871 lines
27 KiB
C++
871 lines
27 KiB
C++
// © 2016 and later: Unicode, Inc. and others.
|
|
// License & terms of use: http://www.unicode.org/copyright.html
|
|
/*
|
|
*******************************************************************************
|
|
*
|
|
* Copyright (C) 2003-2016, International Business Machines
|
|
* Corporation and others. All Rights Reserved.
|
|
*
|
|
*******************************************************************************
|
|
* file name: usprep.cpp
|
|
* encoding: UTF-8
|
|
* tab size: 8 (not used)
|
|
* indentation:4
|
|
*
|
|
* created on: 2003jul2
|
|
* created by: Ram Viswanadha
|
|
*/
|
|
|
|
#include "unicode/utypes.h"
|
|
|
|
#if !UCONFIG_NO_IDNA
|
|
|
|
#include "unicode/usprep.h"
|
|
|
|
#include "unicode/normalizer2.h"
|
|
#include "unicode/ustring.h"
|
|
#include "unicode/uchar.h"
|
|
#include "unicode/uversion.h"
|
|
#include "umutex.h"
|
|
#include "cmemory.h"
|
|
#include "sprpimpl.h"
|
|
#include "ustr_imp.h"
|
|
#include "uhash.h"
|
|
#include "cstring.h"
|
|
#include "udataswp.h"
|
|
#include "ucln_cmn.h"
|
|
#include "ubidi_props.h"
|
|
#include "uprops.h"
|
|
|
|
U_NAMESPACE_USE
|
|
|
|
U_CDECL_BEGIN
|
|
|
|
/*
|
|
Static cache for already opened StringPrep profiles
|
|
*/
|
|
static UHashtable *SHARED_DATA_HASHTABLE = NULL;
|
|
static icu::UInitOnce gSharedDataInitOnce {};
|
|
|
|
static UMutex usprepMutex;
|
|
/* format version of spp file */
|
|
//static uint8_t formatVersion[4]={ 0, 0, 0, 0 };
|
|
|
|
/* the Unicode version of the sprep data */
|
|
static UVersionInfo dataVersion={ 0, 0, 0, 0 };
|
|
|
|
/* Profile names must be aligned to UStringPrepProfileType */
|
|
static const char * const PROFILE_NAMES[] = {
|
|
"rfc3491", /* USPREP_RFC3491_NAMEPREP */
|
|
"rfc3530cs", /* USPREP_RFC3530_NFS4_CS_PREP */
|
|
"rfc3530csci", /* USPREP_RFC3530_NFS4_CS_PREP_CI */
|
|
"rfc3491", /* USPREP_RFC3530_NSF4_CIS_PREP */
|
|
"rfc3530mixp", /* USPREP_RFC3530_NSF4_MIXED_PREP_PREFIX */
|
|
"rfc3491", /* USPREP_RFC3530_NSF4_MIXED_PREP_SUFFIX */
|
|
"rfc3722", /* USPREP_RFC3722_ISCSI */
|
|
"rfc3920node", /* USPREP_RFC3920_NODEPREP */
|
|
"rfc3920res", /* USPREP_RFC3920_RESOURCEPREP */
|
|
"rfc4011", /* USPREP_RFC4011_MIB */
|
|
"rfc4013", /* USPREP_RFC4013_SASLPREP */
|
|
"rfc4505", /* USPREP_RFC4505_TRACE */
|
|
"rfc4518", /* USPREP_RFC4518_LDAP */
|
|
"rfc4518ci", /* USPREP_RFC4518_LDAP_CI */
|
|
};
|
|
|
|
static UBool U_CALLCONV
|
|
isSPrepAcceptable(void * /* context */,
|
|
const char * /* type */,
|
|
const char * /* name */,
|
|
const UDataInfo *pInfo) {
|
|
if(
|
|
pInfo->size>=20 &&
|
|
pInfo->isBigEndian==U_IS_BIG_ENDIAN &&
|
|
pInfo->charsetFamily==U_CHARSET_FAMILY &&
|
|
pInfo->dataFormat[0]==0x53 && /* dataFormat="SPRP" */
|
|
pInfo->dataFormat[1]==0x50 &&
|
|
pInfo->dataFormat[2]==0x52 &&
|
|
pInfo->dataFormat[3]==0x50 &&
|
|
pInfo->formatVersion[0]==3 &&
|
|
pInfo->formatVersion[2]==UTRIE_SHIFT &&
|
|
pInfo->formatVersion[3]==UTRIE_INDEX_SHIFT
|
|
) {
|
|
//uprv_memcpy(formatVersion, pInfo->formatVersion, 4);
|
|
uprv_memcpy(dataVersion, pInfo->dataVersion, 4);
|
|
return true;
|
|
} else {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
static int32_t U_CALLCONV
|
|
getSPrepFoldingOffset(uint32_t data) {
|
|
|
|
return (int32_t)data;
|
|
|
|
}
|
|
|
|
/* hashes an entry */
|
|
static int32_t U_CALLCONV
|
|
hashEntry(const UHashTok parm) {
|
|
UStringPrepKey *b = (UStringPrepKey *)parm.pointer;
|
|
UHashTok namekey, pathkey;
|
|
namekey.pointer = b->name;
|
|
pathkey.pointer = b->path;
|
|
uint32_t unsignedHash = static_cast<uint32_t>(uhash_hashChars(namekey)) +
|
|
37u * static_cast<uint32_t>(uhash_hashChars(pathkey));
|
|
return static_cast<int32_t>(unsignedHash);
|
|
}
|
|
|
|
/* compares two entries */
|
|
static UBool U_CALLCONV
|
|
compareEntries(const UHashTok p1, const UHashTok p2) {
|
|
UStringPrepKey *b1 = (UStringPrepKey *)p1.pointer;
|
|
UStringPrepKey *b2 = (UStringPrepKey *)p2.pointer;
|
|
UHashTok name1, name2, path1, path2;
|
|
name1.pointer = b1->name;
|
|
name2.pointer = b2->name;
|
|
path1.pointer = b1->path;
|
|
path2.pointer = b2->path;
|
|
return ((UBool)(uhash_compareChars(name1, name2) &
|
|
uhash_compareChars(path1, path2)));
|
|
}
|
|
|
|
static void
|
|
usprep_unload(UStringPrepProfile* data){
|
|
udata_close(data->sprepData);
|
|
}
|
|
|
|
static int32_t
|
|
usprep_internal_flushCache(UBool noRefCount){
|
|
UStringPrepProfile *profile = NULL;
|
|
UStringPrepKey *key = NULL;
|
|
int32_t pos = UHASH_FIRST;
|
|
int32_t deletedNum = 0;
|
|
const UHashElement *e;
|
|
|
|
/*
|
|
* if shared data hasn't even been lazy evaluated yet
|
|
* return 0
|
|
*/
|
|
umtx_lock(&usprepMutex);
|
|
if (SHARED_DATA_HASHTABLE == NULL) {
|
|
umtx_unlock(&usprepMutex);
|
|
return 0;
|
|
}
|
|
|
|
/*creates an enumeration to iterate through every element in the table */
|
|
while ((e = uhash_nextElement(SHARED_DATA_HASHTABLE, &pos)) != NULL)
|
|
{
|
|
profile = (UStringPrepProfile *) e->value.pointer;
|
|
key = (UStringPrepKey *) e->key.pointer;
|
|
|
|
if ((noRefCount== false && profile->refCount == 0) ||
|
|
noRefCount== true) {
|
|
deletedNum++;
|
|
uhash_removeElement(SHARED_DATA_HASHTABLE, e);
|
|
|
|
/* unload the data */
|
|
usprep_unload(profile);
|
|
|
|
if(key->name != NULL) {
|
|
uprv_free(key->name);
|
|
key->name=NULL;
|
|
}
|
|
if(key->path != NULL) {
|
|
uprv_free(key->path);
|
|
key->path=NULL;
|
|
}
|
|
uprv_free(profile);
|
|
uprv_free(key);
|
|
}
|
|
|
|
}
|
|
umtx_unlock(&usprepMutex);
|
|
|
|
return deletedNum;
|
|
}
|
|
|
|
/* Works just like ucnv_flushCache()
|
|
static int32_t
|
|
usprep_flushCache(){
|
|
return usprep_internal_flushCache(false);
|
|
}
|
|
*/
|
|
|
|
static UBool U_CALLCONV usprep_cleanup(void){
|
|
if (SHARED_DATA_HASHTABLE != NULL) {
|
|
usprep_internal_flushCache(true);
|
|
if (SHARED_DATA_HASHTABLE != NULL && uhash_count(SHARED_DATA_HASHTABLE) == 0) {
|
|
uhash_close(SHARED_DATA_HASHTABLE);
|
|
SHARED_DATA_HASHTABLE = NULL;
|
|
}
|
|
}
|
|
gSharedDataInitOnce.reset();
|
|
return (SHARED_DATA_HASHTABLE == NULL);
|
|
}
|
|
U_CDECL_END
|
|
|
|
|
|
/** Initializes the cache for resources */
|
|
static void U_CALLCONV
|
|
createCache(UErrorCode &status) {
|
|
SHARED_DATA_HASHTABLE = uhash_open(hashEntry, compareEntries, NULL, &status);
|
|
if (U_FAILURE(status)) {
|
|
SHARED_DATA_HASHTABLE = NULL;
|
|
}
|
|
ucln_common_registerCleanup(UCLN_COMMON_USPREP, usprep_cleanup);
|
|
}
|
|
|
|
static void
|
|
initCache(UErrorCode *status) {
|
|
umtx_initOnce(gSharedDataInitOnce, &createCache, *status);
|
|
}
|
|
|
|
static UBool U_CALLCONV
|
|
loadData(UStringPrepProfile* profile,
|
|
const char* path,
|
|
const char* name,
|
|
const char* type,
|
|
UErrorCode* errorCode) {
|
|
/* load Unicode SPREP data from file */
|
|
UTrie _sprepTrie={ 0,0,0,0,0,0,0 };
|
|
UDataMemory *dataMemory;
|
|
const int32_t *p=NULL;
|
|
const uint8_t *pb;
|
|
UVersionInfo normUnicodeVersion;
|
|
int32_t normUniVer, sprepUniVer, normCorrVer;
|
|
|
|
if(errorCode==NULL || U_FAILURE(*errorCode)) {
|
|
return 0;
|
|
}
|
|
|
|
/* open the data outside the mutex block */
|
|
//TODO: change the path
|
|
dataMemory=udata_openChoice(path, type, name, isSPrepAcceptable, NULL, errorCode);
|
|
if(U_FAILURE(*errorCode)) {
|
|
return false;
|
|
}
|
|
|
|
p=(const int32_t *)udata_getMemory(dataMemory);
|
|
pb=(const uint8_t *)(p+_SPREP_INDEX_TOP);
|
|
utrie_unserialize(&_sprepTrie, pb, p[_SPREP_INDEX_TRIE_SIZE], errorCode);
|
|
_sprepTrie.getFoldingOffset=getSPrepFoldingOffset;
|
|
|
|
|
|
if(U_FAILURE(*errorCode)) {
|
|
udata_close(dataMemory);
|
|
return false;
|
|
}
|
|
|
|
/* in the mutex block, set the data for this process */
|
|
umtx_lock(&usprepMutex);
|
|
if(profile->sprepData==NULL) {
|
|
profile->sprepData=dataMemory;
|
|
dataMemory=NULL;
|
|
uprv_memcpy(&profile->indexes, p, sizeof(profile->indexes));
|
|
uprv_memcpy(&profile->sprepTrie, &_sprepTrie, sizeof(UTrie));
|
|
} else {
|
|
p=(const int32_t *)udata_getMemory(profile->sprepData);
|
|
}
|
|
umtx_unlock(&usprepMutex);
|
|
/* initialize some variables */
|
|
profile->mappingData=(uint16_t *)((uint8_t *)(p+_SPREP_INDEX_TOP)+profile->indexes[_SPREP_INDEX_TRIE_SIZE]);
|
|
|
|
u_getUnicodeVersion(normUnicodeVersion);
|
|
normUniVer = (normUnicodeVersion[0] << 24) + (normUnicodeVersion[1] << 16) +
|
|
(normUnicodeVersion[2] << 8 ) + (normUnicodeVersion[3]);
|
|
sprepUniVer = (dataVersion[0] << 24) + (dataVersion[1] << 16) +
|
|
(dataVersion[2] << 8 ) + (dataVersion[3]);
|
|
normCorrVer = profile->indexes[_SPREP_NORM_CORRECTNS_LAST_UNI_VERSION];
|
|
|
|
if(U_FAILURE(*errorCode)){
|
|
udata_close(dataMemory);
|
|
return false;
|
|
}
|
|
if( normUniVer < sprepUniVer && /* the Unicode version of SPREP file must be less than the Unicode Version of the normalization data */
|
|
normUniVer < normCorrVer && /* the Unicode version of the NormalizationCorrections.txt file should be less than the Unicode Version of the normalization data */
|
|
((profile->indexes[_SPREP_OPTIONS] & _SPREP_NORMALIZATION_ON) > 0) /* normalization turned on*/
|
|
){
|
|
*errorCode = U_INVALID_FORMAT_ERROR;
|
|
udata_close(dataMemory);
|
|
return false;
|
|
}
|
|
profile->isDataLoaded = true;
|
|
|
|
/* if a different thread set it first, then close the extra data */
|
|
if(dataMemory!=NULL) {
|
|
udata_close(dataMemory); /* NULL if it was set correctly */
|
|
}
|
|
|
|
|
|
return profile->isDataLoaded;
|
|
}
|
|
|
|
static UStringPrepProfile*
|
|
usprep_getProfile(const char* path,
|
|
const char* name,
|
|
UErrorCode *status){
|
|
|
|
UStringPrepProfile* profile = NULL;
|
|
|
|
initCache(status);
|
|
|
|
if(U_FAILURE(*status)){
|
|
return NULL;
|
|
}
|
|
|
|
UStringPrepKey stackKey;
|
|
/*
|
|
* const is cast way to save malloc, strcpy and free calls
|
|
* we use the passed in pointers for fetching the data from the
|
|
* hash table which is safe
|
|
*/
|
|
stackKey.name = (char*) name;
|
|
stackKey.path = (char*) path;
|
|
|
|
/* fetch the data from the cache */
|
|
umtx_lock(&usprepMutex);
|
|
profile = (UStringPrepProfile*) (uhash_get(SHARED_DATA_HASHTABLE,&stackKey));
|
|
if(profile != NULL) {
|
|
profile->refCount++;
|
|
}
|
|
umtx_unlock(&usprepMutex);
|
|
|
|
if(profile == NULL) {
|
|
/* else load the data and put the data in the cache */
|
|
LocalMemory<UStringPrepProfile> newProfile;
|
|
if(newProfile.allocateInsteadAndReset() == NULL) {
|
|
*status = U_MEMORY_ALLOCATION_ERROR;
|
|
return NULL;
|
|
}
|
|
|
|
/* load the data */
|
|
if(!loadData(newProfile.getAlias(), path, name, _SPREP_DATA_TYPE, status) || U_FAILURE(*status) ){
|
|
return NULL;
|
|
}
|
|
|
|
/* get the options */
|
|
newProfile->doNFKC = (UBool)((newProfile->indexes[_SPREP_OPTIONS] & _SPREP_NORMALIZATION_ON) > 0);
|
|
newProfile->checkBiDi = (UBool)((newProfile->indexes[_SPREP_OPTIONS] & _SPREP_CHECK_BIDI_ON) > 0);
|
|
|
|
LocalMemory<UStringPrepKey> key;
|
|
LocalMemory<char> keyName;
|
|
LocalMemory<char> keyPath;
|
|
if( key.allocateInsteadAndReset() == NULL ||
|
|
keyName.allocateInsteadAndCopy(static_cast<int32_t>(uprv_strlen(name)+1)) == NULL ||
|
|
(path != NULL &&
|
|
keyPath.allocateInsteadAndCopy(static_cast<int32_t>(uprv_strlen(path)+1)) == NULL)
|
|
) {
|
|
*status = U_MEMORY_ALLOCATION_ERROR;
|
|
usprep_unload(newProfile.getAlias());
|
|
return NULL;
|
|
}
|
|
|
|
umtx_lock(&usprepMutex);
|
|
// If another thread already inserted the same key/value, refcount and cleanup our thread data
|
|
profile = (UStringPrepProfile*) (uhash_get(SHARED_DATA_HASHTABLE,&stackKey));
|
|
if(profile != NULL) {
|
|
profile->refCount++;
|
|
usprep_unload(newProfile.getAlias());
|
|
}
|
|
else {
|
|
/* initialize the key members */
|
|
key->name = keyName.orphan();
|
|
uprv_strcpy(key->name, name);
|
|
if(path != NULL){
|
|
key->path = keyPath.orphan();
|
|
uprv_strcpy(key->path, path);
|
|
}
|
|
profile = newProfile.orphan();
|
|
|
|
/* add the data object to the cache */
|
|
profile->refCount = 1;
|
|
uhash_put(SHARED_DATA_HASHTABLE, key.orphan(), profile, status);
|
|
}
|
|
umtx_unlock(&usprepMutex);
|
|
}
|
|
|
|
return profile;
|
|
}
|
|
|
|
U_CAPI UStringPrepProfile* U_EXPORT2
|
|
usprep_open(const char* path,
|
|
const char* name,
|
|
UErrorCode* status){
|
|
|
|
if(status == NULL || U_FAILURE(*status)){
|
|
return NULL;
|
|
}
|
|
|
|
/* initialize the profile struct members */
|
|
return usprep_getProfile(path,name,status);
|
|
}
|
|
|
|
U_CAPI UStringPrepProfile* U_EXPORT2
|
|
usprep_openByType(UStringPrepProfileType type,
|
|
UErrorCode* status) {
|
|
if(status == NULL || U_FAILURE(*status)){
|
|
return NULL;
|
|
}
|
|
int32_t index = (int32_t)type;
|
|
if (index < 0 || index >= UPRV_LENGTHOF(PROFILE_NAMES)) {
|
|
*status = U_ILLEGAL_ARGUMENT_ERROR;
|
|
return NULL;
|
|
}
|
|
return usprep_open(NULL, PROFILE_NAMES[index], status);
|
|
}
|
|
|
|
U_CAPI void U_EXPORT2
|
|
usprep_close(UStringPrepProfile* profile){
|
|
if(profile==NULL){
|
|
return;
|
|
}
|
|
|
|
umtx_lock(&usprepMutex);
|
|
/* decrement the ref count*/
|
|
if(profile->refCount > 0){
|
|
profile->refCount--;
|
|
}
|
|
umtx_unlock(&usprepMutex);
|
|
|
|
}
|
|
|
|
U_CFUNC void
|
|
uprv_syntaxError(const UChar* rules,
|
|
int32_t pos,
|
|
int32_t rulesLen,
|
|
UParseError* parseError){
|
|
if(parseError == NULL){
|
|
return;
|
|
}
|
|
parseError->offset = pos;
|
|
parseError->line = 0 ; // we are not using line numbers
|
|
|
|
// for pre-context
|
|
int32_t start = (pos < U_PARSE_CONTEXT_LEN)? 0 : (pos - (U_PARSE_CONTEXT_LEN-1));
|
|
int32_t limit = pos;
|
|
|
|
u_memcpy(parseError->preContext,rules+start,limit-start);
|
|
//null terminate the buffer
|
|
parseError->preContext[limit-start] = 0;
|
|
|
|
// for post-context; include error rules[pos]
|
|
start = pos;
|
|
limit = start + (U_PARSE_CONTEXT_LEN-1);
|
|
if (limit > rulesLen) {
|
|
limit = rulesLen;
|
|
}
|
|
if (start < rulesLen) {
|
|
u_memcpy(parseError->postContext,rules+start,limit-start);
|
|
}
|
|
//null terminate the buffer
|
|
parseError->postContext[limit-start]= 0;
|
|
}
|
|
|
|
|
|
static inline UStringPrepType
|
|
getValues(uint16_t trieWord, int16_t& value, UBool& isIndex){
|
|
|
|
UStringPrepType type;
|
|
if(trieWord == 0){
|
|
/*
|
|
* Initial value stored in the mapping table
|
|
* just return USPREP_TYPE_LIMIT .. so that
|
|
* the source codepoint is copied to the destination
|
|
*/
|
|
type = USPREP_TYPE_LIMIT;
|
|
isIndex =false;
|
|
value = 0;
|
|
}else if(trieWord >= _SPREP_TYPE_THRESHOLD){
|
|
type = (UStringPrepType) (trieWord - _SPREP_TYPE_THRESHOLD);
|
|
isIndex =false;
|
|
value = 0;
|
|
}else{
|
|
/* get the type */
|
|
type = USPREP_MAP;
|
|
/* ascertain if the value is index or delta */
|
|
if(trieWord & 0x02){
|
|
isIndex = true;
|
|
value = trieWord >> 2; //mask off the lower 2 bits and shift
|
|
}else{
|
|
isIndex = false;
|
|
value = (int16_t)trieWord;
|
|
value = (value >> 2);
|
|
}
|
|
|
|
if((trieWord>>2) == _SPREP_MAX_INDEX_VALUE){
|
|
type = USPREP_DELETE;
|
|
isIndex =false;
|
|
value = 0;
|
|
}
|
|
}
|
|
return type;
|
|
}
|
|
|
|
// TODO: change to writing to UnicodeString not UChar *
|
|
static int32_t
|
|
usprep_map( const UStringPrepProfile* profile,
|
|
const UChar* src, int32_t srcLength,
|
|
UChar* dest, int32_t destCapacity,
|
|
int32_t options,
|
|
UParseError* parseError,
|
|
UErrorCode* status ){
|
|
|
|
uint16_t result;
|
|
int32_t destIndex=0;
|
|
int32_t srcIndex;
|
|
UBool allowUnassigned = (UBool) ((options & USPREP_ALLOW_UNASSIGNED)>0);
|
|
UStringPrepType type;
|
|
int16_t value;
|
|
UBool isIndex;
|
|
const int32_t* indexes = profile->indexes;
|
|
|
|
// no error checking the caller check for error and arguments
|
|
// no string length check the caller finds out the string length
|
|
|
|
for(srcIndex=0;srcIndex<srcLength;){
|
|
UChar32 ch;
|
|
|
|
U16_NEXT(src,srcIndex,srcLength,ch);
|
|
|
|
result=0;
|
|
|
|
UTRIE_GET16(&profile->sprepTrie,ch,result);
|
|
|
|
type = getValues(result, value, isIndex);
|
|
|
|
// check if the source codepoint is unassigned
|
|
if(type == USPREP_UNASSIGNED && allowUnassigned == false){
|
|
|
|
uprv_syntaxError(src,srcIndex-U16_LENGTH(ch), srcLength,parseError);
|
|
*status = U_STRINGPREP_UNASSIGNED_ERROR;
|
|
return 0;
|
|
|
|
}else if(type == USPREP_MAP){
|
|
|
|
int32_t index, length;
|
|
|
|
if(isIndex){
|
|
index = value;
|
|
if(index >= indexes[_SPREP_ONE_UCHAR_MAPPING_INDEX_START] &&
|
|
index < indexes[_SPREP_TWO_UCHARS_MAPPING_INDEX_START]){
|
|
length = 1;
|
|
}else if(index >= indexes[_SPREP_TWO_UCHARS_MAPPING_INDEX_START] &&
|
|
index < indexes[_SPREP_THREE_UCHARS_MAPPING_INDEX_START]){
|
|
length = 2;
|
|
}else if(index >= indexes[_SPREP_THREE_UCHARS_MAPPING_INDEX_START] &&
|
|
index < indexes[_SPREP_FOUR_UCHARS_MAPPING_INDEX_START]){
|
|
length = 3;
|
|
}else{
|
|
length = profile->mappingData[index++];
|
|
|
|
}
|
|
|
|
/* copy mapping to destination */
|
|
for(int32_t i=0; i< length; i++){
|
|
if(destIndex < destCapacity ){
|
|
dest[destIndex] = profile->mappingData[index+i];
|
|
}
|
|
destIndex++; /* for pre-flighting */
|
|
}
|
|
continue;
|
|
}else{
|
|
// subtract the delta to arrive at the code point
|
|
ch -= value;
|
|
}
|
|
|
|
}else if(type==USPREP_DELETE){
|
|
// just consume the codepoint and continue
|
|
continue;
|
|
}
|
|
//copy the code point into destination
|
|
if(ch <= 0xFFFF){
|
|
if(destIndex < destCapacity ){
|
|
dest[destIndex] = (UChar)ch;
|
|
}
|
|
destIndex++;
|
|
}else{
|
|
if(destIndex+1 < destCapacity ){
|
|
dest[destIndex] = U16_LEAD(ch);
|
|
dest[destIndex+1] = U16_TRAIL(ch);
|
|
}
|
|
destIndex +=2;
|
|
}
|
|
|
|
}
|
|
|
|
return u_terminateUChars(dest, destCapacity, destIndex, status);
|
|
}
|
|
|
|
/*
|
|
1) Map -- For each character in the input, check if it has a mapping
|
|
and, if so, replace it with its mapping.
|
|
|
|
2) Normalize -- Possibly normalize the result of step 1 using Unicode
|
|
normalization.
|
|
|
|
3) Prohibit -- Check for any characters that are not allowed in the
|
|
output. If any are found, return an error.
|
|
|
|
4) Check bidi -- Possibly check for right-to-left characters, and if
|
|
any are found, make sure that the whole string satisfies the
|
|
requirements for bidirectional strings. If the string does not
|
|
satisfy the requirements for bidirectional strings, return an
|
|
error.
|
|
[Unicode3.2] defines several bidirectional categories; each character
|
|
has one bidirectional category assigned to it. For the purposes of
|
|
the requirements below, an "RandALCat character" is a character that
|
|
has Unicode bidirectional categories "R" or "AL"; an "LCat character"
|
|
is a character that has Unicode bidirectional category "L". Note
|
|
|
|
|
|
that there are many characters which fall in neither of the above
|
|
definitions; Latin digits (<U+0030> through <U+0039>) are examples of
|
|
this because they have bidirectional category "EN".
|
|
|
|
In any profile that specifies bidirectional character handling, all
|
|
three of the following requirements MUST be met:
|
|
|
|
1) The characters in section 5.8 MUST be prohibited.
|
|
|
|
2) If a string contains any RandALCat character, the string MUST NOT
|
|
contain any LCat character.
|
|
|
|
3) If a string contains any RandALCat character, a RandALCat
|
|
character MUST be the first character of the string, and a
|
|
RandALCat character MUST be the last character of the string.
|
|
*/
|
|
U_CAPI int32_t U_EXPORT2
|
|
usprep_prepare( const UStringPrepProfile* profile,
|
|
const UChar* src, int32_t srcLength,
|
|
UChar* dest, int32_t destCapacity,
|
|
int32_t options,
|
|
UParseError* parseError,
|
|
UErrorCode* status ){
|
|
|
|
// check error status
|
|
if(U_FAILURE(*status)){
|
|
return 0;
|
|
}
|
|
|
|
//check arguments
|
|
if(profile==NULL ||
|
|
(src==NULL ? srcLength!=0 : srcLength<-1) ||
|
|
(dest==NULL ? destCapacity!=0 : destCapacity<0)) {
|
|
*status=U_ILLEGAL_ARGUMENT_ERROR;
|
|
return 0;
|
|
}
|
|
|
|
//get the string length
|
|
if(srcLength < 0){
|
|
srcLength = u_strlen(src);
|
|
}
|
|
// map
|
|
UnicodeString s1;
|
|
UChar *b1 = s1.getBuffer(srcLength);
|
|
if(b1==NULL){
|
|
*status = U_MEMORY_ALLOCATION_ERROR;
|
|
return 0;
|
|
}
|
|
int32_t b1Len = usprep_map(profile, src, srcLength,
|
|
b1, s1.getCapacity(), options, parseError, status);
|
|
s1.releaseBuffer(U_SUCCESS(*status) ? b1Len : 0);
|
|
|
|
if(*status == U_BUFFER_OVERFLOW_ERROR){
|
|
// redo processing of string
|
|
/* we do not have enough room so grow the buffer*/
|
|
b1 = s1.getBuffer(b1Len);
|
|
if(b1==NULL){
|
|
*status = U_MEMORY_ALLOCATION_ERROR;
|
|
return 0;
|
|
}
|
|
|
|
*status = U_ZERO_ERROR; // reset error
|
|
b1Len = usprep_map(profile, src, srcLength,
|
|
b1, s1.getCapacity(), options, parseError, status);
|
|
s1.releaseBuffer(U_SUCCESS(*status) ? b1Len : 0);
|
|
}
|
|
if(U_FAILURE(*status)){
|
|
return 0;
|
|
}
|
|
|
|
// normalize
|
|
UnicodeString s2;
|
|
if(profile->doNFKC){
|
|
const Normalizer2 *n2 = Normalizer2::getNFKCInstance(*status);
|
|
FilteredNormalizer2 fn2(*n2, *uniset_getUnicode32Instance(*status));
|
|
if(U_FAILURE(*status)){
|
|
return 0;
|
|
}
|
|
fn2.normalize(s1, s2, *status);
|
|
}else{
|
|
s2.fastCopyFrom(s1);
|
|
}
|
|
if(U_FAILURE(*status)){
|
|
return 0;
|
|
}
|
|
|
|
// Prohibit and checkBiDi in one pass
|
|
const UChar *b2 = s2.getBuffer();
|
|
int32_t b2Len = s2.length();
|
|
UCharDirection direction=U_CHAR_DIRECTION_COUNT, firstCharDir=U_CHAR_DIRECTION_COUNT;
|
|
UBool leftToRight=false, rightToLeft=false;
|
|
int32_t rtlPos =-1, ltrPos =-1;
|
|
|
|
for(int32_t b2Index=0; b2Index<b2Len;){
|
|
UChar32 ch = 0;
|
|
U16_NEXT(b2, b2Index, b2Len, ch);
|
|
|
|
uint16_t result;
|
|
UTRIE_GET16(&profile->sprepTrie,ch,result);
|
|
|
|
int16_t value;
|
|
UBool isIndex;
|
|
UStringPrepType type = getValues(result, value, isIndex);
|
|
|
|
if( type == USPREP_PROHIBITED ||
|
|
((result < _SPREP_TYPE_THRESHOLD) && (result & 0x01) /* first bit says it the code point is prohibited*/)
|
|
){
|
|
*status = U_STRINGPREP_PROHIBITED_ERROR;
|
|
uprv_syntaxError(b2, b2Index-U16_LENGTH(ch), b2Len, parseError);
|
|
return 0;
|
|
}
|
|
|
|
if(profile->checkBiDi) {
|
|
direction = ubidi_getClass(ch);
|
|
if(firstCharDir == U_CHAR_DIRECTION_COUNT){
|
|
firstCharDir = direction;
|
|
}
|
|
if(direction == U_LEFT_TO_RIGHT){
|
|
leftToRight = true;
|
|
ltrPos = b2Index-1;
|
|
}
|
|
if(direction == U_RIGHT_TO_LEFT || direction == U_RIGHT_TO_LEFT_ARABIC){
|
|
rightToLeft = true;
|
|
rtlPos = b2Index-1;
|
|
}
|
|
}
|
|
}
|
|
if(profile->checkBiDi == true){
|
|
// satisfy 2
|
|
if( leftToRight == true && rightToLeft == true){
|
|
*status = U_STRINGPREP_CHECK_BIDI_ERROR;
|
|
uprv_syntaxError(b2,(rtlPos>ltrPos) ? rtlPos : ltrPos, b2Len, parseError);
|
|
return 0;
|
|
}
|
|
|
|
//satisfy 3
|
|
if( rightToLeft == true &&
|
|
!((firstCharDir == U_RIGHT_TO_LEFT || firstCharDir == U_RIGHT_TO_LEFT_ARABIC) &&
|
|
(direction == U_RIGHT_TO_LEFT || direction == U_RIGHT_TO_LEFT_ARABIC))
|
|
){
|
|
*status = U_STRINGPREP_CHECK_BIDI_ERROR;
|
|
uprv_syntaxError(b2, rtlPos, b2Len, parseError);
|
|
return false;
|
|
}
|
|
}
|
|
return s2.extract(dest, destCapacity, *status);
|
|
}
|
|
|
|
|
|
/* data swapping ------------------------------------------------------------ */
|
|
|
|
U_CAPI int32_t U_EXPORT2
|
|
usprep_swap(const UDataSwapper *ds,
|
|
const void *inData, int32_t length, void *outData,
|
|
UErrorCode *pErrorCode) {
|
|
const UDataInfo *pInfo;
|
|
int32_t headerSize;
|
|
|
|
const uint8_t *inBytes;
|
|
uint8_t *outBytes;
|
|
|
|
const int32_t *inIndexes;
|
|
int32_t indexes[16];
|
|
|
|
int32_t i, offset, count, size;
|
|
|
|
/* udata_swapDataHeader checks the arguments */
|
|
headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode);
|
|
if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
|
|
return 0;
|
|
}
|
|
|
|
/* check data format and format version */
|
|
pInfo=(const UDataInfo *)((const char *)inData+4);
|
|
if(!(
|
|
pInfo->dataFormat[0]==0x53 && /* dataFormat="SPRP" */
|
|
pInfo->dataFormat[1]==0x50 &&
|
|
pInfo->dataFormat[2]==0x52 &&
|
|
pInfo->dataFormat[3]==0x50 &&
|
|
pInfo->formatVersion[0]==3
|
|
)) {
|
|
udata_printError(ds, "usprep_swap(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized as StringPrep .spp data\n",
|
|
pInfo->dataFormat[0], pInfo->dataFormat[1],
|
|
pInfo->dataFormat[2], pInfo->dataFormat[3],
|
|
pInfo->formatVersion[0]);
|
|
*pErrorCode=U_UNSUPPORTED_ERROR;
|
|
return 0;
|
|
}
|
|
|
|
inBytes=(const uint8_t *)inData+headerSize;
|
|
outBytes=(uint8_t *)outData+headerSize;
|
|
|
|
inIndexes=(const int32_t *)inBytes;
|
|
|
|
if(length>=0) {
|
|
length-=headerSize;
|
|
if(length<16*4) {
|
|
udata_printError(ds, "usprep_swap(): too few bytes (%d after header) for StringPrep .spp data\n",
|
|
length);
|
|
*pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
/* read the first 16 indexes (ICU 2.8/format version 3: _SPREP_INDEX_TOP==16, might grow) */
|
|
for(i=0; i<16; ++i) {
|
|
indexes[i]=udata_readInt32(ds, inIndexes[i]);
|
|
}
|
|
|
|
/* calculate the total length of the data */
|
|
size=
|
|
16*4+ /* size of indexes[] */
|
|
indexes[_SPREP_INDEX_TRIE_SIZE]+
|
|
indexes[_SPREP_INDEX_MAPPING_DATA_SIZE];
|
|
|
|
if(length>=0) {
|
|
if(length<size) {
|
|
udata_printError(ds, "usprep_swap(): too few bytes (%d after header) for all of StringPrep .spp data\n",
|
|
length);
|
|
*pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
|
|
return 0;
|
|
}
|
|
|
|
/* copy the data for inaccessible bytes */
|
|
if(inBytes!=outBytes) {
|
|
uprv_memcpy(outBytes, inBytes, size);
|
|
}
|
|
|
|
offset=0;
|
|
|
|
/* swap the int32_t indexes[] */
|
|
count=16*4;
|
|
ds->swapArray32(ds, inBytes, count, outBytes, pErrorCode);
|
|
offset+=count;
|
|
|
|
/* swap the UTrie */
|
|
count=indexes[_SPREP_INDEX_TRIE_SIZE];
|
|
utrie_swap(ds, inBytes+offset, count, outBytes+offset, pErrorCode);
|
|
offset+=count;
|
|
|
|
/* swap the uint16_t mappingTable[] */
|
|
count=indexes[_SPREP_INDEX_MAPPING_DATA_SIZE];
|
|
ds->swapArray16(ds, inBytes+offset, count, outBytes+offset, pErrorCode);
|
|
//offset+=count;
|
|
}
|
|
|
|
return headerSize+size;
|
|
}
|
|
|
|
#endif /* #if !UCONFIG_NO_IDNA */
|