AuroraMiddleware/scuffed-code
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
4bd1ee008c
scuffed-code/icu4c/source/common/ucptrie.cpp
Markus Scherer 98f9170004 ICU-20250 make UnicodeSet(intprop=value) faster 
- fastpath for UnicodeSet.add(new last range)
- fewer UnicodeSet memory allocations:
  initial internal list array, exponential array growth,
  allocate strings list/set only when first one is added
- faster CodePointTrie.getRange(): fewer calls to filter function
- revert UnicodeSet(intprop=value) from trie ranges to range starts + lookup
- cache per-int-prop range starts: fewer lookups
2018-11-12 16:05:57 -08:00

602 lines
22 KiB
C++
Raw Blame History

// © 2017 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
// ucptrie.cpp (modified from utrie2.cpp)
// created: 2017dec29 Markus W. Scherer
// #define UCPTRIE_DEBUG
#ifdef UCPTRIE_DEBUG
# include <stdio.h>
#endif
#include "unicode/utypes.h"
#include "unicode/ucptrie.h"
#include "unicode/utf.h"
#include "unicode/utf8.h"
#include "unicode/utf16.h"
#include "cmemory.h"
#include "uassert.h"
#include "ucptrie_impl.h"
U_CAPI UCPTrie * U_EXPORT2
ucptrie_openFromBinary(UCPTrieType type, UCPTrieValueWidth valueWidth,
const void *data, int32_t length, int32_t *pActualLength,
UErrorCode *pErrorCode) {
if (U_FAILURE(*pErrorCode)) {
return nullptr;
}
if (length <= 0 || (U_POINTER_MASK_LSB(data, 3) != 0) ||
type < UCPTRIE_TYPE_ANY || UCPTRIE_TYPE_SMALL < type ||
valueWidth < UCPTRIE_VALUE_BITS_ANY || UCPTRIE_VALUE_BITS_8 < valueWidth) {
*pErrorCode = U_ILLEGAL_ARGUMENT_ERROR;
return nullptr;
}
// Enough data for a trie header?
if (length < (int32_t)sizeof(UCPTrieHeader)) {
*pErrorCode = U_INVALID_FORMAT_ERROR;
return nullptr;
}
// Check the signature.
const UCPTrieHeader *header = (const UCPTrieHeader *)data;
if (header->signature != UCPTRIE_SIG) {
*pErrorCode = U_INVALID_FORMAT_ERROR;
return nullptr;
}
int32_t options = header->options;
int32_t typeInt = (options >> 6) & 3;
int32_t valueWidthInt = options & UCPTRIE_OPTIONS_VALUE_BITS_MASK;
if (typeInt > UCPTRIE_TYPE_SMALL || valueWidthInt > UCPTRIE_VALUE_BITS_8 ||
(options & UCPTRIE_OPTIONS_RESERVED_MASK) != 0) {
*pErrorCode = U_INVALID_FORMAT_ERROR;
return nullptr;
}
UCPTrieType actualType = (UCPTrieType)typeInt;
UCPTrieValueWidth actualValueWidth = (UCPTrieValueWidth)valueWidthInt;
if (type < 0) {
type = actualType;
}
if (valueWidth < 0) {
valueWidth = actualValueWidth;
}
if (type != actualType || valueWidth != actualValueWidth) {
*pErrorCode = U_INVALID_FORMAT_ERROR;
return nullptr;
}
// Get the length values and offsets.
UCPTrie tempTrie;
uprv_memset(&tempTrie, 0, sizeof(tempTrie));
tempTrie.indexLength = header->indexLength;
tempTrie.dataLength =
((options & UCPTRIE_OPTIONS_DATA_LENGTH_MASK) << 4) | header->dataLength;
tempTrie.index3NullOffset = header->index3NullOffset;
tempTrie.dataNullOffset =
((options & UCPTRIE_OPTIONS_DATA_NULL_OFFSET_MASK) << 8) | header->dataNullOffset;
tempTrie.highStart = header->shiftedHighStart << UCPTRIE_SHIFT_2;
tempTrie.shifted12HighStart = (tempTrie.highStart + 0xfff) >> 12;
tempTrie.type = type;
tempTrie.valueWidth = valueWidth;
// Calculate the actual length.
int32_t actualLength = (int32_t)sizeof(UCPTrieHeader) + tempTrie.indexLength * 2;
if (valueWidth == UCPTRIE_VALUE_BITS_16) {
actualLength += tempTrie.dataLength * 2;
} else if (valueWidth == UCPTRIE_VALUE_BITS_32) {
actualLength += tempTrie.dataLength * 4;
} else {
actualLength += tempTrie.dataLength;
}
if (length < actualLength) {
*pErrorCode = U_INVALID_FORMAT_ERROR; // Not enough bytes.
return nullptr;
}
// Allocate the trie.
UCPTrie *trie = (UCPTrie *)uprv_malloc(sizeof(UCPTrie));
if (trie == nullptr) {
*pErrorCode = U_MEMORY_ALLOCATION_ERROR;
return nullptr;
}
uprv_memcpy(trie, &tempTrie, sizeof(tempTrie));
#ifdef UCPTRIE_DEBUG
trie->name = "fromSerialized";
#endif
// Set the pointers to its index and data arrays.
const uint16_t *p16 = (const uint16_t *)(header + 1);
trie->index = p16;
p16 += trie->indexLength;
// Get the data.
int32_t nullValueOffset = trie->dataNullOffset;
if (nullValueOffset >= trie->dataLength) {
nullValueOffset = trie->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET;
}
switch (valueWidth) {
case UCPTRIE_VALUE_BITS_16:
trie->data.ptr16 = p16;
trie->nullValue = trie->data.ptr16[nullValueOffset];
break;
case UCPTRIE_VALUE_BITS_32:
trie->data.ptr32 = (const uint32_t *)p16;
trie->nullValue = trie->data.ptr32[nullValueOffset];
break;
case UCPTRIE_VALUE_BITS_8:
trie->data.ptr8 = (const uint8_t *)p16;
trie->nullValue = trie->data.ptr8[nullValueOffset];
break;
default:
// Unreachable because valueWidth was checked above.
*pErrorCode = U_INVALID_FORMAT_ERROR;
return nullptr;
}
if (pActualLength != nullptr) {
*pActualLength = actualLength;
}
return trie;
}
U_CAPI void U_EXPORT2
ucptrie_close(UCPTrie *trie) {
uprv_free(trie);
}
U_CAPI UCPTrieType U_EXPORT2
ucptrie_getType(const UCPTrie *trie) {
return (UCPTrieType)trie->type;
}
U_CAPI UCPTrieValueWidth U_EXPORT2
ucptrie_getValueWidth(const UCPTrie *trie) {
return (UCPTrieValueWidth)trie->valueWidth;
}
U_CAPI int32_t U_EXPORT2
ucptrie_internalSmallIndex(const UCPTrie *trie, UChar32 c) {
int32_t i1 = c >> UCPTRIE_SHIFT_1;
if (trie->type == UCPTRIE_TYPE_FAST) {
U_ASSERT(0xffff < c && c < trie->highStart);
i1 += UCPTRIE_BMP_INDEX_LENGTH - UCPTRIE_OMITTED_BMP_INDEX_1_LENGTH;
} else {
U_ASSERT((uint32_t)c < (uint32_t)trie->highStart && trie->highStart > UCPTRIE_SMALL_LIMIT);
i1 += UCPTRIE_SMALL_INDEX_LENGTH;
}
int32_t i3Block = trie->index[
(int32_t)trie->index[i1] + ((c >> UCPTRIE_SHIFT_2) & UCPTRIE_INDEX_2_MASK)];
int32_t i3 = (c >> UCPTRIE_SHIFT_3) & UCPTRIE_INDEX_3_MASK;
int32_t dataBlock;
if ((i3Block & 0x8000) == 0) {
// 16-bit indexes
dataBlock = trie->index[i3Block + i3];
} else {
// 18-bit indexes stored in groups of 9 entries per 8 indexes.
i3Block = (i3Block & 0x7fff) + (i3 & ~7) + (i3 >> 3);
i3 &= 7;
dataBlock = ((int32_t)trie->index[i3Block++] << (2 + (2 * i3))) & 0x30000;
dataBlock |= trie->index[i3Block + i3];
}
return dataBlock + (c & UCPTRIE_SMALL_DATA_MASK);
}
U_CAPI int32_t U_EXPORT2
ucptrie_internalSmallU8Index(const UCPTrie *trie, int32_t lt1, uint8_t t2, uint8_t t3) {
UChar32 c = (lt1 << 12) | (t2 << 6) | t3;
if (c >= trie->highStart) {
// Possible because the UTF-8 macro compares with shifted12HighStart which may be higher.
return trie->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET;
}
return ucptrie_internalSmallIndex(trie, c);
}
U_CAPI int32_t U_EXPORT2
ucptrie_internalU8PrevIndex(const UCPTrie *trie, UChar32 c,
const uint8_t *start, const uint8_t *src) {
int32_t i, length;
// Support 64-bit pointers by avoiding cast of arbitrary difference.
if ((src - start) <= 7) {
i = length = (int32_t)(src - start);
} else {
i = length = 7;
start = src - 7;
}
c = utf8_prevCharSafeBody(start, 0, &i, c, -1);
i = length - i; // Number of bytes read backward from src.
int32_t idx = _UCPTRIE_CP_INDEX(trie, 0xffff, c);
return (idx << 3) | i;
}
namespace {
inline uint32_t getValue(UCPTrieData data, UCPTrieValueWidth valueWidth, int32_t dataIndex) {
switch (valueWidth) {
case UCPTRIE_VALUE_BITS_16:
return data.ptr16[dataIndex];
case UCPTRIE_VALUE_BITS_32:
return data.ptr32[dataIndex];
case UCPTRIE_VALUE_BITS_8:
return data.ptr8[dataIndex];
default:
// Unreachable if the trie is properly initialized.
return 0xffffffff;
}
}
} // namespace
U_CAPI uint32_t U_EXPORT2
ucptrie_get(const UCPTrie *trie, UChar32 c) {
int32_t dataIndex;
if ((uint32_t)c <= 0x7f) {
// linear ASCII
dataIndex = c;
} else {
UChar32 fastMax = trie->type == UCPTRIE_TYPE_FAST ? 0xffff : UCPTRIE_SMALL_MAX;
dataIndex = _UCPTRIE_CP_INDEX(trie, fastMax, c);
}
return getValue(trie->data, (UCPTrieValueWidth)trie->valueWidth, dataIndex);
}
namespace {
constexpr int32_t MAX_UNICODE = 0x10ffff;
inline uint32_t maybeFilterValue(uint32_t value, uint32_t trieNullValue, uint32_t nullValue,
UCPMapValueFilter *filter, const void *context) {
if (value == trieNullValue) {
value = nullValue;
} else if (filter != nullptr) {
value = filter(context, value);
}
return value;
}
UChar32 getRange(const void *t, UChar32 start,
UCPMapValueFilter *filter, const void *context, uint32_t *pValue) {
if ((uint32_t)start > MAX_UNICODE) {
return U_SENTINEL;
}
const UCPTrie *trie = reinterpret_cast<const UCPTrie *>(t);
UCPTrieValueWidth valueWidth = (UCPTrieValueWidth)trie->valueWidth;
if (start >= trie->highStart) {
if (pValue != nullptr) {
int32_t di = trie->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET;
uint32_t value = getValue(trie->data, valueWidth, di);
if (filter != nullptr) { value = filter(context, value); }
*pValue = value;
}
return MAX_UNICODE;
}
uint32_t nullValue = trie->nullValue;
if (filter != nullptr) { nullValue = filter(context, nullValue); }
const uint16_t *index = trie->index;
int32_t prevI3Block = -1;
int32_t prevBlock = -1;
UChar32 c = start;
uint32_t trieValue, value;
bool haveValue = false;
do {
int32_t i3Block;
int32_t i3;
int32_t i3BlockLength;
int32_t dataBlockLength;
if (c <= 0xffff && (trie->type == UCPTRIE_TYPE_FAST || c <= UCPTRIE_SMALL_MAX)) {
i3Block = 0;
i3 = c >> UCPTRIE_FAST_SHIFT;
i3BlockLength = trie->type == UCPTRIE_TYPE_FAST ?
UCPTRIE_BMP_INDEX_LENGTH : UCPTRIE_SMALL_INDEX_LENGTH;
dataBlockLength = UCPTRIE_FAST_DATA_BLOCK_LENGTH;
} else {
// Use the multi-stage index.
int32_t i1 = c >> UCPTRIE_SHIFT_1;
if (trie->type == UCPTRIE_TYPE_FAST) {
U_ASSERT(0xffff < c && c < trie->highStart);
i1 += UCPTRIE_BMP_INDEX_LENGTH - UCPTRIE_OMITTED_BMP_INDEX_1_LENGTH;
} else {
U_ASSERT(c < trie->highStart && trie->highStart > UCPTRIE_SMALL_LIMIT);
i1 += UCPTRIE_SMALL_INDEX_LENGTH;
}
i3Block = trie->index[
(int32_t)trie->index[i1] + ((c >> UCPTRIE_SHIFT_2) & UCPTRIE_INDEX_2_MASK)];
if (i3Block == prevI3Block && (c - start) >= UCPTRIE_CP_PER_INDEX_2_ENTRY) {
// The index-3 block is the same as the previous one, and filled with value.
U_ASSERT((c & (UCPTRIE_CP_PER_INDEX_2_ENTRY - 1)) == 0);
c += UCPTRIE_CP_PER_INDEX_2_ENTRY;
continue;
}
prevI3Block = i3Block;
if (i3Block == trie->index3NullOffset) {
// This is the index-3 null block.
if (haveValue) {
if (nullValue != value) {
return c - 1;
}
} else {
trieValue = trie->nullValue;
value = nullValue;
if (pValue != nullptr) { *pValue = nullValue; }
haveValue = true;
}
prevBlock = trie->dataNullOffset;
c = (c + UCPTRIE_CP_PER_INDEX_2_ENTRY) & ~(UCPTRIE_CP_PER_INDEX_2_ENTRY - 1);
continue;
}
i3 = (c >> UCPTRIE_SHIFT_3) & UCPTRIE_INDEX_3_MASK;
i3BlockLength = UCPTRIE_INDEX_3_BLOCK_LENGTH;
dataBlockLength = UCPTRIE_SMALL_DATA_BLOCK_LENGTH;
}
// Enumerate data blocks for one index-3 block.
do {
int32_t block;
if ((i3Block & 0x8000) == 0) {
block = index[i3Block + i3];
} else {
// 18-bit indexes stored in groups of 9 entries per 8 indexes.
int32_t group = (i3Block & 0x7fff) + (i3 & ~7) + (i3 >> 3);
int32_t gi = i3 & 7;
block = ((int32_t)index[group++] << (2 + (2 * gi))) & 0x30000;
block |= index[group + gi];
}
if (block == prevBlock && (c - start) >= dataBlockLength) {
// The block is the same as the previous one, and filled with value.
U_ASSERT((c & (dataBlockLength - 1)) == 0);
c += dataBlockLength;
} else {
int32_t dataMask = dataBlockLength - 1;
prevBlock = block;
if (block == trie->dataNullOffset) {
// This is the data null block.
if (haveValue) {
if (nullValue != value) {
return c - 1;
}
} else {
trieValue = trie->nullValue;
value = nullValue;
if (pValue != nullptr) { *pValue = nullValue; }
haveValue = true;
}
c = (c + dataBlockLength) & ~dataMask;
} else {
int32_t di = block + (c & dataMask);
uint32_t trieValue2 = getValue(trie->data, valueWidth, di);
if (haveValue) {
if (trieValue2 != trieValue) {
if (filter == nullptr ||
maybeFilterValue(trieValue2, trie->nullValue, nullValue,
filter, context) != value) {
return c - 1;
}
trieValue = trieValue2; // may or may not help
}
} else {
trieValue = trieValue2;
value = maybeFilterValue(trieValue2, trie->nullValue, nullValue,
filter, context);
if (pValue != nullptr) { *pValue = value; }
haveValue = true;
}
while ((++c & dataMask) != 0) {
trieValue2 = getValue(trie->data, valueWidth, ++di);
if (trieValue2 != trieValue) {
if (filter == nullptr ||
maybeFilterValue(trieValue2, trie->nullValue, nullValue,
filter, context) != value) {
return c - 1;
}
trieValue = trieValue2; // may or may not help
}
}
}
}
} while (++i3 < i3BlockLength);
} while (c < trie->highStart);
U_ASSERT(haveValue);
int32_t di = trie->dataLength - UCPTRIE_HIGH_VALUE_NEG_DATA_OFFSET;
uint32_t highValue = getValue(trie->data, valueWidth, di);
if (maybeFilterValue(highValue, trie->nullValue, nullValue,
filter, context) != value) {
return c - 1;
} else {
return MAX_UNICODE;
}
}
} // namespace
U_CFUNC UChar32
ucptrie_internalGetRange(UCPTrieGetRange *getRange,
const void *trie, UChar32 start,
UCPMapRangeOption option, uint32_t surrogateValue,
UCPMapValueFilter *filter, const void *context, uint32_t *pValue) {
if (option == UCPMAP_RANGE_NORMAL) {
return getRange(trie, start, filter, context, pValue);
}
uint32_t value;
if (pValue == nullptr) {
// We need to examine the range value even if the caller does not want it.
pValue = &value;
}
UChar32 surrEnd = option == UCPMAP_RANGE_FIXED_ALL_SURROGATES ? 0xdfff : 0xdbff;
UChar32 end = getRange(trie, start, filter, context, pValue);
if (end < 0xd7ff || start > surrEnd) {
return end;
}
// The range overlaps with surrogates, or ends just before the first one.
if (*pValue == surrogateValue) {
if (end >= surrEnd) {
// Surrogates followed by a non-surrogateValue range,
// or surrogates are part of a larger surrogateValue range.
return end;
}
} else {
if (start <= 0xd7ff) {
return 0xd7ff; // Non-surrogateValue range ends before surrogateValue surrogates.
}
// Start is a surrogate with a non-surrogateValue code *unit* value.
// Return a surrogateValue code *point* range.
*pValue = surrogateValue;
if (end > surrEnd) {
return surrEnd; // Surrogate range ends before non-surrogateValue rest of range.
}
}
// See if the surrogateValue surrogate range can be merged with
// an immediately following range.
uint32_t value2;
UChar32 end2 = getRange(trie, surrEnd + 1, filter, context, &value2);
if (value2 == surrogateValue) {
return end2;
}
return surrEnd;
}
U_CAPI UChar32 U_EXPORT2
ucptrie_getRange(const UCPTrie *trie, UChar32 start,
UCPMapRangeOption option, uint32_t surrogateValue,
UCPMapValueFilter *filter, const void *context, uint32_t *pValue) {
return ucptrie_internalGetRange(getRange, trie, start,
option, surrogateValue,
filter, context, pValue);
}
U_CAPI int32_t U_EXPORT2
ucptrie_toBinary(const UCPTrie *trie,
void *data, int32_t capacity,
UErrorCode *pErrorCode) {
if (U_FAILURE(*pErrorCode)) {
return 0;
}
UCPTrieType type = (UCPTrieType)trie->type;
UCPTrieValueWidth valueWidth = (UCPTrieValueWidth)trie->valueWidth;
if (type < UCPTRIE_TYPE_FAST || UCPTRIE_TYPE_SMALL < type ||
valueWidth < UCPTRIE_VALUE_BITS_16 || UCPTRIE_VALUE_BITS_8 < valueWidth ||
capacity < 0 ||
(capacity > 0 && (data == nullptr || (U_POINTER_MASK_LSB(data, 3) != 0)))) {
*pErrorCode = U_ILLEGAL_ARGUMENT_ERROR;
return 0;
}
int32_t length = (int32_t)sizeof(UCPTrieHeader) + trie->indexLength * 2;
switch (valueWidth) {
case UCPTRIE_VALUE_BITS_16:
length += trie->dataLength * 2;
break;
case UCPTRIE_VALUE_BITS_32:
length += trie->dataLength * 4;
break;
case UCPTRIE_VALUE_BITS_8:
length += trie->dataLength;
break;
default:
// unreachable
break;
}
if (capacity < length) {
*pErrorCode = U_BUFFER_OVERFLOW_ERROR;
return length;
}
char *bytes = (char *)data;
UCPTrieHeader *header = (UCPTrieHeader *)bytes;
header->signature = UCPTRIE_SIG; // "Tri3"
header->options = (uint16_t)(
((trie->dataLength & 0xf0000) >> 4) |
((trie->dataNullOffset & 0xf0000) >> 8) |
(trie->type << 6) |
valueWidth);
header->indexLength = (uint16_t)trie->indexLength;
header->dataLength = (uint16_t)trie->dataLength;
header->index3NullOffset = trie->index3NullOffset;
header->dataNullOffset = (uint16_t)trie->dataNullOffset;
header->shiftedHighStart = trie->highStart >> UCPTRIE_SHIFT_2;
bytes += sizeof(UCPTrieHeader);
uprv_memcpy(bytes, trie->index, trie->indexLength * 2);
bytes += trie->indexLength * 2;
switch (valueWidth) {
case UCPTRIE_VALUE_BITS_16:
uprv_memcpy(bytes, trie->data.ptr16, trie->dataLength * 2);
break;
case UCPTRIE_VALUE_BITS_32:
uprv_memcpy(bytes, trie->data.ptr32, trie->dataLength * 4);
break;
case UCPTRIE_VALUE_BITS_8:
uprv_memcpy(bytes, trie->data.ptr8, trie->dataLength);
break;
default:
// unreachable
break;
}
return length;
}
namespace {
#ifdef UCPTRIE_DEBUG
long countNull(const UCPTrie *trie) {
uint32_t nullValue=trie->nullValue;
int32_t length=trie->dataLength;
long count=0;
switch (trie->valueWidth) {
case UCPTRIE_VALUE_BITS_16:
for(int32_t i=0; i<length; ++i) {
if(trie->data.ptr16[i]==nullValue) { ++count; }
}
break;
case UCPTRIE_VALUE_BITS_32:
for(int32_t i=0; i<length; ++i) {
if(trie->data.ptr32[i]==nullValue) { ++count; }
}
break;
case UCPTRIE_VALUE_BITS_8:
for(int32_t i=0; i<length; ++i) {
if(trie->data.ptr8[i]==nullValue) { ++count; }
}
break;
default:
// unreachable
break;
}
return count;
}
U_CFUNC void
ucptrie_printLengths(const UCPTrie *trie, const char *which) {
long indexLength=trie->indexLength;
long dataLength=(long)trie->dataLength;
long totalLength=(long)sizeof(UCPTrieHeader)+indexLength*2+
dataLength*(trie->valueWidth==UCPTRIE_VALUE_BITS_16 ? 2 :
trie->valueWidth==UCPTRIE_VALUE_BITS_32 ? 4 : 1);
printf("**UCPTrieLengths(%s %s)** index:%6ld data:%6ld countNull:%6ld serialized:%6ld\n",
which, trie->name, indexLength, dataLength, countNull(trie), totalLength);
}
#endif
} // namespace
// UCPMap ----
// Initially, this is the same as UCPTrie. This may well change.
U_CAPI uint32_t U_EXPORT2
ucpmap_get(const UCPMap *map, UChar32 c) {
return ucptrie_get(reinterpret_cast<const UCPTrie *>(map), c);
}
U_CAPI UChar32 U_EXPORT2
ucpmap_getRange(const UCPMap *map, UChar32 start,
UCPMapRangeOption option, uint32_t surrogateValue,
UCPMapValueFilter *filter, const void *context, uint32_t *pValue) {
return ucptrie_getRange(reinterpret_cast<const UCPTrie *>(map), start,
option, surrogateValue,
filter, context, pValue);
}
Reference in New Issue View Git Blame Copy Permalink