DXR is a code search and navigation tool aimed at making sense of large projects. It supports full-text and regex searches as well as structural queries.

Header

Mercurial (31ec81b5d7bb)

VCS Links

Line Code
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

#include "mozilla/FloatingPoint.h"

#include "Key.h"
#include "jsfriendapi.h"
#include "nsAlgorithm.h"
#include "nsJSUtils.h"
#include "xpcpublic.h"
#include "mozilla/Endian.h"
#include <algorithm>

USING_INDEXEDDB_NAMESPACE

/*
 Here's how we encode keys:

 Basic strategy is the following

 Numbers: 1 n n n n n n n n    ("n"s are encoded 64bit float)
 Dates:   2 n n n n n n n n    ("n"s are encoded 64bit float)
 Strings: 3 s s s ... 0        ("s"s are encoded unicode bytes)
 Arrays:  4 i i i ... 0        ("i"s are encoded array items)


 When encoding floats, 64bit IEEE 754 are almost sortable, except that
 positive sort lower than negative, and negative sort descending. So we use
 the following encoding:
 
 value < 0 ?
   (-to64bitInt(value)) :
   (to64bitInt(value) | 0x8000000000000000)


 When encoding strings, we use variable-size encoding per the following table
 
 Chars 0         - 7E           are encoded as 0xxxxxxx with 1 added
 Chars 7F        - (3FFF+7F)    are encoded as 10xxxxxx xxxxxxxx with 7F subtracted
 Chars (3FFF+80) - FFFF         are encoded as 11xxxxxx xxxxxxxx xx000000

 This ensures that the first byte is never encoded as 0, which means that the
 string terminator (per basic-stategy table) sorts before any character.
 The reason that (3FFF+80) - FFFF is encoded "shifted up" 6 bits is to maximize
 the chance that the last character is 0. See below for why.


 When encoding Arrays, we use an additional trick. Rather than adding a byte
 containing the value '4' to indicate type, we instead add 4 to the next byte.
 This is usually the byte containing the type of the first item in the array.
 So simple examples are

 ["foo"]       7 s s s 0 0                              // 7 is 3 + 4
 [1, 2]        5 n n n n n n n n 1 n n n n n n n n 0    // 5 is 1 + 4

 Whe do this iteratively if the first item in the array is also an array

 [["foo"]]    11 s s s 0 0 0

 However, to avoid overflow in the byte, we only do this 3 times. If the first
 item in an array is an array, and that array also has an array as first item,
 we simply write out the total value accumulated so far and then follow the
 "normal" rules.

 [[["foo"]]]  12 3 s s s 0 0 0 0

 There is another edge case that can happen though, which is that the array
 doesn't have a first item to which we can add 4 to the type. Instead the
 next byte would normally be the array terminator (per basic-strategy table)
 so we simply add the 4 there.

 [[]]         8 0             // 8 is 4 + 4 + 0
 []           4               // 4 is 4 + 0
 [[], "foo"]  8 3 s s s 0 0   // 8 is 4 + 4 + 0

 Note that the max-3-times rule kicks in before we get a chance to add to the
 array terminator

 [[[]]]       12 0 0 0        // 12 is 4 + 4 + 4

 We could use a much higher number than 3 at no complexity or performance cost,
 however it seems unlikely that it'll make a practical difference, and the low
 limit makes testing eaiser.


 As a final optimization we do a post-encoding step which drops all 0s at the
 end of the encoded buffer.
 
 "foo"         // 3 s s s
 1             // 1 bf f0
 ["a", "b"]    // 7 s 3 s
 [1, 2]        // 5 bf f0 0 0 0 0 0 0 1 c0
 [[]]          // 8
*/

const int MaxArrayCollapse = 3;

const int MaxRecursionDepth = 256;

nsresult
Key::EncodeJSValInternal(JSContext* aCx, const jsval aVal,
                         uint8_t aTypeOffset, uint16_t aRecursionDepth)
{
  NS_ENSURE_TRUE(aRecursionDepth < MaxRecursionDepth, NS_ERROR_DOM_INDEXEDDB_DATA_ERR);

  static_assert(eMaxType * MaxArrayCollapse < 256,
                "Unable to encode jsvals.");

  if (JSVAL_IS_STRING(aVal)) {
    nsDependentJSString str;
    if (!str.init(aCx, aVal)) {
      return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
    }
    EncodeString(str, aTypeOffset);
    return NS_OK;
  }

  if (JSVAL_IS_INT(aVal)) {
    EncodeNumber((double)JSVAL_TO_INT(aVal), eFloat + aTypeOffset);
    return NS_OK;
  }

  if (JSVAL_IS_DOUBLE(aVal)) {
    double d = JSVAL_TO_DOUBLE(aVal);
    if (mozilla::IsNaN(d)) {
      return NS_ERROR_DOM_INDEXEDDB_DATA_ERR;
    }
    EncodeNumber(d, eFloat + aTypeOffset);
    return NS_OK;
  }

  if (!JSVAL_IS_PRIMITIVE(aVal)) {
    JS::Rooted<JSObject*> obj(aCx, JSVAL_TO_OBJECT(aVal));
    if (JS_IsArrayObject(aCx, obj)) {
      aTypeOffset += eMaxType;

      if (aTypeOffset == eMaxType * MaxArrayCollapse) {
        mBuffer.Append(aTypeOffset);
        aTypeOffset = 0;
      }
      NS_ASSERTION((aTypeOffset % eMaxType) == 0 &&
                   aTypeOffset < (eMaxType * MaxArrayCollapse),
                   "Wrong typeoffset");

      uint32_t length;
      if (!JS_GetArrayLength(aCx, obj, &length)) {
        return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
      }

      for (uint32_t index = 0; index < length; index++) {
        JS::Rooted<JS::Value> val(aCx);
        if (!JS_GetElement(aCx, obj, index, &val)) {
          return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
        }

        nsresult rv = EncodeJSValInternal(aCx, val, aTypeOffset,
                                          aRecursionDepth + 1);
        if (NS_FAILED(rv)) {
          return rv;
        }

        aTypeOffset = 0;
      }

      mBuffer.Append(eTerminator + aTypeOffset);

      return NS_OK;
    }

    if (JS_ObjectIsDate(aCx, obj)) {
      if (!js_DateIsValid(obj))  {
        return NS_ERROR_DOM_INDEXEDDB_DATA_ERR;
      }
      EncodeNumber(js_DateGetMsecSinceEpoch(obj), eDate + aTypeOffset);
      return NS_OK;
    }
  }

  return NS_ERROR_DOM_INDEXEDDB_DATA_ERR;
}

// static
nsresult
Key::DecodeJSValInternal(const unsigned char*& aPos, const unsigned char* aEnd,
                         JSContext* aCx, uint8_t aTypeOffset, JS::MutableHandle<JS::Value> aVal,
                         uint16_t aRecursionDepth)
{
  NS_ENSURE_TRUE(aRecursionDepth < MaxRecursionDepth, NS_ERROR_DOM_INDEXEDDB_DATA_ERR);

  if (*aPos - aTypeOffset >= eArray) {
    JS::Rooted<JSObject*> array(aCx, JS_NewArrayObject(aCx, 0, nullptr));
    if (!array) {
      NS_WARNING("Failed to make array!");
      return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
    }

    aTypeOffset += eMaxType;

    if (aTypeOffset == eMaxType * MaxArrayCollapse) {
      ++aPos;
      aTypeOffset = 0;
    }

    uint32_t index = 0;
    JS::Rooted<JS::Value> val(aCx);
    while (aPos < aEnd && *aPos - aTypeOffset != eTerminator) {
      nsresult rv = DecodeJSValInternal(aPos, aEnd, aCx, aTypeOffset,
                                        &val, aRecursionDepth + 1);
      NS_ENSURE_SUCCESS(rv, rv);

      aTypeOffset = 0;

      if (!JS_SetElement(aCx, array, index++, &val)) {
        NS_WARNING("Failed to set array element!");
        return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
      }
    }

    NS_ASSERTION(aPos >= aEnd || (*aPos % eMaxType) == eTerminator,
                 "Should have found end-of-array marker");
    ++aPos;

    aVal.setObject(*array);
  }
  else if (*aPos - aTypeOffset == eString) {
    nsString key;
    DecodeString(aPos, aEnd, key);
    if (!xpc::StringToJsval(aCx, key, aVal)) {
      return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
    }
  }
  else if (*aPos - aTypeOffset == eDate) {
    double msec = static_cast<double>(DecodeNumber(aPos, aEnd));
    JSObject* date = JS_NewDateObjectMsec(aCx, msec);
    if (!date) {
      NS_WARNING("Failed to make date!");
      return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR;
    }

    aVal.setObject(*date);
  }
  else if (*aPos - aTypeOffset == eFloat) {
    aVal.setDouble(DecodeNumber(aPos, aEnd));
  }
  else {
    NS_NOTREACHED("Unknown key type!");
  }

  return NS_OK;
}

#define ONE_BYTE_LIMIT 0x7E
#define TWO_BYTE_LIMIT (0x3FFF+0x7F)

#define ONE_BYTE_ADJUST 1
#define TWO_BYTE_ADJUST (-0x7F)
#define THREE_BYTE_SHIFT 6

void
Key::EncodeString(const nsAString& aString, uint8_t aTypeOffset)
{
  // First measure how long the encoded string will be.

  // The +2 is for initial 3 and trailing 0. We'll compensate for multi-byte
  // chars below.
  uint32_t size = aString.Length() + 2;
  
  const PRUnichar* start = aString.BeginReading();
  const PRUnichar* end = aString.EndReading();
  for (const PRUnichar* iter = start; iter < end; ++iter) {
    if (*iter > ONE_BYTE_LIMIT) {
      size += *iter > TWO_BYTE_LIMIT ? 2 : 1;
    }
  }

  // Allocate memory for the new size
  uint32_t oldLen = mBuffer.Length();
  char* buffer;
  if (!mBuffer.GetMutableData(&buffer, oldLen + size)) {
    return;
  }
  buffer += oldLen;

  // Write type marker
  *(buffer++) = eString + aTypeOffset;

  // Encode string
  for (const PRUnichar* iter = start; iter < end; ++iter) {
    if (*iter <= ONE_BYTE_LIMIT) {
      *(buffer++) = *iter + ONE_BYTE_ADJUST;
    }
    else if (*iter <= TWO_BYTE_LIMIT) {
      PRUnichar c = PRUnichar(*iter) + TWO_BYTE_ADJUST + 0x8000;
      *(buffer++) = (char)(c >> 8);
      *(buffer++) = (char)(c & 0xFF);
    }
    else {
      uint32_t c = (uint32_t(*iter) << THREE_BYTE_SHIFT) | 0x00C00000;
      *(buffer++) = (char)(c >> 16);
      *(buffer++) = (char)(c >> 8);
      *(buffer++) = (char)c;
    }
  }

  // Write end marker
  *(buffer++) = eTerminator;
  
  NS_ASSERTION(buffer == mBuffer.EndReading(), "Wrote wrong number of bytes");
}

// static
void
Key::DecodeString(const unsigned char*& aPos, const unsigned char* aEnd,
                  nsString& aString)
{
  NS_ASSERTION(*aPos % eMaxType == eString, "Don't call me!");

  const unsigned char* buffer = aPos + 1;

  // First measure how big the decoded string will be.
  uint32_t size = 0;
  const unsigned char* iter; 
  for (iter = buffer; iter < aEnd && *iter != eTerminator; ++iter) {
    if (*iter & 0x80) {
      iter += (*iter & 0x40) ? 2 : 1;
    }
    ++size;
  }
  
  // Set end so that we don't have to check for null termination in the loop
  // below
  if (iter < aEnd) {
    aEnd = iter;
  }

  PRUnichar* out;
  if (size && !aString.GetMutableData(&out, size)) {
    return;
  }

  for (iter = buffer; iter < aEnd;) {
    if (!(*iter & 0x80)) {
      *out = *(iter++) - ONE_BYTE_ADJUST;
    }
    else if (!(*iter & 0x40)) {
      PRUnichar c = (PRUnichar(*(iter++)) << 8);
      if (iter < aEnd) {
        c |= *(iter++);
      }
      *out = c - TWO_BYTE_ADJUST - 0x8000;
    }
    else {
      uint32_t c = uint32_t(*(iter++)) << (16 - THREE_BYTE_SHIFT);
      if (iter < aEnd) {
        c |= uint32_t(*(iter++)) << (8 - THREE_BYTE_SHIFT);
      }
      if (iter < aEnd) {
        c |= *(iter++) >> THREE_BYTE_SHIFT;
      }
      *out = (PRUnichar)c;
    }
    
    ++out;
  }
  
  NS_ASSERTION(!size || out == aString.EndReading(),
               "Should have written the whole string");
  
  aPos = iter + 1;
}

union Float64Union {
  double d;
  uint64_t u;
}; 

void
Key::EncodeNumber(double aFloat, uint8_t aType)
{
  // Allocate memory for the new size
  uint32_t oldLen = mBuffer.Length();
  char* buffer;
  if (!mBuffer.GetMutableData(&buffer, oldLen + 1 + sizeof(double))) {
    return;
  }
  buffer += oldLen;

  *(buffer++) = aType;

  Float64Union pun;
  pun.d = aFloat;
  // Note: The subtraction from 0 below is necessary to fix
  // MSVC build warning C4146 (negating an unsigned value).
  uint64_t number = pun.u & PR_UINT64(0x8000000000000000) ?
                    (0 - pun.u) :
                    (pun.u | PR_UINT64(0x8000000000000000));

  mozilla::BigEndian::writeUint64(buffer, number);
}

// static
double
Key::DecodeNumber(const unsigned char*& aPos, const unsigned char* aEnd)
{
  NS_ASSERTION(*aPos % eMaxType == eFloat ||
               *aPos % eMaxType == eDate, "Don't call me!");

  ++aPos;

  uint64_t number = 0;
  memcpy(&number, aPos, std::min<size_t>(sizeof(number), aEnd - aPos));
  number = mozilla::NativeEndian::swapFromBigEndian(number);

  aPos += sizeof(number);

  Float64Union pun;
  // Note: The subtraction from 0 below is necessary to fix
  // MSVC build warning C4146 (negating an unsigned value).
  pun.u = number & PR_UINT64(0x8000000000000000) ?
          (number & ~PR_UINT64(0x8000000000000000)) :
          (0 - number);

  return pun.d;
}