|
1 /* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */ |
|
2 /* vim: set ts=2 et sw=2 tw=80: */ |
|
3 /* This Source Code Form is subject to the terms of the Mozilla Public |
|
4 * License, v. 2.0. If a copy of the MPL was not distributed with this |
|
5 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
|
6 |
|
7 #include "mozilla/FloatingPoint.h" |
|
8 |
|
9 #include "Key.h" |
|
10 #include "ReportInternalError.h" |
|
11 |
|
12 #include "jsfriendapi.h" |
|
13 #include "nsAlgorithm.h" |
|
14 #include "nsJSUtils.h" |
|
15 #include "xpcpublic.h" |
|
16 #include "mozilla/Endian.h" |
|
17 #include <algorithm> |
|
18 |
|
19 USING_INDEXEDDB_NAMESPACE |
|
20 |
|
21 /* |
|
22 Here's how we encode keys: |
|
23 |
|
24 Basic strategy is the following |
|
25 |
|
26 Numbers: 1 n n n n n n n n ("n"s are encoded 64bit float) |
|
27 Dates: 2 n n n n n n n n ("n"s are encoded 64bit float) |
|
28 Strings: 3 s s s ... 0 ("s"s are encoded unicode bytes) |
|
29 Arrays: 4 i i i ... 0 ("i"s are encoded array items) |
|
30 |
|
31 |
|
32 When encoding floats, 64bit IEEE 754 are almost sortable, except that |
|
33 positive sort lower than negative, and negative sort descending. So we use |
|
34 the following encoding: |
|
35 |
|
36 value < 0 ? |
|
37 (-to64bitInt(value)) : |
|
38 (to64bitInt(value) | 0x8000000000000000) |
|
39 |
|
40 |
|
41 When encoding strings, we use variable-size encoding per the following table |
|
42 |
|
43 Chars 0 - 7E are encoded as 0xxxxxxx with 1 added |
|
44 Chars 7F - (3FFF+7F) are encoded as 10xxxxxx xxxxxxxx with 7F subtracted |
|
45 Chars (3FFF+80) - FFFF are encoded as 11xxxxxx xxxxxxxx xx000000 |
|
46 |
|
47 This ensures that the first byte is never encoded as 0, which means that the |
|
48 string terminator (per basic-stategy table) sorts before any character. |
|
49 The reason that (3FFF+80) - FFFF is encoded "shifted up" 6 bits is to maximize |
|
50 the chance that the last character is 0. See below for why. |
|
51 |
|
52 |
|
53 When encoding Arrays, we use an additional trick. Rather than adding a byte |
|
54 containing the value '4' to indicate type, we instead add 4 to the next byte. |
|
55 This is usually the byte containing the type of the first item in the array. |
|
56 So simple examples are |
|
57 |
|
58 ["foo"] 7 s s s 0 0 // 7 is 3 + 4 |
|
59 [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 |
|
60 |
|
61 Whe do this iteratively if the first item in the array is also an array |
|
62 |
|
63 [["foo"]] 11 s s s 0 0 0 |
|
64 |
|
65 However, to avoid overflow in the byte, we only do this 3 times. If the first |
|
66 item in an array is an array, and that array also has an array as first item, |
|
67 we simply write out the total value accumulated so far and then follow the |
|
68 "normal" rules. |
|
69 |
|
70 [[["foo"]]] 12 3 s s s 0 0 0 0 |
|
71 |
|
72 There is another edge case that can happen though, which is that the array |
|
73 doesn't have a first item to which we can add 4 to the type. Instead the |
|
74 next byte would normally be the array terminator (per basic-strategy table) |
|
75 so we simply add the 4 there. |
|
76 |
|
77 [[]] 8 0 // 8 is 4 + 4 + 0 |
|
78 [] 4 // 4 is 4 + 0 |
|
79 [[], "foo"] 8 3 s s s 0 0 // 8 is 4 + 4 + 0 |
|
80 |
|
81 Note that the max-3-times rule kicks in before we get a chance to add to the |
|
82 array terminator |
|
83 |
|
84 [[[]]] 12 0 0 0 // 12 is 4 + 4 + 4 |
|
85 |
|
86 We could use a much higher number than 3 at no complexity or performance cost, |
|
87 however it seems unlikely that it'll make a practical difference, and the low |
|
88 limit makes testing eaiser. |
|
89 |
|
90 |
|
91 As a final optimization we do a post-encoding step which drops all 0s at the |
|
92 end of the encoded buffer. |
|
93 |
|
94 "foo" // 3 s s s |
|
95 1 // 1 bf f0 |
|
96 ["a", "b"] // 7 s 3 s |
|
97 [1, 2] // 5 bf f0 0 0 0 0 0 0 1 c0 |
|
98 [[]] // 8 |
|
99 */ |
|
100 |
|
101 const int MaxArrayCollapse = 3; |
|
102 |
|
103 const int MaxRecursionDepth = 256; |
|
104 |
|
105 nsresult |
|
106 Key::EncodeJSValInternal(JSContext* aCx, JS::Handle<JS::Value> aVal, |
|
107 uint8_t aTypeOffset, uint16_t aRecursionDepth) |
|
108 { |
|
109 NS_ENSURE_TRUE(aRecursionDepth < MaxRecursionDepth, NS_ERROR_DOM_INDEXEDDB_DATA_ERR); |
|
110 |
|
111 static_assert(eMaxType * MaxArrayCollapse < 256, |
|
112 "Unable to encode jsvals."); |
|
113 |
|
114 if (aVal.isString()) { |
|
115 nsDependentJSString str; |
|
116 if (!str.init(aCx, aVal)) { |
|
117 IDB_REPORT_INTERNAL_ERR(); |
|
118 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
119 } |
|
120 EncodeString(str, aTypeOffset); |
|
121 return NS_OK; |
|
122 } |
|
123 |
|
124 if (aVal.isNumber()) { |
|
125 double d = aVal.toNumber(); |
|
126 if (mozilla::IsNaN(d)) { |
|
127 return NS_ERROR_DOM_INDEXEDDB_DATA_ERR; |
|
128 } |
|
129 EncodeNumber(d, eFloat + aTypeOffset); |
|
130 return NS_OK; |
|
131 } |
|
132 |
|
133 if (aVal.isObject()) { |
|
134 JS::Rooted<JSObject*> obj(aCx, &aVal.toObject()); |
|
135 if (JS_IsArrayObject(aCx, obj)) { |
|
136 aTypeOffset += eMaxType; |
|
137 |
|
138 if (aTypeOffset == eMaxType * MaxArrayCollapse) { |
|
139 mBuffer.Append(aTypeOffset); |
|
140 aTypeOffset = 0; |
|
141 } |
|
142 NS_ASSERTION((aTypeOffset % eMaxType) == 0 && |
|
143 aTypeOffset < (eMaxType * MaxArrayCollapse), |
|
144 "Wrong typeoffset"); |
|
145 |
|
146 uint32_t length; |
|
147 if (!JS_GetArrayLength(aCx, obj, &length)) { |
|
148 IDB_REPORT_INTERNAL_ERR(); |
|
149 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
150 } |
|
151 |
|
152 for (uint32_t index = 0; index < length; index++) { |
|
153 JS::Rooted<JS::Value> val(aCx); |
|
154 if (!JS_GetElement(aCx, obj, index, &val)) { |
|
155 IDB_REPORT_INTERNAL_ERR(); |
|
156 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
157 } |
|
158 |
|
159 nsresult rv = EncodeJSValInternal(aCx, val, aTypeOffset, |
|
160 aRecursionDepth + 1); |
|
161 if (NS_FAILED(rv)) { |
|
162 return rv; |
|
163 } |
|
164 |
|
165 aTypeOffset = 0; |
|
166 } |
|
167 |
|
168 mBuffer.Append(eTerminator + aTypeOffset); |
|
169 |
|
170 return NS_OK; |
|
171 } |
|
172 |
|
173 if (JS_ObjectIsDate(aCx, obj)) { |
|
174 if (!js_DateIsValid(obj)) { |
|
175 return NS_ERROR_DOM_INDEXEDDB_DATA_ERR; |
|
176 } |
|
177 EncodeNumber(js_DateGetMsecSinceEpoch(obj), eDate + aTypeOffset); |
|
178 return NS_OK; |
|
179 } |
|
180 } |
|
181 |
|
182 return NS_ERROR_DOM_INDEXEDDB_DATA_ERR; |
|
183 } |
|
184 |
|
185 // static |
|
186 nsresult |
|
187 Key::DecodeJSValInternal(const unsigned char*& aPos, const unsigned char* aEnd, |
|
188 JSContext* aCx, uint8_t aTypeOffset, JS::MutableHandle<JS::Value> aVal, |
|
189 uint16_t aRecursionDepth) |
|
190 { |
|
191 NS_ENSURE_TRUE(aRecursionDepth < MaxRecursionDepth, NS_ERROR_DOM_INDEXEDDB_DATA_ERR); |
|
192 |
|
193 if (*aPos - aTypeOffset >= eArray) { |
|
194 JS::Rooted<JSObject*> array(aCx, JS_NewArrayObject(aCx, 0)); |
|
195 if (!array) { |
|
196 NS_WARNING("Failed to make array!"); |
|
197 IDB_REPORT_INTERNAL_ERR(); |
|
198 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
199 } |
|
200 |
|
201 aTypeOffset += eMaxType; |
|
202 |
|
203 if (aTypeOffset == eMaxType * MaxArrayCollapse) { |
|
204 ++aPos; |
|
205 aTypeOffset = 0; |
|
206 } |
|
207 |
|
208 uint32_t index = 0; |
|
209 JS::Rooted<JS::Value> val(aCx); |
|
210 while (aPos < aEnd && *aPos - aTypeOffset != eTerminator) { |
|
211 nsresult rv = DecodeJSValInternal(aPos, aEnd, aCx, aTypeOffset, |
|
212 &val, aRecursionDepth + 1); |
|
213 NS_ENSURE_SUCCESS(rv, rv); |
|
214 |
|
215 aTypeOffset = 0; |
|
216 |
|
217 if (!JS_SetElement(aCx, array, index++, val)) { |
|
218 NS_WARNING("Failed to set array element!"); |
|
219 IDB_REPORT_INTERNAL_ERR(); |
|
220 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
221 } |
|
222 } |
|
223 |
|
224 NS_ASSERTION(aPos >= aEnd || (*aPos % eMaxType) == eTerminator, |
|
225 "Should have found end-of-array marker"); |
|
226 ++aPos; |
|
227 |
|
228 aVal.setObject(*array); |
|
229 } |
|
230 else if (*aPos - aTypeOffset == eString) { |
|
231 nsString key; |
|
232 DecodeString(aPos, aEnd, key); |
|
233 if (!xpc::StringToJsval(aCx, key, aVal)) { |
|
234 IDB_REPORT_INTERNAL_ERR(); |
|
235 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
236 } |
|
237 } |
|
238 else if (*aPos - aTypeOffset == eDate) { |
|
239 double msec = static_cast<double>(DecodeNumber(aPos, aEnd)); |
|
240 JSObject* date = JS_NewDateObjectMsec(aCx, msec); |
|
241 if (!date) { |
|
242 IDB_WARNING("Failed to make date!"); |
|
243 return NS_ERROR_DOM_INDEXEDDB_UNKNOWN_ERR; |
|
244 } |
|
245 |
|
246 aVal.setObject(*date); |
|
247 } |
|
248 else if (*aPos - aTypeOffset == eFloat) { |
|
249 aVal.setDouble(DecodeNumber(aPos, aEnd)); |
|
250 } |
|
251 else { |
|
252 NS_NOTREACHED("Unknown key type!"); |
|
253 } |
|
254 |
|
255 return NS_OK; |
|
256 } |
|
257 |
|
258 #define ONE_BYTE_LIMIT 0x7E |
|
259 #define TWO_BYTE_LIMIT (0x3FFF+0x7F) |
|
260 |
|
261 #define ONE_BYTE_ADJUST 1 |
|
262 #define TWO_BYTE_ADJUST (-0x7F) |
|
263 #define THREE_BYTE_SHIFT 6 |
|
264 |
|
265 void |
|
266 Key::EncodeString(const nsAString& aString, uint8_t aTypeOffset) |
|
267 { |
|
268 // First measure how long the encoded string will be. |
|
269 |
|
270 // The +2 is for initial 3 and trailing 0. We'll compensate for multi-byte |
|
271 // chars below. |
|
272 uint32_t size = aString.Length() + 2; |
|
273 |
|
274 const char16_t* start = aString.BeginReading(); |
|
275 const char16_t* end = aString.EndReading(); |
|
276 for (const char16_t* iter = start; iter < end; ++iter) { |
|
277 if (*iter > ONE_BYTE_LIMIT) { |
|
278 size += *iter > TWO_BYTE_LIMIT ? 2 : 1; |
|
279 } |
|
280 } |
|
281 |
|
282 // Allocate memory for the new size |
|
283 uint32_t oldLen = mBuffer.Length(); |
|
284 char* buffer; |
|
285 if (!mBuffer.GetMutableData(&buffer, oldLen + size)) { |
|
286 return; |
|
287 } |
|
288 buffer += oldLen; |
|
289 |
|
290 // Write type marker |
|
291 *(buffer++) = eString + aTypeOffset; |
|
292 |
|
293 // Encode string |
|
294 for (const char16_t* iter = start; iter < end; ++iter) { |
|
295 if (*iter <= ONE_BYTE_LIMIT) { |
|
296 *(buffer++) = *iter + ONE_BYTE_ADJUST; |
|
297 } |
|
298 else if (*iter <= TWO_BYTE_LIMIT) { |
|
299 char16_t c = char16_t(*iter) + TWO_BYTE_ADJUST + 0x8000; |
|
300 *(buffer++) = (char)(c >> 8); |
|
301 *(buffer++) = (char)(c & 0xFF); |
|
302 } |
|
303 else { |
|
304 uint32_t c = (uint32_t(*iter) << THREE_BYTE_SHIFT) | 0x00C00000; |
|
305 *(buffer++) = (char)(c >> 16); |
|
306 *(buffer++) = (char)(c >> 8); |
|
307 *(buffer++) = (char)c; |
|
308 } |
|
309 } |
|
310 |
|
311 // Write end marker |
|
312 *(buffer++) = eTerminator; |
|
313 |
|
314 NS_ASSERTION(buffer == mBuffer.EndReading(), "Wrote wrong number of bytes"); |
|
315 } |
|
316 |
|
317 // static |
|
318 void |
|
319 Key::DecodeString(const unsigned char*& aPos, const unsigned char* aEnd, |
|
320 nsString& aString) |
|
321 { |
|
322 NS_ASSERTION(*aPos % eMaxType == eString, "Don't call me!"); |
|
323 |
|
324 const unsigned char* buffer = aPos + 1; |
|
325 |
|
326 // First measure how big the decoded string will be. |
|
327 uint32_t size = 0; |
|
328 const unsigned char* iter; |
|
329 for (iter = buffer; iter < aEnd && *iter != eTerminator; ++iter) { |
|
330 if (*iter & 0x80) { |
|
331 iter += (*iter & 0x40) ? 2 : 1; |
|
332 } |
|
333 ++size; |
|
334 } |
|
335 |
|
336 // Set end so that we don't have to check for null termination in the loop |
|
337 // below |
|
338 if (iter < aEnd) { |
|
339 aEnd = iter; |
|
340 } |
|
341 |
|
342 char16_t* out; |
|
343 if (size && !aString.GetMutableData(&out, size)) { |
|
344 return; |
|
345 } |
|
346 |
|
347 for (iter = buffer; iter < aEnd;) { |
|
348 if (!(*iter & 0x80)) { |
|
349 *out = *(iter++) - ONE_BYTE_ADJUST; |
|
350 } |
|
351 else if (!(*iter & 0x40)) { |
|
352 char16_t c = (char16_t(*(iter++)) << 8); |
|
353 if (iter < aEnd) { |
|
354 c |= *(iter++); |
|
355 } |
|
356 *out = c - TWO_BYTE_ADJUST - 0x8000; |
|
357 } |
|
358 else { |
|
359 uint32_t c = uint32_t(*(iter++)) << (16 - THREE_BYTE_SHIFT); |
|
360 if (iter < aEnd) { |
|
361 c |= uint32_t(*(iter++)) << (8 - THREE_BYTE_SHIFT); |
|
362 } |
|
363 if (iter < aEnd) { |
|
364 c |= *(iter++) >> THREE_BYTE_SHIFT; |
|
365 } |
|
366 *out = (char16_t)c; |
|
367 } |
|
368 |
|
369 ++out; |
|
370 } |
|
371 |
|
372 NS_ASSERTION(!size || out == aString.EndReading(), |
|
373 "Should have written the whole string"); |
|
374 |
|
375 aPos = iter + 1; |
|
376 } |
|
377 |
|
378 union Float64Union { |
|
379 double d; |
|
380 uint64_t u; |
|
381 }; |
|
382 |
|
383 void |
|
384 Key::EncodeNumber(double aFloat, uint8_t aType) |
|
385 { |
|
386 // Allocate memory for the new size |
|
387 uint32_t oldLen = mBuffer.Length(); |
|
388 char* buffer; |
|
389 if (!mBuffer.GetMutableData(&buffer, oldLen + 1 + sizeof(double))) { |
|
390 return; |
|
391 } |
|
392 buffer += oldLen; |
|
393 |
|
394 *(buffer++) = aType; |
|
395 |
|
396 Float64Union pun; |
|
397 pun.d = aFloat; |
|
398 // Note: The subtraction from 0 below is necessary to fix |
|
399 // MSVC build warning C4146 (negating an unsigned value). |
|
400 uint64_t number = pun.u & PR_UINT64(0x8000000000000000) ? |
|
401 (0 - pun.u) : |
|
402 (pun.u | PR_UINT64(0x8000000000000000)); |
|
403 |
|
404 mozilla::BigEndian::writeUint64(buffer, number); |
|
405 } |
|
406 |
|
407 // static |
|
408 double |
|
409 Key::DecodeNumber(const unsigned char*& aPos, const unsigned char* aEnd) |
|
410 { |
|
411 NS_ASSERTION(*aPos % eMaxType == eFloat || |
|
412 *aPos % eMaxType == eDate, "Don't call me!"); |
|
413 |
|
414 ++aPos; |
|
415 |
|
416 uint64_t number = 0; |
|
417 memcpy(&number, aPos, std::min<size_t>(sizeof(number), aEnd - aPos)); |
|
418 number = mozilla::NativeEndian::swapFromBigEndian(number); |
|
419 |
|
420 aPos += sizeof(number); |
|
421 |
|
422 Float64Union pun; |
|
423 // Note: The subtraction from 0 below is necessary to fix |
|
424 // MSVC build warning C4146 (negating an unsigned value). |
|
425 pun.u = number & PR_UINT64(0x8000000000000000) ? |
|
426 (number & ~PR_UINT64(0x8000000000000000)) : |
|
427 (0 - number); |
|
428 |
|
429 return pun.d; |
|
430 } |