.putInt(out, kv.getSerializedSize(withTags)); 2307 return (long) kv.write(out, withTags) + Bytes.SIZEOF_INT; 2308 } 2309 2310 @Override 2311 public int write(OutputStream out, boolean withTags) throws IOException { 2312 int len = getSerializedSize(withTags); 2313 out.write(this.bytes, this.offset, len); 2314 return len; 2315 } 2316 2317 @Override 2318 public int getSerializedSize(boolean withTags) { 2319 if (withTags) { 2320 return this.length; 2321 } 2322 return this.getKeyLength() + this.getValueLength() + KEYVALUE_INFRASTRUCTURE_SIZE; 2323 } 2324 2325 @Override 2326 public void write(ByteBuffer buf, int offset) { 2327 ByteBufferUtils.copyFromArrayToBuffer(buf, offset, this.bytes, this.offset, this.length); 2328 } 2329 2330 /** 2331 * Avoids redundant comparisons for better performance. 2332 * 2333 * TODO get rid of this wart 2334 */ 2335 public interface SamePrefixComparator<T> { 2336 /** 2337 * Compare two keys assuming that the first n bytes are the same. 2338 * @param commonPrefix How many bytes are the same. 2339 */ 2340 int compareIgnoringPrefix(int commonPrefix, byte[] left, int loffset, int llength, 2341 byte[] right, int roffset, int rlength 2342 ); 2343 } 2344 2345 /** 2346 * HeapSize implementation 2347 * 2348 * We do not count the bytes in the rowCache because it should be empty for a KeyValue in the 2349 * MemStore. 2350 */ 2351 @Override 2352 public long heapSize() { 2353 /* 2354 * Deep object overhead for this KV consists of two parts. The first part is the KV object 2355 * itself, while the second part is the backing byte[]. We will only count the array overhead 2356 * from the byte[] only if this is the first KV in there. 2357 */ 2358 return ClassSize.align(FIXED_OVERHEAD) + 2359 (offset == 0 2360 ? ClassSize.sizeOfByteArray(length) // count both length and object overhead 2361 : length); // only count the number of bytes 2362 } 2363 2364 /** 2365 * A simple form of KeyValue that creates a keyvalue with only the key part of the byte[] 2366 * Mainly used in places where we need to compare two cells. Avoids copying of bytes 2367 * In places like block index keys, we need to compare the key byte[] with a cell. 2368 * Hence create a Keyvalue(aka Cell) that would help in comparing as two cells 2369 */ 2370 public static class KeyOnlyKeyValue extends KeyValue { 2371 private short rowLen = -1; 2372 public KeyOnlyKeyValue() { 2373 2374 } 2375 public KeyOnlyKeyValue(byte[] b) { 2376 this(b, 0, b.length); 2377 } 2378 2379 public KeyOnlyKeyValue(byte[] b, int offset, int length) { 2380 this.bytes = b; 2381 this.length = length; 2382 this.offset = offset; 2383 this.rowLen = Bytes.toShort(this.bytes, this.offset); 2384 } 2385 2386 public void set(KeyOnlyKeyValue keyOnlyKeyValue) { 2387 this.bytes = keyOnlyKeyValue.bytes; 2388 this.length = keyOnlyKeyValue.length; 2389 this.offset = keyOnlyKeyValue.offset; 2390 this.rowLen = keyOnlyKeyValue.rowLen; 2391 } 2392 2393 public void clear() { 2394 rowLen = -1; 2395 bytes = null; 2396 offset = 0; 2397 length = 0; 2398 } 2399 2400 @Override 2401 public int getKeyOffset() { 2402 return this.offset; 2403 } 2404 2405 /** 2406 * A setter that helps to avoid object creation every time and whenever 2407 * there is a need to create new KeyOnlyKeyValue. 2408 * @param key 2409 * @param offset 2410 * @param length 2411 */ 2412 public void setKey(byte[] key, int offset, int length) { 2413 this.bytes = key; 2414 this.offset = offset; 2415 this.length = length; 2416 this.rowLen = Bytes.toShort(this.bytes, this.offset); 2417 } 2418 2419 @Override 2420 public byte[] getKey() { 2421 int keylength = getKeyLength(); 2422 byte[] key = new byte[keylength]; 2423 System.arraycopy(this.bytes, getKeyOffset(), key, 0, keylength); 2424 return key; 2425 } 2426 2427 @Override 2428 public byte[] getRowArray() { 2429 return bytes; 2430 } 2431 2432 @Override 2433 public int getRowOffset() { 2434 return getKeyOffset() + Bytes.SIZEOF_SHORT; 2435 } 2436 2437 @Override 2438 public byte[] getFamilyArray() { 2439 return bytes; 2440 } 2441 2442 @Override 2443 public byte getFamilyLength() { 2444 return this.bytes[getFamilyOffset() - 1]; 2445 } 2446 2447 @Override 2448 public int getFamilyOffset() { 2449 return this.offset + Bytes.SIZEOF_SHORT + getRowLength() + Bytes.SIZEOF_BYTE; 2450 } 2451 2452 @Override 2453 public byte[] getQualifierArray() { 2454 return bytes; 2455 } 2456 2457 @Override 2458 public int getQualifierLength() { 2459 return getQualifierLength(getRowLength(), getFamilyLength()); 2460 } 2461 2462 @Override 2463 public int getQualifierOffset() { 2464 return getFamilyOffset() + getFamilyLength(); 2465 } 2466 2467 @Override 2468 public int getKeyLength() { 2469 return length; 2470 } 2471 2472 @Override 2473 public short getRowLength() { 2474 return rowLen; 2475 } 2476 2477 @Override 2478 public byte getTypeByte() { 2479 return this.bytes[this.offset + getKeyLength() - 1]; 2480 } 2481 2482 private int getQualifierLength(int rlength, int flength) { 2483 return getKeyLength() - (int) getKeyDataStructureSize(rlength, flength, 0); 2484 } 2485 2486 @Override 2487 public long getTimestamp() { 2488 int tsOffset = getTimestampOffset(); 2489 return Bytes.toLong(this.bytes, tsOffset); 2490 } 2491 2492 @Override 2493 public int getTimestampOffset() { 2494 return getKeyOffset() + getKeyLength() - TIMESTAMP_TYPE_SIZE; 2495 } 2496 2497 @Override 2498 public byte[] getTagsArray() { 2499 return HConstants.EMPTY_BYTE_ARRAY; 2500 } 2501 2502 @Override 2503 public int getTagsOffset() { 2504 return 0; 2505 } 2506 2507 @Override 2508 public byte[] getValueArray() { 2509 throw new IllegalArgumentException("KeyOnlyKeyValue does not work with values."); 2510 } 2511 2512 @Override 2513 public int getValueOffset() { 2514 throw new IllegalArgumentException("KeyOnlyKeyValue does not work with values."); 2515 } 2516 2517 @Override 2518 public int getValueLength() { 2519 throw new IllegalArgumentException("KeyOnlyKeyValue does not work with values."); 2520 } 2521 2522 @Override 2523 public int getTagsLength() { 2524 return 0; 2525 } 2526 2527 @Override 2528 public String toString() { 2529 if (this.bytes == null || this.bytes.length == 0) { 2530 return "empty"; 2531 } 2532 return keyToString(this.bytes, this.offset, getKeyLength()) + "/vlen=0/mvcc=0"; 2533 } 2534 2535 @Override 2536 public int hashCode() { 2537 return super.hashCode(); 2538 } 2539 2540 @Override 2541 public boolean equals(Object other) { 2542 return super.equals(other); 2543 } 2544 2545 @Override 2546 public long heapSize() { 2547 return super.heapSize() + Bytes.SIZEOF_SHORT; 2548 } 2549 2550 @Override 2551 public int write(OutputStream out, boolean withTags) throws IOException { 2552 // This type of Cell is used only to maintain some internal states. We never allow this type 2553 // of Cell to be returned back over the RPC 2554 throw new IllegalStateException("A reader should never return this type of a Cell"); 2555 } 2556 } 2557 2558 @Override 2559 public ExtendedCell deepClone() { 2560 byte[] copy = Bytes.copy(this.bytes, this.offset, this.length); 2561 KeyValue kv = new KeyValue(copy, 0, copy.length); 2562 kv.setSequenceId(this.getSequenceId()); 2563 return kv; 2564 } 2565}