| 1 |
/* |
| 2 |
* Licensed under the Apache License, Version 2.0 (the "License"); |
| 3 |
* you may not use this file except in compliance with the License. |
| 4 |
* You may obtain a copy of the License at |
| 5 |
* |
| 6 |
* http://www.apache.org/licenses/LICENSE-2.0 |
| 7 |
* |
| 8 |
* Unless required by applicable law or agreed to in writing, software |
| 9 |
* distributed under the License is distributed on an "AS IS" BASIS, |
| 10 |
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 11 |
* See the License for the specific language governing permissions and |
| 12 |
* limitations under the License. |
| 13 |
*/ |
| 14 |
package io.airlift.slice; |
| 15 |
|
| 16 |
import java.util.OptionalInt; |
| 17 |
|
| 18 |
import static io.airlift.slice.Preconditions.checkArgument; |
| 19 |
import static io.airlift.slice.Preconditions.checkPositionIndex; |
| 20 |
import static io.airlift.slice.Preconditions.checkPositionIndexes; |
| 21 |
import static java.lang.Character.MAX_CODE_POINT; |
| 22 |
import static java.lang.Character.MAX_SURROGATE; |
| 23 |
import static java.lang.Character.MIN_SURROGATE; |
| 24 |
import static java.lang.Integer.toHexString; |
| 25 |
|
| 26 |
/** |
| 27 |
* Utility methods for UTF-8 encoded slices. |
| 28 |
*/ |
| 29 |
public final class SliceUtf8 |
| 30 |
{ |
| 31 |
private SliceUtf8() {} |
| 32 |
|
| 33 |
private static final int REPLACEMENT_CODE_POINT = 0xFFFD; |
| 34 |
|
| 35 |
private static final int TOP_MASK32 = 0x8080_8080; |
| 36 |
private static final long TOP_MASK64 = 0x8080_8080_8080_8080L; |
| 37 |
|
| 38 |
private static final int[] LOWER_CODE_POINTS; |
| 39 |
private static final int[] UPPER_CODE_POINTS; |
| 40 |
private static final boolean[] WHITESPACE_CODE_POINTS; |
| 41 |
|
| 42 |
static { |
| 43 |
LOWER_CODE_POINTS = new int[MAX_CODE_POINT + 1]; |
| 44 |
UPPER_CODE_POINTS = new int[MAX_CODE_POINT + 1]; |
| 45 |
WHITESPACE_CODE_POINTS = new boolean[MAX_CODE_POINT + 1]; |
| 46 |
for (int codePoint = 0; codePoint <= MAX_CODE_POINT; codePoint++) { |
| 47 |
int type = Character.getType(codePoint); |
| 48 |
if (type != Character.SURROGATE) { |
| 49 |
LOWER_CODE_POINTS[codePoint] = Character.toLowerCase(codePoint); |
| 50 |
UPPER_CODE_POINTS[codePoint] = Character.toUpperCase(codePoint); |
| 51 |
WHITESPACE_CODE_POINTS[codePoint] = Character.isWhitespace(codePoint); |
| 52 |
} |
| 53 |
else { |
| 54 |
LOWER_CODE_POINTS[codePoint] = REPLACEMENT_CODE_POINT; |
| 55 |
UPPER_CODE_POINTS[codePoint] = REPLACEMENT_CODE_POINT; |
| 56 |
WHITESPACE_CODE_POINTS[codePoint] = false; |
| 57 |
} |
| 58 |
} |
| 59 |
} |
| 60 |
|
| 61 |
/** |
| 62 |
* Does the slice contain only 7-bit ASCII characters. |
| 63 |
*/ |
| 64 |
public static boolean isAscii(Slice utf8) |
| 65 |
{ |
| 66 |
int length = utf8.length(); |
| 67 |
int offset = 0; |
| 68 |
|
| 69 |
// Length rounded to 8 bytes |
| 70 |
int length8 = length & 0x7FFF_FFF8; |
| 71 |
for (; offset < length8; offset += 8) { |
| 72 |
if ((utf8.getLongUnchecked(offset) & TOP_MASK64) != 0) { |
| 73 |
return false; |
| 74 |
} |
| 75 |
} |
| 76 |
// Enough bytes left for 32 bits? |
| 77 |
if (offset + 4 < length) { |
| 78 |
if ((utf8.getIntUnchecked(offset) & TOP_MASK32) != 0) { |
| 79 |
return false; |
| 80 |
} |
| 81 |
|
| 82 |
offset += 4; |
| 83 |
} |
| 84 |
// Do the rest one by one |
| 85 |
for (; offset < length; offset++) { |
| 86 |
if ((utf8.getByteUnchecked(offset) & 0x80) != 0) { |
| 87 |
return false; |
| 88 |
} |
| 89 |
} |
| 90 |
|
| 91 |
return true; |
| 92 |
} |
| 93 |
|
| 94 |
/** |
| 95 |
* Counts the code points within UTF-8 encoded slice. |
| 96 |
* <p> |
| 97 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 98 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 99 |
*/ |
| 100 |
public static int countCodePoints(Slice utf8) |
| 101 |
{ |
| 102 |
return countCodePoints(utf8, 0, utf8.length()); |
| 103 |
} |
| 104 |
|
| 105 |
/** |
| 106 |
* Counts the code points within UTF-8 encoded slice up to {@code length}. |
| 107 |
* <p> |
| 108 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 109 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 110 |
*/ |
| 111 |
public static int countCodePoints(Slice utf8, int offset, int length) |
| 112 |
{ |
| 113 |
checkPositionIndexes(offset, offset + length, utf8.length()); |
| 114 |
|
| 115 |
// Quick exit if empty string |
| 116 |
if (length == 0) { |
| 117 |
return 0; |
| 118 |
} |
| 119 |
|
| 120 |
int continuationBytesCount = 0; |
| 121 |
// Length rounded to 8 bytes |
| 122 |
int length8 = length & 0x7FFF_FFF8; |
| 123 |
for (; offset < length8; offset += 8) { |
| 124 |
// Count bytes which are NOT the start of a code point |
| 125 |
continuationBytesCount += countContinuationBytes(utf8.getLongUnchecked(offset)); |
| 126 |
} |
| 127 |
// Enough bytes left for 32 bits? |
| 128 |
if (offset + 4 < length) { |
| 129 |
// Count bytes which are NOT the start of a code point |
| 130 |
continuationBytesCount += countContinuationBytes(utf8.getIntUnchecked(offset)); |
| 131 |
|
| 132 |
offset += 4; |
| 133 |
} |
| 134 |
// Do the rest one by one |
| 135 |
for (; offset < length; offset++) { |
| 136 |
// Count bytes which are NOT the start of a code point |
| 137 |
continuationBytesCount += countContinuationBytes(utf8.getByteUnchecked(offset)); |
| 138 |
} |
| 139 |
|
| 140 |
assert continuationBytesCount <= length; |
| 141 |
return length - continuationBytesCount; |
| 142 |
} |
| 143 |
|
| 144 |
/** |
| 145 |
* Gets the substring starting at {@code codePointStart} and extending for |
| 146 |
* {@code codePointLength} code points. |
| 147 |
* <p> |
| 148 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 149 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 150 |
*/ |
| 151 |
public static Slice substring(Slice utf8, int codePointStart, int codePointLength) |
| 152 |
{ |
| 153 |
checkArgument(codePointStart >= 0, "codePointStart is negative"); |
| 154 |
checkArgument(codePointLength >= 0, "codePointLength is negative"); |
| 155 |
|
| 156 |
int indexStart = offsetOfCodePoint(utf8, codePointStart); |
| 157 |
if (indexStart < 0) { |
| 158 |
throw new IllegalArgumentException("UTF-8 does not contain " + codePointStart + " code points"); |
| 159 |
} |
| 160 |
if (codePointLength == 0) { |
| 161 |
return Slices.EMPTY_SLICE; |
| 162 |
} |
| 163 |
int indexEnd = offsetOfCodePoint(utf8, indexStart, codePointLength - 1); |
| 164 |
if (indexEnd < 0) { |
| 165 |
throw new IllegalArgumentException("UTF-8 does not contain " + (codePointStart + codePointLength) + " code points"); |
| 166 |
} |
| 167 |
indexEnd += lengthOfCodePoint(utf8, indexEnd); |
| 168 |
if (indexEnd > utf8.length()) { |
| 169 |
throw new InvalidUtf8Exception("UTF-8 is not well formed"); |
| 170 |
} |
| 171 |
return utf8.slice(indexStart, indexEnd - indexStart); |
| 172 |
} |
| 173 |
|
| 174 |
/** |
| 175 |
* Reverses the slice code point by code point. |
| 176 |
* <p> |
| 177 |
* Note: Invalid UTF-8 sequences are copied directly to the output. |
| 178 |
*/ |
| 179 |
public static Slice reverse(Slice utf8) |
| 180 |
{ |
| 181 |
int length = utf8.length(); |
| 182 |
Slice reverse = Slices.allocate(length); |
| 183 |
|
| 184 |
int forwardPosition = 0; |
| 185 |
int reversePosition = length; |
| 186 |
while (forwardPosition < length) { |
| 187 |
int codePointLength = lengthOfCodePointSafe(utf8, forwardPosition); |
| 188 |
|
| 189 |
// backup the reverse pointer |
| 190 |
reversePosition -= codePointLength; |
| 191 |
if (reversePosition < 0) { |
| 192 |
// this should not happen |
| 193 |
throw new InvalidUtf8Exception("UTF-8 is not well formed"); |
| 194 |
} |
| 195 |
// copy the character |
| 196 |
copyUtf8SequenceUnsafe(utf8, forwardPosition, reverse, reversePosition, codePointLength); |
| 197 |
|
| 198 |
forwardPosition += codePointLength; |
| 199 |
} |
| 200 |
return reverse; |
| 201 |
} |
| 202 |
|
| 203 |
/** |
| 204 |
* Converts slice to upper case code point by code point. This method does |
| 205 |
* not perform perform locale-sensitive, context-sensitive, or one-to-many |
| 206 |
* mappings required for some languages. Specifically, this will return |
| 207 |
* incorrect results for Lithuanian, Turkish, and Azeri. |
| 208 |
* <p> |
| 209 |
* Note: Invalid UTF-8 sequences are copied directly to the output. |
| 210 |
*/ |
| 211 |
public static Slice toUpperCase(Slice utf8) |
| 212 |
{ |
| 213 |
return translateCodePoints(utf8, UPPER_CODE_POINTS); |
| 214 |
} |
| 215 |
|
| 216 |
/** |
| 217 |
* Converts slice to lower case code point by code point. This method does |
| 218 |
* not perform perform locale-sensitive, context-sensitive, or one-to-many |
| 219 |
* mappings required for some languages. Specifically, this will return |
| 220 |
* incorrect results for Lithuanian, Turkish, and Azeri. |
| 221 |
* <p> |
| 222 |
* Note: Invalid UTF-8 sequences are copied directly to the output. |
| 223 |
*/ |
| 224 |
public static Slice toLowerCase(Slice utf8) |
| 225 |
{ |
| 226 |
return translateCodePoints(utf8, LOWER_CODE_POINTS); |
| 227 |
} |
| 228 |
|
| 229 |
private static Slice translateCodePoints(Slice utf8, int[] codePointTranslationMap) |
| 230 |
{ |
| 231 |
int length = utf8.length(); |
| 232 |
Slice newUtf8 = Slices.allocate(length); |
| 233 |
|
| 234 |
int position = 0; |
| 235 |
int upperPosition = 0; |
| 236 |
while (position < length) { |
| 237 |
int codePoint = tryGetCodePointAt(utf8, position); |
| 238 |
if (codePoint >= 0) { |
| 239 |
int upperCodePoint = codePointTranslationMap[codePoint]; |
| 240 |
|
| 241 |
// grow slice if necessary |
| 242 |
int nextUpperPosition = upperPosition + lengthOfCodePoint(upperCodePoint); |
| 243 |
if (nextUpperPosition > length) { |
| 244 |
newUtf8 = Slices.ensureSize(newUtf8, nextUpperPosition); |
| 245 |
} |
| 246 |
|
| 247 |
// write new byte |
| 248 |
setCodePointAt(upperCodePoint, newUtf8, upperPosition); |
| 249 |
|
| 250 |
position += lengthOfCodePoint(codePoint); |
| 251 |
upperPosition = nextUpperPosition; |
| 252 |
} |
| 253 |
else { |
| 254 |
int skipLength = -codePoint; |
| 255 |
|
| 256 |
// grow slice if necessary |
| 257 |
int nextUpperPosition = upperPosition + skipLength; |
| 258 |
if (nextUpperPosition > length) { |
| 259 |
newUtf8 = Slices.ensureSize(newUtf8, nextUpperPosition); |
| 260 |
} |
| 261 |
|
| 262 |
copyUtf8SequenceUnsafe(utf8, position, newUtf8, upperPosition, skipLength); |
| 263 |
position += skipLength; |
| 264 |
upperPosition = nextUpperPosition; |
| 265 |
} |
| 266 |
} |
| 267 |
return newUtf8.slice(0, upperPosition); |
| 268 |
} |
| 269 |
|
| 270 |
private static void copyUtf8SequenceUnsafe(Slice source, int sourcePosition, Slice destination, int destinationPosition, int length) |
| 271 |
{ |
| 272 |
switch (length) { |
| 273 |
case 1: |
| 274 |
destination.setByteUnchecked(destinationPosition, source.getByteUnchecked(sourcePosition)); |
| 275 |
break; |
| 276 |
case 2: |
| 277 |
destination.setShortUnchecked(destinationPosition, source.getShortUnchecked(sourcePosition)); |
| 278 |
break; |
| 279 |
case 3: |
| 280 |
destination.setShortUnchecked(destinationPosition, source.getShortUnchecked(sourcePosition)); |
| 281 |
destination.setByteUnchecked(destinationPosition + 2, source.getByteUnchecked(sourcePosition + 2)); |
| 282 |
break; |
| 283 |
case 4: |
| 284 |
destination.setIntUnchecked(destinationPosition, source.getIntUnchecked(sourcePosition)); |
| 285 |
break; |
| 286 |
case 5: |
| 287 |
destination.setIntUnchecked(destinationPosition, source.getIntUnchecked(sourcePosition)); |
| 288 |
destination.setByteUnchecked(destinationPosition + 4, source.getByteUnchecked(sourcePosition + 4)); |
| 289 |
break; |
| 290 |
case 6: |
| 291 |
destination.setIntUnchecked(destinationPosition, source.getIntUnchecked(sourcePosition)); |
| 292 |
destination.setShortUnchecked(destinationPosition + 4, source.getShortUnchecked(sourcePosition + 4)); |
| 293 |
break; |
| 294 |
default: |
| 295 |
throw new IllegalStateException("Invalid code point length " + length); |
| 296 |
} |
| 297 |
} |
| 298 |
|
| 299 |
/** |
| 300 |
* Removes all white space characters from the left string of the string. |
| 301 |
* <p> |
| 302 |
* Note: Invalid UTF-8 sequences are not trimmed. |
| 303 |
*/ |
| 304 |
public static Slice leftTrim(Slice utf8) |
| 305 |
{ |
| 306 |
int length = utf8.length(); |
| 307 |
|
| 308 |
int position = firstNonWhitespacePosition(utf8); |
| 309 |
return utf8.slice(position, length - position); |
| 310 |
} |
| 311 |
|
| 312 |
private static int firstNonWhitespacePosition(Slice utf8) |
| 313 |
{ |
| 314 |
int length = utf8.length(); |
| 315 |
|
| 316 |
int position = 0; |
| 317 |
while (position < length) { |
| 318 |
int codePoint = tryGetCodePointAt(utf8, position); |
| 319 |
if (codePoint < 0) { |
| 320 |
break; |
| 321 |
} |
| 322 |
if (!WHITESPACE_CODE_POINTS[codePoint]) { |
| 323 |
break; |
| 324 |
} |
| 325 |
position += lengthOfCodePoint(codePoint); |
| 326 |
} |
| 327 |
return position; |
| 328 |
} |
| 329 |
|
| 330 |
/** |
| 331 |
* Removes all white space characters from the right side of the string. |
| 332 |
* <p> |
| 333 |
* Note: Invalid UTF-8 sequences are not trimmed. |
| 334 |
*/ |
| 335 |
public static Slice rightTrim(Slice utf8) |
| 336 |
{ |
| 337 |
int position = lastNonWhitespacePosition(utf8, 0); |
| 338 |
return utf8.slice(0, position); |
| 339 |
} |
| 340 |
|
| 341 |
private static int lastNonWhitespacePosition(Slice utf8, int minPosition) |
| 342 |
{ |
| 343 |
int length = utf8.length(); |
| 344 |
|
| 345 |
int position = length; |
| 346 |
while (minPosition < position) { |
| 347 |
// decode the code point before position if possible |
| 348 |
int codePoint; |
| 349 |
byte unsignedByte = utf8.getByte(position - 1); |
| 350 |
if (!isContinuationByte(unsignedByte)) { |
| 351 |
codePoint = unsignedByte & 0xFF; |
| 352 |
} |
| 353 |
else if (minPosition <= position -2 && !isContinuationByte(utf8.getByte(position - 2))) { |
| 354 |
codePoint = tryGetCodePointAt(utf8, position - 2); |
| 355 |
} |
| 356 |
else if (minPosition <= position -3 && !isContinuationByte(utf8.getByte(position - 3))) { |
| 357 |
codePoint = tryGetCodePointAt(utf8, position - 3); |
| 358 |
} |
| 359 |
else if (minPosition <= position -4 && !isContinuationByte(utf8.getByte(position - 4))) { |
| 360 |
codePoint = tryGetCodePointAt(utf8, position - 4); |
| 361 |
} |
| 362 |
else { |
| 363 |
break; |
| 364 |
} |
| 365 |
|
| 366 |
if (codePoint < 0 || !WHITESPACE_CODE_POINTS[codePoint]) { |
| 367 |
break; |
| 368 |
} |
| 369 |
position -= lengthOfCodePoint(codePoint); |
| 370 |
} |
| 371 |
return position; |
| 372 |
} |
| 373 |
|
| 374 |
/** |
| 375 |
* Removes all white space characters from the left and right side of the string. |
| 376 |
* <p> |
| 377 |
* Note: Invalid UTF-8 sequences are not trimmed. |
| 378 |
*/ |
| 379 |
public static Slice trim(Slice utf8) |
| 380 |
{ |
| 381 |
int start = firstNonWhitespacePosition(utf8); |
| 382 |
int end = lastNonWhitespacePosition(utf8, start); |
| 383 |
return utf8.slice(start, end - start); |
| 384 |
} |
| 385 |
|
| 386 |
public static Slice fixInvalidUtf8(Slice slice) |
| 387 |
{ |
| 388 |
return fixInvalidUtf8(slice, OptionalInt.of(REPLACEMENT_CODE_POINT)); |
| 389 |
} |
| 390 |
|
| 391 |
public static Slice fixInvalidUtf8(Slice slice, OptionalInt replacementCodePoint) |
| 392 |
{ |
| 393 |
if (isAscii(slice)) { |
| 394 |
return slice; |
| 395 |
} |
| 396 |
|
| 397 |
int replacementCodePointValue = -1; |
| 398 |
int replacementCodePointLength = 0; |
| 399 |
if (replacementCodePoint.isPresent()) { |
| 400 |
replacementCodePointValue = replacementCodePoint.getAsInt(); |
| 401 |
replacementCodePointLength = lengthOfCodePoint(replacementCodePointValue); |
| 402 |
} |
| 403 |
|
| 404 |
int length = slice.length(); |
| 405 |
Slice utf8 = Slices.allocate(length); |
| 406 |
|
| 407 |
int dataPosition = 0; |
| 408 |
int utf8Position = 0; |
| 409 |
while (dataPosition < length) { |
| 410 |
int codePoint = tryGetCodePointAt(slice, dataPosition); |
| 411 |
int codePointLength; |
| 412 |
if (codePoint >= 0) { |
| 413 |
codePointLength = lengthOfCodePoint(codePoint); |
| 414 |
dataPosition += codePointLength; |
| 415 |
} |
| 416 |
else { |
| 417 |
// negative number carries the number of invalid bytes |
| 418 |
dataPosition += (-codePoint); |
| 419 |
if (replacementCodePointValue < 0) { |
| 420 |
continue; |
| 421 |
} |
| 422 |
codePoint = replacementCodePointValue; |
| 423 |
codePointLength = replacementCodePointLength; |
| 424 |
} |
| 425 |
utf8 = Slices.ensureSize(utf8, utf8Position + codePointLength); |
| 426 |
utf8Position += setCodePointAt(codePoint, utf8, utf8Position); |
| 427 |
} |
| 428 |
return utf8.slice(0, utf8Position); |
| 429 |
} |
| 430 |
|
| 431 |
/** |
| 432 |
* Tries to get the UTF-8 encoded code point at the {@code position}. A positive |
| 433 |
* return value means the UTF-8 sequence at the position is valid, and the result |
| 434 |
* is the code point. A negative return value means the UTF-8 sequence at the |
| 435 |
* position is invalid, and the length of the invalid sequence is the absolute |
| 436 |
* value of the result. |
| 437 |
* @return the code point or negative the number of bytes in the invalid UTF-8 sequence. |
| 438 |
*/ |
| 439 |
public static int tryGetCodePointAt(Slice utf8, int position) |
| 440 |
{ |
| 441 |
// |
| 442 |
// Process first byte |
| 443 |
byte firstByte = utf8.getByte(position); |
| 444 |
|
| 445 |
int length = lengthOfCodePointFromStartByteSafe(firstByte); |
| 446 |
if (length < 0) { |
| 447 |
return length; |
| 448 |
} |
| 449 |
|
| 450 |
if (length == 1) { |
| 451 |
// normal ASCII |
| 452 |
// 0xxx_xxxx |
| 453 |
return firstByte; |
| 454 |
} |
| 455 |
|
| 456 |
// |
| 457 |
// Process second byte |
| 458 |
if (position + 1 >= utf8.length()) { |
| 459 |
return -1; |
| 460 |
} |
| 461 |
|
| 462 |
byte secondByte = utf8.getByteUnchecked(position + 1); |
| 463 |
if (!isContinuationByte(secondByte)) { |
| 464 |
return -1; |
| 465 |
} |
| 466 |
|
| 467 |
if (length == 2) { |
| 468 |
// 110x_xxxx 10xx_xxxx |
| 469 |
return ((firstByte & 0b0001_1111) << 6) | |
| 470 |
(secondByte & 0b0011_1111); |
| 471 |
} |
| 472 |
|
| 473 |
// |
| 474 |
// Process third byte |
| 475 |
if (position + 2 >= utf8.length()) { |
| 476 |
return -2; |
| 477 |
} |
| 478 |
|
| 479 |
byte thirdByte = utf8.getByteUnchecked(position + 2); |
| 480 |
if (!isContinuationByte(thirdByte)) { |
| 481 |
return -2; |
| 482 |
} |
| 483 |
|
| 484 |
if (length == 3) { |
| 485 |
// 1110_xxxx 10xx_xxxx 10xx_xxxx |
| 486 |
int codePoint = ((firstByte & 0b0000_1111) << 12) | |
| 487 |
((secondByte & 0b0011_1111) << 6) | |
| 488 |
(thirdByte & 0b0011_1111); |
| 489 |
|
| 490 |
// surrogates are invalid |
| 491 |
if (MIN_SURROGATE <= codePoint && codePoint <= MAX_SURROGATE) { |
| 492 |
return -3; |
| 493 |
} |
| 494 |
return codePoint; |
| 495 |
} |
| 496 |
|
| 497 |
// |
| 498 |
// Process forth byte |
| 499 |
if (position + 3 >= utf8.length()) { |
| 500 |
return -3; |
| 501 |
} |
| 502 |
|
| 503 |
byte forthByte = utf8.getByteUnchecked(position + 3); |
| 504 |
if (!isContinuationByte(forthByte)) { |
| 505 |
return -3; |
| 506 |
} |
| 507 |
|
| 508 |
if (length == 4) { |
| 509 |
// 1111_0xxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 510 |
int codePoint = ((firstByte & 0b0000_0111) << 18) | |
| 511 |
((secondByte & 0b0011_1111) << 12) | |
| 512 |
((thirdByte & 0b0011_1111) << 6) | |
| 513 |
(forthByte & 0b0011_1111); |
| 514 |
// 4 byte code points have a limited valid range |
| 515 |
if (codePoint < 0x11_0000) { |
| 516 |
return codePoint; |
| 517 |
} |
| 518 |
return -4; |
| 519 |
} |
| 520 |
|
| 521 |
// |
| 522 |
// Process fifth byte |
| 523 |
if (position + 4 >= utf8.length()) { |
| 524 |
return -4; |
| 525 |
} |
| 526 |
|
| 527 |
byte fifthByte = utf8.getByteUnchecked(position + 4); |
| 528 |
if (!isContinuationByte(fifthByte)) { |
| 529 |
return -4; |
| 530 |
} |
| 531 |
|
| 532 |
if (length == 5) { |
| 533 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 534 |
return -5; |
| 535 |
} |
| 536 |
|
| 537 |
// |
| 538 |
// Process sixth byte |
| 539 |
if (position + 5 >= utf8.length()) { |
| 540 |
return -5; |
| 541 |
} |
| 542 |
|
| 543 |
byte sixthByte = utf8.getByteUnchecked(position + 5); |
| 544 |
if (!isContinuationByte(sixthByte)) { |
| 545 |
return -5; |
| 546 |
} |
| 547 |
|
| 548 |
if (length == 6) { |
| 549 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 550 |
return -6; |
| 551 |
} |
| 552 |
|
| 553 |
// for longer sequence, which can't happen |
| 554 |
return -1; |
| 555 |
} |
| 556 |
|
| 557 |
static int lengthOfCodePointFromStartByteSafe(byte startByte) |
| 558 |
{ |
| 559 |
int unsignedStartByte = startByte & 0xFF; |
| 560 |
if (unsignedStartByte < 0b1000_0000) { |
| 561 |
// normal ASCII |
| 562 |
// 0xxx_xxxx |
| 563 |
return 1; |
| 564 |
} |
| 565 |
if (unsignedStartByte < 0b1100_0000) { |
| 566 |
// illegal bytes |
| 567 |
// 10xx_xxxx |
| 568 |
return -1; |
| 569 |
} |
| 570 |
if (unsignedStartByte < 0b1110_0000) { |
| 571 |
// 110x_xxxx 10xx_xxxx |
| 572 |
return 2; |
| 573 |
} |
| 574 |
if (unsignedStartByte < 0b1111_0000) { |
| 575 |
// 1110_xxxx 10xx_xxxx 10xx_xxxx |
| 576 |
return 3; |
| 577 |
} |
| 578 |
if (unsignedStartByte < 0b1111_1000) { |
| 579 |
// 1111_0xxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 580 |
return 4; |
| 581 |
} |
| 582 |
if (unsignedStartByte < 0b1111_1100) { |
| 583 |
// 1111_10xx 10xx_xxxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 584 |
return 5; |
| 585 |
} |
| 586 |
if (unsignedStartByte < 0b1111_1110) { |
| 587 |
// 1111_110x 10xx_xxxx 10xx_xxxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 588 |
return 6; |
| 589 |
} |
| 590 |
return -1; |
| 591 |
} |
| 592 |
|
| 593 |
/** |
| 594 |
* Finds the index of the first byte of the code point at a position, or |
| 595 |
* {@code -1} if the position is not withing the slice. |
| 596 |
* <p> |
| 597 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 598 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 599 |
*/ |
| 600 |
public static int offsetOfCodePoint(Slice utf8, int codePointCount) |
| 601 |
{ |
| 602 |
return offsetOfCodePoint(utf8, 0, codePointCount); |
| 603 |
} |
| 604 |
|
| 605 |
/** |
| 606 |
* Starting from {@code position} bytes in {@code utf8}, finds the |
| 607 |
* index of the first byte of the code point {@code codePointCount} |
| 608 |
* in the slice. If the slice does not contain |
| 609 |
* {@code codePointCount} code points after {@code position}, {@code -1} |
| 610 |
* is returned. |
| 611 |
* <p> |
| 612 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 613 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 614 |
*/ |
| 615 |
public static int offsetOfCodePoint(Slice utf8, int position, int codePointCount) |
| 616 |
{ |
| 617 |
checkPositionIndex(position, utf8.length()); |
| 618 |
checkArgument(codePointCount >= 0, "codePointPosition is negative"); |
| 619 |
|
| 620 |
// Quick exit if we are sure that the position is after the end |
| 621 |
if (utf8.length() - position <= codePointCount) { |
| 622 |
return -1; |
| 623 |
} |
| 624 |
if (codePointCount == 0) { |
| 625 |
return position; |
| 626 |
} |
| 627 |
|
| 628 |
int correctIndex = codePointCount + position; |
| 629 |
// Length rounded to 8 bytes |
| 630 |
int length8 = utf8.length() & 0x7FFF_FFF8; |
| 631 |
// While we have enough bytes left and we need at least 8 characters process 8 bytes at once |
| 632 |
while (position < length8 && correctIndex >= position + 8) { |
| 633 |
// Count bytes which are NOT the start of a code point |
| 634 |
correctIndex += countContinuationBytes(utf8.getLongUnchecked(position)); |
| 635 |
|
| 636 |
position += 8; |
| 637 |
} |
| 638 |
// Length rounded to 4 bytes |
| 639 |
int length4 = utf8.length() & 0x7FFF_FFFC; |
| 640 |
// While we have enough bytes left and we need at least 4 characters process 4 bytes at once |
| 641 |
while (position < length4 && correctIndex >= position + 4) { |
| 642 |
// Count bytes which are NOT the start of a code point |
| 643 |
correctIndex += countContinuationBytes(utf8.getIntUnchecked(position)); |
| 644 |
|
| 645 |
position += 4; |
| 646 |
} |
| 647 |
// Do the rest one by one, always check the last byte to find the end of the code point |
| 648 |
while (position < utf8.length()) { |
| 649 |
// Count bytes which are NOT the start of a code point |
| 650 |
correctIndex += countContinuationBytes(utf8.getByteUnchecked(position)); |
| 651 |
if (position == correctIndex) { |
| 652 |
break; |
| 653 |
} |
| 654 |
|
| 655 |
position++; |
| 656 |
} |
| 657 |
|
| 658 |
if (position == correctIndex && correctIndex < utf8.length()) { |
| 659 |
return correctIndex; |
| 660 |
} |
| 661 |
return -1; |
| 662 |
} |
| 663 |
|
| 664 |
/** |
| 665 |
* Gets the UTF-8 sequence length of the code point at {@code position}. |
| 666 |
* <p> |
| 667 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 668 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 669 |
*/ |
| 670 |
public static int lengthOfCodePoint(Slice utf8, int position) |
| 671 |
{ |
| 672 |
return lengthOfCodePointFromStartByte(utf8.getByte(position)); |
| 673 |
} |
| 674 |
|
| 675 |
/** |
| 676 |
* Gets the UTF-8 sequence length of the code point at {@code position}. |
| 677 |
* <p> |
| 678 |
* Truncated UTF-8 sequences, 5 and 6 byte sequences, and invalid code points |
| 679 |
* are handled by this method without throwing an exception. |
| 680 |
*/ |
| 681 |
public static int lengthOfCodePointSafe(Slice utf8, int position) |
| 682 |
{ |
| 683 |
int length = lengthOfCodePointFromStartByteSafe(utf8.getByte(position)); |
| 684 |
if (length < 0) { |
| 685 |
return -length; |
| 686 |
} |
| 687 |
|
| 688 |
if (length == 1 || position + 1 >= utf8.length() || !isContinuationByte(utf8.getByteUnchecked(position + 1))) { |
| 689 |
return 1; |
| 690 |
} |
| 691 |
|
| 692 |
if (length == 2 || position + 2 >= utf8.length() || !isContinuationByte(utf8.getByteUnchecked(position + 2))) { |
| 693 |
return 2; |
| 694 |
} |
| 695 |
|
| 696 |
if (length == 3 || position + 3 >= utf8.length() || !isContinuationByte(utf8.getByteUnchecked(position + 3))) { |
| 697 |
return 3; |
| 698 |
} |
| 699 |
|
| 700 |
if (length == 4 || position + 4 >= utf8.length() || !isContinuationByte(utf8.getByteUnchecked(position + 4))) { |
| 701 |
return 4; |
| 702 |
} |
| 703 |
|
| 704 |
if (length == 5 || position + 5 >= utf8.length() || !isContinuationByte(utf8.getByteUnchecked(position + 5))) { |
| 705 |
return 5; |
| 706 |
} |
| 707 |
|
| 708 |
if (length == 6) { |
| 709 |
return 6; |
| 710 |
} |
| 711 |
|
| 712 |
return 1; |
| 713 |
} |
| 714 |
|
| 715 |
/** |
| 716 |
* Gets the UTF-8 sequence length of the code point. |
| 717 |
* |
| 718 |
* @throws InvalidCodePointException if code point is not within a valid range |
| 719 |
*/ |
| 720 |
public static int lengthOfCodePoint(int codePoint) |
| 721 |
{ |
| 722 |
if (codePoint < 0) { |
| 723 |
throw new InvalidCodePointException(codePoint); |
| 724 |
} |
| 725 |
if (codePoint < 0x80) { |
| 726 |
// normal ASCII |
| 727 |
// 0xxx_xxxx |
| 728 |
return 1; |
| 729 |
} |
| 730 |
if (codePoint < 0x800) { |
| 731 |
return 2; |
| 732 |
} |
| 733 |
if (codePoint < 0x1_0000) { |
| 734 |
return 3; |
| 735 |
} |
| 736 |
if (codePoint < 0x11_0000) { |
| 737 |
return 4; |
| 738 |
} |
| 739 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 740 |
throw new InvalidCodePointException(codePoint); |
| 741 |
} |
| 742 |
|
| 743 |
/** |
| 744 |
* Gets the UTF-8 sequence length using the sequence start byte. |
| 745 |
* <p> |
| 746 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 747 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 748 |
*/ |
| 749 |
public static int lengthOfCodePointFromStartByte(byte startByte) |
| 750 |
{ |
| 751 |
int unsignedStartByte = startByte & 0xFF; |
| 752 |
if (unsignedStartByte < 0x80) { |
| 753 |
// normal ASCII |
| 754 |
// 0xxx_xxxx |
| 755 |
return 1; |
| 756 |
} |
| 757 |
if (unsignedStartByte < 0xc0) { |
| 758 |
// illegal bytes |
| 759 |
// 10xx_xxxx |
| 760 |
throw new InvalidUtf8Exception("Illegal start 0x" + toHexString(unsignedStartByte).toUpperCase() + " of code point"); |
| 761 |
} |
| 762 |
if (unsignedStartByte < 0xe0) { |
| 763 |
// 110x_xxxx 10xx_xxxx |
| 764 |
return 2; |
| 765 |
} |
| 766 |
if (unsignedStartByte < 0xf0) { |
| 767 |
// 1110_xxxx 10xx_xxxx 10xx_xxxx |
| 768 |
return 3; |
| 769 |
} |
| 770 |
if (unsignedStartByte < 0xf8) { |
| 771 |
// 1111_0xxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 772 |
return 4; |
| 773 |
} |
| 774 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 775 |
throw new InvalidUtf8Exception("Illegal start 0x" + toHexString(unsignedStartByte).toUpperCase() + " of code point"); |
| 776 |
} |
| 777 |
|
| 778 |
/** |
| 779 |
* Gets the UTF-8 encoded code point at the {@code position}. |
| 780 |
* <p> |
| 781 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 782 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 783 |
*/ |
| 784 |
public static int getCodePointAt(Slice utf8, int position) |
| 785 |
{ |
| 786 |
int unsignedStartByte = utf8.getByte(position) & 0xFF; |
| 787 |
if (unsignedStartByte < 0x80) { |
| 788 |
// normal ASCII |
| 789 |
// 0xxx_xxxx |
| 790 |
return unsignedStartByte; |
| 791 |
} |
| 792 |
if (unsignedStartByte < 0xc0) { |
| 793 |
// illegal bytes |
| 794 |
// 10xx_xxxx |
| 795 |
throw new InvalidUtf8Exception("Illegal start 0x" + toHexString(unsignedStartByte).toUpperCase() + " of code point"); |
| 796 |
} |
| 797 |
if (unsignedStartByte < 0xe0) { |
| 798 |
// 110x_xxxx 10xx_xxxx |
| 799 |
if (position + 1 >= utf8.length()) { |
| 800 |
throw new InvalidUtf8Exception("UTF-8 sequence truncated"); |
| 801 |
} |
| 802 |
return ((unsignedStartByte & 0b0001_1111) << 6) | |
| 803 |
(utf8.getByte(position + 1) & 0b0011_1111); |
| 804 |
} |
| 805 |
if (unsignedStartByte < 0xf0) { |
| 806 |
// 1110_xxxx 10xx_xxxx 10xx_xxxx |
| 807 |
if (position + 2 >= utf8.length()) { |
| 808 |
throw new InvalidUtf8Exception("UTF-8 sequence truncated"); |
| 809 |
} |
| 810 |
return ((unsignedStartByte & 0b0000_1111) << 12) | |
| 811 |
((utf8.getByteUnchecked(position + 1) & 0b0011_1111) << 6) | |
| 812 |
(utf8.getByteUnchecked(position + 2) & 0b0011_1111); |
| 813 |
} |
| 814 |
if (unsignedStartByte < 0xf8) { |
| 815 |
// 1111_0xxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 816 |
if (position + 3 >= utf8.length()) { |
| 817 |
throw new InvalidUtf8Exception("UTF-8 sequence truncated"); |
| 818 |
} |
| 819 |
return ((unsignedStartByte & 0b0000_0111) << 18) | |
| 820 |
((utf8.getByteUnchecked(position + 1) & 0b0011_1111) << 12) | |
| 821 |
((utf8.getByteUnchecked(position + 2) & 0b0011_1111) << 6) | |
| 822 |
(utf8.getByteUnchecked(position + 3) & 0b0011_1111); |
| 823 |
} |
| 824 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 825 |
throw new InvalidUtf8Exception("Illegal start 0x" + toHexString(unsignedStartByte).toUpperCase() + " of code point"); |
| 826 |
} |
| 827 |
|
| 828 |
/** |
| 829 |
* Gets the UTF-8 encoded code point before the {@code position}. |
| 830 |
* <p> |
| 831 |
* Note: This method does not explicitly check for valid UTF-8, and may |
| 832 |
* return incorrect results or throw an exception for invalid UTF-8. |
| 833 |
*/ |
| 834 |
public static int getCodePointBefore(Slice utf8, int position) |
| 835 |
{ |
| 836 |
byte unsignedByte = utf8.getByte(position - 1); |
| 837 |
if (!isContinuationByte(unsignedByte)) { |
| 838 |
return unsignedByte & 0xFF; |
| 839 |
} |
| 840 |
if (!isContinuationByte(utf8.getByte(position - 2))) { |
| 841 |
return getCodePointAt(utf8, position - 2); |
| 842 |
} |
| 843 |
if (!isContinuationByte(utf8.getByte(position - 3))) { |
| 844 |
return getCodePointAt(utf8, position - 3); |
| 845 |
} |
| 846 |
if (!isContinuationByte(utf8.getByte(position - 4))) { |
| 847 |
return getCodePointAt(utf8, position - 4); |
| 848 |
} |
| 849 |
|
| 850 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 851 |
throw new InvalidUtf8Exception("UTF-8 is not well formed"); |
| 852 |
} |
| 853 |
|
| 854 |
private static boolean isContinuationByte(byte b) |
| 855 |
{ |
| 856 |
return (b & 0b1100_0000) == 0b1000_0000; |
| 857 |
} |
| 858 |
|
| 859 |
/** |
| 860 |
* Convert the code point to UTF-8. |
| 861 |
* <p> |
| 862 |
* |
| 863 |
* @throws InvalidCodePointException if code point is not within a valid range |
| 864 |
*/ |
| 865 |
public static Slice codePointToUtf8(int codePoint) |
| 866 |
{ |
| 867 |
Slice utf8 = Slices.allocate(lengthOfCodePoint(codePoint)); |
| 868 |
setCodePointAt(codePoint, utf8, 0); |
| 869 |
return utf8; |
| 870 |
} |
| 871 |
|
| 872 |
/** |
| 873 |
* Sets the UTF-8 sequence for code point at the {@code position}. |
| 874 |
* |
| 875 |
* @throws InvalidCodePointException if code point is not within a valid range |
| 876 |
*/ |
| 877 |
public static int setCodePointAt(int codePoint, Slice utf8, int position) |
| 878 |
{ |
| 879 |
if (codePoint < 0) { |
| 880 |
throw new InvalidCodePointException(codePoint); |
| 881 |
} |
| 882 |
if (codePoint < 0x80) { |
| 883 |
// normal ASCII |
| 884 |
// 0xxx_xxxx |
| 885 |
utf8.setByte(position, codePoint); |
| 886 |
return 1; |
| 887 |
} |
| 888 |
if (codePoint < 0x800) { |
| 889 |
// 110x_xxxx 10xx_xxxx |
| 890 |
utf8.setByte(position, 0b1100_0000 | (codePoint >>> 6)); |
| 891 |
utf8.setByte(position + 1, 0b1000_0000 | (codePoint & 0b0011_1111)); |
| 892 |
return 2; |
| 893 |
} |
| 894 |
if (MIN_SURROGATE <= codePoint && codePoint <= MAX_SURROGATE) { |
| 895 |
throw new InvalidCodePointException(codePoint); |
| 896 |
} |
| 897 |
if (codePoint < 0x1_0000) { |
| 898 |
// 1110_xxxx 10xx_xxxx 10xx_xxxx |
| 899 |
utf8.setByte(position, 0b1110_0000 | ((codePoint >>> 12) & 0b0000_1111)); |
| 900 |
utf8.setByte(position + 1, 0b1000_0000 | ((codePoint >>> 6) & 0b0011_1111)); |
| 901 |
utf8.setByte(position + 2, 0b1000_0000 | (codePoint & 0b0011_1111)); |
| 902 |
return 3; |
| 903 |
} |
| 904 |
if (codePoint < 0x11_0000) { |
| 905 |
// 1111_0xxx 10xx_xxxx 10xx_xxxx 10xx_xxxx |
| 906 |
utf8.setByte(position, 0b1111_0000 | ((codePoint >>> 18) & 0b0000_0111)); |
| 907 |
utf8.setByte(position + 1, 0b1000_0000 | ((codePoint >>> 12) & 0b0011_1111)); |
| 908 |
utf8.setByte(position + 2, 0b1000_0000 | ((codePoint >>> 6) & 0b0011_1111)); |
| 909 |
utf8.setByte(position + 3, 0b1000_0000 | (codePoint & 0b0011_1111)); |
| 910 |
return 4; |
| 911 |
} |
| 912 |
// Per RFC3629, UTF-8 is limited to 4 bytes, so more bytes are illegal |
| 913 |
throw new InvalidCodePointException(codePoint); |
| 914 |
} |
| 915 |
|
| 916 |
private static int countContinuationBytes(byte i8) |
| 917 |
{ |
| 918 |
// see below |
| 919 |
int value = i8 & 0xff; |
| 920 |
return (value >>> 7) & (~value >>> 6); |
| 921 |
} |
| 922 |
|
| 923 |
private static int countContinuationBytes(int i32) |
| 924 |
{ |
| 925 |
// see below |
| 926 |
i32 = ((i32 & TOP_MASK32) >>> 1) & (~i32); |
| 927 |
return Integer.bitCount(i32); |
| 928 |
} |
| 929 |
|
| 930 |
private static int countContinuationBytes(long i64) |
| 931 |
{ |
| 932 |
// Count the number of bytes that match 0b10xx_xxxx as follows: |
| 933 |
// 1. Mask off the 8th bit of every byte and shift it into the 7th position. |
| 934 |
// 2. Then invert the bytes, which turns the 0 in the 7th bit to a one. |
| 935 |
// 3. And together the restults of step 1 and 2, giving us a one in the 7th |
| 936 |
// position if the byte matched. |
| 937 |
// 4. Count the number of bits in the result, which is the number of bytes |
| 938 |
// that matched. |
| 939 |
i64 = ((i64 & TOP_MASK64) >>> 1) & (~i64); |
| 940 |
return Long.bitCount(i64); |
| 941 |
} |
| 942 |
} |