/************* history ************ * * COMPONENT: JRD MODULE: INTL.CPP * generated by Marion V2.5 2/6/90 * from dev db on 4-JAN-1995 ***************************************************************** * * PR 2002-06-02 Added ugly c hack in * intl_back_compat_alloc_func_lookup. * When someone has time we need to change the references to * return (void*) function to something more C++ like * * 42 4711 3 11 17 tamlin 2001 * Added silly numbers before my name, and converted it to C++. * * 18850 daves 4-JAN-1995 * Fix gds__alloc usage * * 18837 deej 31-DEC-1994 * fixing up HARBOR_MERGE * * 18821 deej 27-DEC-1994 * HARBOR MERGE * * 18789 jdavid 19-DEC-1994 * Cast some functions * * 17508 jdavid 15-JUL-1994 * Bring it up to date * * 17500 daves 13-JUL-1994 * Bug 6645: Different calculation of partial keys * * 17202 katz 24-MAY-1994 * PC_PLATFORM requires the .dll extension * * 17191 katz 23-MAY-1994 * OS/2 requires the .dll extension * * 17180 katz 23-MAY-1994 * Define location of DLL on OS/2 * * 17149 katz 20-MAY-1994 * In JRD, isc_arg_number arguments are SLONG's not int's * * 16633 daves 19-APR-1994 * Bug 6202: International licensing uses INTERNATIONAL product code * * 16555 katz 17-APR-1994 * The last argument of calls to ERR_post should be 0 * * 16521 katz 14-APR-1994 * Borland C needs a decorated symbol to lookup * * 16403 daves 8-APR-1994 * Bug 6441: Emit an error whenever transliteration from ttype_binary attempted * * 16141 katz 28-MAR-1994 * Don't declare return value from ISC_lookup_entrypoint as API_ROUTINE * * The contents of this file are subject to the Interbase Public * License Version 1.0 (the "License"); you may not use this file * except in compliance with the License. You may obtain a copy * of the License at http://www.Inprise.com/IPL.html * * Software distributed under the License is distributed on an * "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, either express * or implied. See the License for the specific language governing * rights and limitations under the License. * * The Original Code was created by Inprise Corporation * and its predecessors. Portions created by Inprise Corporation are * Copyright (C) Inprise Corporation. * * All Rights Reserved. * Contributor(s): ______________________________________. * * 2002.10.29 Sean Leyne - Removed obsolete "Netware" port * * 2002.10.30 Sean Leyne - Removed support for obsolete "PC_PLATFORM" define * */ /* * PROGRAM: JRD Intl * MODULE: intl.cpp * DESCRIPTION: International text support routines * * copyright (c) 1992, 1993 by Borland International */ #include "firebird.h" #include #include "../jrd/common.h" #include #include "../jrd/jrd.h" #include "../jrd/req.h" #include "../jrd/val.h" #include "gen/iberror.h" #include "../jrd/intl.h" #include "../jrd/intl_classes.h" #include "../jrd/ods.h" #include "../jrd/btr.h" #include "../intl/charsets.h" #include "../intl/country_codes.h" #include "../jrd/gdsassert.h" #include "../jrd/license.h" #ifdef INTL_BUILTIN #include "../intl/ld_proto.h" #endif #include "../jrd/all_proto.h" #include "../jrd/cvt_proto.h" #include "../jrd/err_proto.h" #include "../jrd/evl_proto.h" #include "../jrd/fun_proto.h" #include "../jrd/gds_proto.h" #include "../jrd/iberr_proto.h" #include "../jrd/intl_proto.h" #include "../jrd/isc_proto.h" #include "../jrd/thd.h" #include "../jrd/evl_string.h" #include "../common/classes/init.h" #include "../jrd/plugin_manager.h" using namespace Jrd; #ifdef DEV_BUILD #define isprintable(x) ((((unsigned char)(x)) & 0x7F) >= ' ') #ifdef DEBUG_INTL #define INTL_TRACE(args) gds__log args #else #define INTL_TRACE(args) #endif #else #define INTL_TRACE(args) #endif #define IS_TEXT(x) (((x)->dsc_dtype == dtype_text) ||\ ((x)->dsc_dtype == dtype_varying)||\ ((x)->dsc_dtype == dtype_cstring)) #define TTYPE_TO_CHARSET(tt) ((SSHORT)((tt) & 0x00FF)) #define TTYPE_TO_COLLATION(tt) ((SSHORT)((tt) >> 8)) typedef unsigned char FILECHAR; static bool all_spaces(thread_db*, CHARSET_ID, const BYTE*, USHORT, USHORT); static void pad_spaces(thread_db*, CHARSET_ID, BYTE *, USHORT); static FPTR_SHORT lookup_init_function(USHORT, SSHORT, SSHORT); static void finish_texttype_init(TEXTTYPE); static USHORT nc_to_wc(TEXTTYPE, UCS2_CHAR*, USHORT, const UCHAR*, USHORT, SSHORT*, USHORT*); static USHORT wc_to_wc(TEXTTYPE, UCS2_CHAR*, USHORT, const UCS2_CHAR*, USHORT, SSHORT*, USHORT*); /* Name of module that implements text-type (n) */ #ifdef VMS /* Note: MUST be only the file name. The VMS lib$find_shared_image * call insists on file name only, not any "path" components. */ const char* INTL_MODULE1 = "FBINTL"; const char* INTL_MODULE2 = "FBINTL2"; #elif defined (WIN_NT) /* prefixed with $INTERBASE */ const char* INTL_MODULE1 = "fbintl.dll"; const char* INTL_MODULE2 = "fbintl2.dll"; #else /* prefixed with $INTERBASE */ const char* INTL_MODULE1 = "fbintl"; const char* INTL_MODULE2 = "fbintl2"; #endif const char* INTL_LOOKUP_ENTRY1 = "LD_lookup"; const char* INTL_LOOKUP_ENTRY2 = "LD2_lookup"; const char* INTL_USER_ENTRY = "USER_TEXTTYPE_%03d"; // We need all the structure definitions from the old interface #define INTL_ENGINE_INTERNAL #include "../jrd/intlobj.h" // storage for the loadable modules namespace { Firebird::InitInstance intlBCPlugins; } static bool bcLoaded = false; static const char *INTL_PLUGIN_DIR = "intl"; class CharsetIDGetter { public: static CHARSET_ID generate(const void *sender, const CsConvert& Item) { return Item.getToCS(); } }; // Classes and structures used internally to this file and intl implementation class CharSetContainer { public: CharSetContainer(MemoryPool& p, USHORT cs_id); CharSet getCharSet() { return cs; } TextType lookupCollation(thread_db* tdbb, USHORT tt_id); CsConvert lookupConverter(thread_db* tdbb, CHARSET_ID to_cs); static CharSetContainer* lookupCharset(thread_db* tdbb, SSHORT ttype, ISC_STATUS *status); private: Firebird::SortedArray, CHARSET_ID, CharsetIDGetter> charset_converters; Firebird::Array charset_collations; Firebird::SortedArray impossible_conversions; CharSet cs; }; CharSetContainer* CharSetContainer::lookupCharset(thread_db* tdbb, SSHORT ttype, ISC_STATUS *status) { /************************************** * * l o o k u p C h a r s e t * ************************************** * * Functional description * * Lookup a character set descriptor. * * First, search the appropriate vector that hangs * off the dbb. If not found, then call the lower * level lookup routine to allocate it, or return * null if we don't know about the charset. * * Returns: * *charset - if no errors; * - if error & err non NULL * NULL - if error & err NULL * **************************************/ CharSetContainer *cs = NULL; SET_TDBB(tdbb); Database* dbb = tdbb->tdbb_database; USHORT id = TTYPE_TO_CHARSET(ttype); if (id == CS_dynamic) id = tdbb->tdbb_attachment->att_charset; if (id >= dbb->dbb_charsets.size()) dbb->dbb_charsets.resize(id + 10); else cs = dbb->dbb_charsets[id]; // allocate a new character set object if we couldn't find one. if (!cs) { cs = FB_NEW(*dbb->dbb_permanent) CharSetContainer(*dbb->dbb_permanent, id); if (cs->getCharSet() == NULL) { delete cs; return NULL; } dbb->dbb_charsets[id] = cs; } fb_assert(cs != NULL); return cs; } CharSetContainer::CharSetContainer(MemoryPool& p, USHORT cs_id) : charset_converters(p), charset_collations(p), impossible_conversions(p), cs(NULL) { typedef USHORT (*CSInitFunc)(charset*, SSHORT, SSHORT); CSInitFunc csInitFunc = reinterpret_cast(lookup_init_function(type_charset, cs_id, 0)); charset* csL = FB_NEW(p) charset; memset(csL, 0, sizeof(charset)); if (!csInitFunc || (*csInitFunc)(csL, cs_id, 0) != 0) { delete csL; csL = NULL; } this->cs = csL; } CsConvert CharSetContainer::lookupConverter(thread_db* tdbb, CHARSET_ID to_cs) { size_t pos; if (charset_converters.find(to_cs, pos)) return charset_converters[pos]; if (impossible_conversions.find(to_cs, pos)) return NULL; if (to_cs == CS_UNICODE_UCS2) { return cs.getConvToUnicode(); } if (cs.getId() == CS_UNICODE_UCS2) { CharSet to_charset = INTL_charset_lookup(tdbb, to_cs, NULL); if (to_charset == NULL) return NULL; return to_charset.getConvFromUnicode(); } typedef USHORT (*CVTInitFunc)(csconvert*, SSHORT, SSHORT); CVTInitFunc cvtInitFunc = reinterpret_cast(lookup_init_function(type_csconvert, to_cs, cs.getId())); if (!cvtInitFunc) { impossible_conversions.add(to_cs); return NULL; } csconvert* cvt = FB_NEW(*tdbb->tdbb_database->dbb_permanent) csconvert; memset(cvt, 0, sizeof(csconvert)); if ((*cvtInitFunc)(cvt, to_cs, cs.getId()) != 0) { impossible_conversions.add(to_cs); delete cvt; return NULL; } CsConvert converter = cvt; fb_assert(converter.getFromCS() == cs.getId()); fb_assert(converter.getToCS() == to_cs); charset_converters.add(converter); return cvt; } TextType CharSetContainer::lookupCollation(thread_db* tdbb, USHORT tt_id) { const USHORT id = TTYPE_TO_COLLATION(tt_id); if (id < charset_collations.getCount() && charset_collations[id] != NULL) return charset_collations[id]; typedef USHORT (*TTInitFunc)(TEXTTYPE, SSHORT, SSHORT); TTInitFunc ttInitFunc = reinterpret_cast(lookup_init_function(type_texttype, tt_id, 0)); if (!ttInitFunc) return NULL; TEXTTYPE tt = FB_NEW(*tdbb->tdbb_database->dbb_permanent) texttype; memset(tt, 0, sizeof(texttype)); if ((*ttInitFunc)(tt, tt_id, 0) != 0) { delete tt; return NULL; } finish_texttype_init(tt); if (charset_collations.getCount() <= id) charset_collations.grow(id+1); charset_collations[id] = tt; return tt; } /* Below are templates for functions used in TextType implementation */ class MBStrConverter { public: MBStrConverter(thread_db* tdbb, TextType obj, const UCHAR* &str, SSHORT &len) { SSHORT err_code; USHORT err_pos; SSHORT out_len = obj.to_wc(NULL, 0, str, len, &err_code, &err_pos); if (out_len > (int) sizeof(tempBuffer)) out_str = FB_NEW(*tdbb->getDefaultPool()) UCS2_CHAR[out_len/2]; else out_str = tempBuffer; len = obj.to_wc(out_str, out_len, str, len, &err_code, &err_pos); str = reinterpret_cast(out_str); } ~MBStrConverter() { if (out_str != tempBuffer) delete[] out_str; } private: UCS2_CHAR tempBuffer[100], *out_str; }; class NullStrConverter { public: NullStrConverter(thread_db* tdbb, TextType obj, const UCHAR *str, SSHORT len) { } }; template class UpcaseConverter : public PrevConverter { public: UpcaseConverter(thread_db* tdbb, TextType obj, const UCHAR* &str, SSHORT &len) : PrevConverter(tdbb, obj, str, len) { if (len > (int) sizeof(tempBuffer)) out_str = FB_NEW(*tdbb->getDefaultPool()) UCHAR[len]; else out_str = tempBuffer; obj.str_to_upper(len, str, len, out_str); str = out_str; } ~UpcaseConverter() { if (out_str != tempBuffer) delete[] out_str; } private: UCHAR tempBuffer[100], *out_str; }; template class ContainsObjectImpl : public ContainsObject { public: ContainsObjectImpl(MemoryPool& pool, const CharType* str, SSHORT str_len) : evaluator(pool, str, str_len) { } void reset() { evaluator.reset(); } bool result() { return evaluator.getResult(); }; bool process(thread_db* tdbb, Jrd::TextType ttype, const UCHAR* str, SSHORT length) { StrConverter cvt(tdbb, ttype, str, length); fb_assert(length % sizeof(CharType) == 0); return evaluator.processNextChunk( reinterpret_cast(str), length / sizeof(CharType)); } ~ContainsObjectImpl() {}; static ContainsObject* create(thread_db* tdbb, TextType ttype, const UCHAR* str, SSHORT length) { StrConverter cvt(tdbb, ttype, str, length); fb_assert(length % sizeof(CharType) == 0); return FB_NEW(*tdbb->getDefaultPool()) ContainsObjectImpl(*tdbb->getDefaultPool(), reinterpret_cast(str), length / sizeof(CharType)); } static bool evaluate(thread_db* tdbb, TextType ttype, const UCHAR* s, SSHORT sl, const UCHAR* p, SSHORT pl) { StrConverter cvt1(tdbb, ttype, p, pl), cvt2(tdbb, ttype, s, sl); fb_assert(pl % sizeof(CharType) == 0); fb_assert(sl % sizeof(CharType) == 0); Firebird::ContainsEvaluator evaluator(*tdbb->getDefaultPool(), reinterpret_cast(p), pl / sizeof(CharType)); evaluator.processNextChunk(reinterpret_cast(s), sl / sizeof(CharType)); return evaluator.getResult(); } static void ttype_init(TEXTTYPE txtobj) { txtobj->texttype_fn_contains = evaluate; txtobj->texttype_fn_contains_create = create; } private: Firebird::ContainsEvaluator evaluator; }; template class LikeObjectImpl : public LikeObject { public: LikeObjectImpl(MemoryPool& pool, const CharType* str, SSHORT str_len, CharType escape) : evaluator(pool, str, str_len, escape, SQL_MATCH_ANY_CHARS, SQL_MATCH_1_CHAR) { } void reset() { evaluator.reset(); } bool result() { return evaluator.getResult(); }; bool process(thread_db* tdbb, Jrd::TextType ttype, const UCHAR* str, SSHORT length) { StrConverter cvt(tdbb, ttype, str, length); fb_assert(length % sizeof(CharType) == 0); return evaluator.processNextChunk( reinterpret_cast(str), length / sizeof(CharType)); } ~LikeObjectImpl() {}; static LikeObject* create(thread_db* tdbb, TextType ttype, const UCHAR* str, SSHORT length, UCS2_CHAR escape) { StrConverter cvt(tdbb, ttype, str, length); fb_assert(length % sizeof(CharType) == 0); return FB_NEW(*tdbb->getDefaultPool()) LikeObjectImpl(*tdbb->getDefaultPool(), reinterpret_cast(str), length / sizeof(CharType), escape); } static bool evaluate(thread_db* tdbb, TextType ttype, const UCHAR* s, SSHORT sl, const UCHAR* p, SSHORT pl, UCS2_CHAR escape) { StrConverter cvt1(tdbb, ttype, p, pl), cvt2(tdbb, ttype, s, sl); fb_assert(pl % sizeof(CharType) == 0); fb_assert(sl % sizeof(CharType) == 0); Firebird::LikeEvaluator evaluator(*tdbb->getDefaultPool(), reinterpret_cast(p), pl / sizeof(CharType), escape, SQL_MATCH_ANY_CHARS, SQL_MATCH_1_CHAR); evaluator.processNextChunk(reinterpret_cast(s), sl / sizeof(CharType)); return evaluator.getResult(); } static void ttype_init(TEXTTYPE txtobj) { txtobj->texttype_fn_like = evaluate; txtobj->texttype_fn_like_create = create; } private: Firebird::LikeEvaluator evaluator; }; static void finish_texttype_init(TEXTTYPE txtobj) { /************************************** * * f i n i s h _ t e x t t y p e _ i n i t * ************************************** * * Functional description * * Finish initializing a text object with pointers to * internal routines. * This is also a handy place to check the licensing bits * for the text object. * * Returns: * The TEXTTYPE_init bit in texttype_flags is set if the * object is sucessfully initialized. * **************************************/ typedef LikeObjectImpl nc_like; typedef ContainsObjectImpl, UCHAR> nc_contains; typedef LikeObjectImpl wc_like; typedef ContainsObjectImpl, UCS2_CHAR> wc_contains; typedef LikeObjectImpl mb_like; typedef ContainsObjectImpl, UCS2_CHAR> mb_contains; if ((txtobj->texttype_fn_to_wc == NULL) && (txtobj->texttype_bytes_per_char == 1)) { /* Finish initialization of a narrow character object */ txtobj->texttype_fn_to_wc = nc_to_wc; txtobj->texttype_fn_matches = EVL_nc_matches; txtobj->texttype_fn_sleuth_merge = EVL_nc_sleuth_merge; txtobj->texttype_fn_sleuth_check = EVL_nc_sleuth_check; if (!txtobj->texttype_fn_mbtowc) txtobj->texttype_fn_mbtowc = INTL_builtin_nc_mbtowc; nc_like::ttype_init(txtobj); nc_contains::ttype_init(txtobj); } else if ((txtobj->texttype_fn_to_wc == NULL) && (txtobj->texttype_bytes_per_char == 2)) { /* Finish initialization of a wide character object */ txtobj->texttype_fn_to_wc = reinterpret_cast(wc_to_wc); txtobj->texttype_fn_matches = reinterpret_cast(EVL_wc_matches); txtobj->texttype_fn_sleuth_merge = reinterpret_cast(EVL_wc_sleuth_merge); txtobj->texttype_fn_sleuth_check = reinterpret_cast(EVL_wc_sleuth_check); if (!txtobj->texttype_fn_mbtowc) txtobj->texttype_fn_mbtowc = INTL_builtin_wc_mbtowc; wc_like::ttype_init(txtobj); wc_contains::ttype_init(txtobj); } else if (txtobj->texttype_fn_to_wc != NULL) { /* Finish initialization of a multibyte character object */ txtobj->texttype_fn_matches = EVL_mb_matches; txtobj->texttype_fn_sleuth_merge = EVL_mb_sleuth_merge; txtobj->texttype_fn_sleuth_check = EVL_mb_sleuth_check; if (!txtobj->texttype_fn_mbtowc) txtobj->texttype_fn_mbtowc = INTL_builtin_mb_mbtowc; mb_like::ttype_init(txtobj); mb_contains::ttype_init(txtobj); } else fb_assert(0); txtobj->texttype_flags |= TEXTTYPE_init; } static FPTR_SHORT lookup_init_function( USHORT type, SSHORT parm1, SSHORT parm2) { /************************************** * * lookup_init_function * ************************************** * * Functional description * Find the allocator function for the requested international * character set using the c/IB/FB 6.0 interface. * Search algorithm is: * Look in intllib * Look in intllib2 * Look for a normal UDF entry * Abort with an error. * * Returns: * FALSE - no errors * TRUE - error occurred, and parameter was NULL; * - error occurred, and parameter non-NULL; * * ***************************************/ if (!bcLoaded) { intlBCPlugins().addSearchPath(INTL_PLUGIN_DIR); bcLoaded = true; } PluginManager::Plugin intlMod1 = intlBCPlugins().findPlugin(INTL_MODULE1); PluginManager::Plugin intlMod2 = intlBCPlugins().findPlugin(INTL_MODULE2); INTL_TRACE(("INTL: looking for obj %d ttype %d\n", objtype, parm1)); FPTR_SHORT function = INTL_builtin_lookup(type, parm1, parm2); if (function) return function; typedef USHORT (*pfn_lookup) (USHORT, FPTR_SHORT*, SSHORT, SSHORT); pfn_lookup lookup_fn; #ifdef INTL_BUILTIN if (LD_lookup(type, &function, parm1, parm2) != 0) function = NULL; else return function; #else /* Look for an InterBase supplied object to implement the text type */ /* The flu.cpp uses searchpath which expects a file name not a path */ INTL_TRACE(("INTL: trying %s %s\n", INTL_MODULE1, INTL_LOOKUP_ENTRY1)); Firebird::string tempStr(INTL_LOOKUP_ENTRY1); if ( intlMod1 && (lookup_fn = (pfn_lookup) intlMod1.lookupSymbol(tempStr)) ) { INTL_TRACE(("INTL: calling lookup %s %s\n", INTL_MODULE1, INTL_LOOKUP_ENTRY1)); if ((*lookup_fn) (type, &function, parm1, parm2) != 0) { function = NULL; } else { return function; } } #endif /* Still not found, check the set of supplimental international objects */ #ifdef INTL_BUILTIN if (LD2_lookup(type, &function, parm1, parm2) != 0) function = NULL; else return function; #else /* Look for an InterBase supplied object to implement the text type */ /* The flu.cpp uses searchpath which expects a file name not a path */ INTL_TRACE(("INTL: trying %s %s\n", INTL_MODULE2, INTL_LOOKUP_ENTRY2)); tempStr = INTL_LOOKUP_ENTRY2; if ( intlMod2 && (lookup_fn = (pfn_lookup) intlMod2.lookupSymbol(tempStr)) ) { INTL_TRACE(("INTL: calling lookup %s %s\n", INTL_MODULE2, INTL_LOOKUP_ENTRY2)); if ((*lookup_fn) (type, &function, parm1, parm2) != 0) { function = NULL; } else { return function; } } #endif /* Still not found, check if there is a UDF in the database defined the right way */ USHORT argcount; Firebird::string entry; switch (type) { case type_texttype: entry.printf(INTL_USER_ENTRY, parm1); argcount = 2; break; case type_charset: entry.printf("USER_CHARSET_%03d", parm1); argcount = 2; break; case type_csconvert: entry.printf("USER_TRANSLATE_%03d_%03d", parm1, parm2); argcount = 3; break; default: BUGCHECK(1); break; } INTL_TRACE(("INTL: trying user fn %s\n", entry.c_str())); UserFunction* function_block = FUN_lookup_function(entry, false); if (function_block) { INTL_TRACE(("INTL: found a user fn, validating\n")); if ((function_block->fun_count == argcount) && (function_block->fun_args == argcount) && (function_block->fun_return_arg == 0) && (function_block->fun_entrypoint != NULL) && (function_block->fun_rpt[0].fun_mechanism == FUN_value) && (function_block->fun_rpt[0].fun_desc.dsc_dtype == dtype_short) && (function_block->fun_rpt[1].fun_desc.dsc_dtype == dtype_short) && (function_block->fun_rpt[argcount - 1]. fun_desc.dsc_dtype == dtype_short) && (function_block->fun_rpt[argcount].fun_mechanism == FUN_reference) && (function_block->fun_rpt[argcount].fun_desc.dsc_dtype == dtype_text)) { return (FPTR_SHORT)function_block->fun_entrypoint; } } return NULL; } CHARSET_ID INTL_charset(thread_db* tdbb, USHORT ttype, FPTR_ERROR err) { /************************************** * * I N T L _ c h a r s e t * ************************************** * * Functional description * Return the character set ID for a piece of text. * **************************************/ switch (ttype) { case ttype_none: return (CS_NONE); case ttype_ascii: return (CS_ASCII); case ttype_unicode_fss: return (CS_UNICODE_FSS); case ttype_binary: return (CS_BINARY); case ttype_dynamic: SET_TDBB(tdbb); return (tdbb->tdbb_attachment->att_charset); default: return (TTYPE_TO_CHARSET(ttype)); } } int INTL_compare(thread_db* tdbb, const dsc* pText1, const dsc* pText2, FPTR_ERROR err) { /************************************** * * I N T L _ c o m p a r e * ************************************** * * Functional description * Compare two pieces of international text. * **************************************/ SET_TDBB(tdbb); fb_assert(pText1 != NULL); fb_assert(pText2 != NULL); fb_assert(IS_TEXT(pText1) && IS_TEXT(pText2)); fb_assert(INTL_data_or_binary(pText1) || INTL_data_or_binary(pText2)); fb_assert(err); /* normal compare routine from CVT_compare */ /* trailing spaces in strings are ignored for comparision */ UCHAR* p1; USHORT t1; USHORT length1 = CVT_get_string_ptr(pText1, &t1, &p1, NULL, 0, err); UCHAR* p2; USHORT t2; USHORT length2 = CVT_get_string_ptr(pText2, &t2, &p2, NULL, 0, err); /* YYY - by SQL II compare_type must be explicit in the SQL statement if there is any doubt */ SSHORT compare_type = MAX(t1, t2); /* YYY */ UCHAR buffer[MAX_KEY]; if (t1 != t2) { CHARSET_ID cs1 = INTL_charset(tdbb, t1, err); CHARSET_ID cs2 = INTL_charset(tdbb, t2, err); if (cs1 != cs2) { if (compare_type != t2) { /* convert pText2 to pText1's type, if possible */ /* YYY - should failure to convert really return an error here? Support joining a 437 & Latin1 Column, and we pick the compare_type as 437, still only want the equal values.... But then, what about < operations, which make no sense if the string cannot be expressed... */ length2 = INTL_convert_bytes(tdbb, cs1, buffer, sizeof(buffer), cs2, p2, length2, err); p2 = buffer; } else { /* convert pText1 to pText2's type, if possible */ length1 = INTL_convert_bytes(tdbb, cs2, buffer, sizeof(buffer), cs1, p1, length1, err); p1 = buffer; } } } TextType obj = INTL_texttype_lookup(tdbb, compare_type, err, NULL); return obj.compare(length1, p1, length2, p2); } USHORT INTL_convert_bytes(thread_db* tdbb, CHARSET_ID dest_type, BYTE* dest_ptr, USHORT dest_len, CHARSET_ID src_type, const BYTE* src_ptr, USHORT src_len, FPTR_ERROR err) { /************************************** * * I N T L _ c o n v e r t _ b y t e s * ************************************** * * Functional description * Given a string of bytes in one character set, convert it to another * character set. * * If (dest_ptr) is NULL, return the count of bytes needed to convert * the string. This does not guarantee the string can be converted, * the purpose of this is to allocate a large enough buffer. * * RETURNS: * Length of resulting string, in bytes. * calls (err) if conversion error occurs. * **************************************/ USHORT len; USHORT len2; SSHORT err_code = 0; USHORT err_position; SET_TDBB(tdbb); fb_assert(src_ptr != NULL); fb_assert(src_type != dest_type); fb_assert(err != NULL); const UCHAR* const start_dest_ptr = dest_ptr; if ((dest_type == CS_BINARY) || (dest_type == CS_NONE)) { /* See if we just need a length estimate */ if (dest_ptr == NULL) return (src_len); len = MIN(dest_len, src_len); if (len) do { *dest_ptr++ = *src_ptr++; } while (--len); /* See if only space characters are remaining */ len = src_len - MIN(dest_len, src_len); if (!len || all_spaces(tdbb, src_type, src_ptr, len, 0)) return (dest_ptr - start_dest_ptr); else (*err) (isc_arith_except, 0); } else if (src_len == 0) return (0); else if (src_type == CS_BINARY) (*err)(isc_arith_except, isc_arg_gds, isc_transliteration_failed, 0); else /* character sets are known to be different */ { /* Do we know an object from cs1 to cs2? */ CsConvert cs_obj = INTL_convert_lookup(tdbb, dest_type, src_type); if (cs_obj != NULL) { len = cs_obj.convert(dest_ptr, dest_len, src_ptr, src_len, &err_code, &err_position); if (!err_code || ((err_code == CS_TRUNCATION_ERROR) && all_spaces(tdbb, src_type, src_ptr, src_len, err_position))) { return (len); } else if (err_code == CS_TRUNCATION_ERROR) (*err) (isc_arith_except, 0); else (*err) (isc_arith_except, isc_arg_gds, isc_transliteration_failed, 0); } /* Find a CS1 to UNICODE object */ CharSet from_cs = INTL_charset_lookup(tdbb, src_type, NULL); if (from_cs == NULL) (*err)(isc_arith_except, isc_arg_gds, isc_text_subtype, isc_arg_number, (ISC_STATUS) src_type, 0); /* ** allocate a temporary buffer that is large enough. */ BYTE* tmp_buffer = (BYTE *) FB_NEW(*tdbb->getDefaultPool()) char[(SLONG) src_len * sizeof(UCS2_CHAR)]; cs_obj = from_cs.getConvToUnicode(); fb_assert(cs_obj != NULL); len = cs_obj.convert(tmp_buffer, src_len * 2, src_ptr, src_len, &err_code, &err_position); if (err_code && !((err_code == CS_TRUNCATION_ERROR) && all_spaces(tdbb, src_type, src_ptr, src_len, err_position))) { delete [] tmp_buffer; if (err_code == CS_TRUNCATION_ERROR) (*err) (isc_arith_except, 0); else (*err) (isc_arith_except, isc_arg_gds, isc_transliteration_failed, 0); } /* Find a UNICODE to CS2 object */ CharSet to_cs = INTL_charset_lookup(tdbb, dest_type, NULL); if (to_cs == NULL) { delete [] tmp_buffer; (*err) (isc_arith_except, isc_arg_gds, isc_text_subtype, isc_arg_number, (ISC_STATUS) dest_type, 0); } cs_obj = to_cs.getConvFromUnicode(); fb_assert(cs_obj != NULL); len2 = cs_obj.convert(dest_ptr, dest_len, tmp_buffer, len, &err_code, &err_position); if (err_code && !((err_code == CS_TRUNCATION_ERROR) && all_spaces(tdbb, CS_UNICODE_UCS2, tmp_buffer, len, err_position))) { delete [] tmp_buffer; if (err_code == CS_TRUNCATION_ERROR) (*err) (isc_arith_except, 0); else (*err) (isc_arith_except, isc_arg_gds, isc_transliteration_failed, 0); } delete [] tmp_buffer; return (len2); } return (0); /* to remove compiler errors. This should never be executed */ } CsConvert INTL_convert_lookup(thread_db* tdbb, CHARSET_ID to_cs, CHARSET_ID from_cs) { /************************************** * * I N T L _ c o n v e r t _ l o o k u p * ************************************** * * Functional description * **************************************/ CharSetContainer *charset; SET_TDBB(tdbb); Database* dbb = tdbb->tdbb_database; CHECK_DBB(dbb); if (from_cs == CS_dynamic) from_cs = tdbb->tdbb_attachment->att_charset; if (to_cs == CS_dynamic) to_cs = tdbb->tdbb_attachment->att_charset; /* Should from_cs == to_cs? be handled better? YYY */ fb_assert(from_cs != CS_dynamic); fb_assert(to_cs != CS_dynamic); charset = CharSetContainer::lookupCharset(tdbb, from_cs, NULL); if (charset == NULL) return NULL; return charset->lookupConverter(tdbb, to_cs); } int INTL_convert_string(dsc* to, const dsc* from, FPTR_ERROR err) { /************************************** * * I N T L _ c o n v e r t _ s t r i n g * ************************************** * * Functional description * Convert a string from one type to another * * RETURNS: * 0 if no error in conversion * non-zero otherwise. * CVC: Unfortunately, this function puts the source in the 2nd param, * as opposed to the CVT routines, so const help mitigating coding mistakes. * **************************************/ /* Note: This function is called from outside the engine as well as inside - we likely can't get rid of JRD_get_thread_data here */ thread_db* tdbb = JRD_get_thread_data(); if (tdbb == NULL) /* are we in the Engine? */ return (1); /* no, then can't access intl gah */ fb_assert(to != NULL); fb_assert(from != NULL); fb_assert(IS_TEXT(to) && IS_TEXT(from)); CHARSET_ID from_cs = INTL_charset(tdbb, INTL_TTYPE(from), err); CHARSET_ID to_cs = INTL_charset(tdbb, INTL_TTYPE(to), err); UCHAR* p = to->dsc_address; /* Must convert dtype(cstring,text,vary) and ttype(ascii,binary,..intl..) */ UCHAR* from_ptr; USHORT from_type; const USHORT from_len = CVT_get_string_ptr(from, &from_type, &from_ptr, NULL, 0, err); USHORT to_size, to_len, to_fill; to_size = to_len = TEXT_LEN(to); USHORT from_fill; const UCHAR* q = from_ptr; switch (to->dsc_dtype) { case dtype_text: if ((from_cs != to_cs) && (to_cs != CS_BINARY) && (to_cs != CS_NONE) && (from_cs != CS_NONE)) { to_len = INTL_convert_bytes(tdbb, to_cs, to->dsc_address, to_size, from_cs, from_ptr, from_len, err); to_fill = to_size - to_len; from_fill = 0; /* Convert_bytes handles source truncation */ p += to_len; } else { /* binary string can always be converted TO by byte-copy */ to_len = MIN(from_len, to_size); from_fill = from_len - to_len; to_fill = to_size - to_len; if (to_len) do *p++ = *q++; while (--to_len); } if (to_fill > 0) pad_spaces(tdbb, to_cs, p, to_fill); break; case dtype_cstring: if ((from_cs != to_cs) && (to_cs != CS_BINARY) && (to_cs != CS_NONE) && (from_cs != CS_NONE)) { to_len = INTL_convert_bytes(tdbb, to_cs, to->dsc_address, to_size, from_cs, from_ptr, from_len, err); to->dsc_address[to_len] = 0; from_fill = 0; /* Convert_bytes handles source truncation */ } else { /* binary string can always be converted TO by byte-copy */ to_len = MIN(from_len, to_size); from_fill = from_len - to_len; if (to_len) do *p++ = *q++; while (--to_len); *p = 0; } break; case dtype_varying: if ((from_cs != to_cs) && (to_cs != CS_BINARY) && (to_cs != CS_NONE) && (from_cs != CS_NONE)) { to_len = INTL_convert_bytes(tdbb, to_cs, reinterpret_cast(((vary*) to->dsc_address)->vary_string), to_size, from_cs, from_ptr, from_len, err); ((vary*) to->dsc_address)->vary_length = to_len; from_fill = 0; /* Convert_bytes handles source truncation */ } else { /* binary string can always be converted TO by byte-copy */ to_len = MIN(from_len, to_size); from_fill = from_len - to_len; ((vary*) p)->vary_length = to_len; p = reinterpret_cast(((vary*) p)->vary_string); if (to_len) do *p++ = *q++; while (--to_len); } break; } if (from_fill) /* Make sure remaining characters on From string are spaces */ if (!all_spaces(tdbb, from_cs, q, from_fill, 0)) (*err) (isc_arith_except, 0); return 0; } int INTL_data(const dsc* pText) { /************************************** * * I N T L _ d a t a * ************************************** * * Functional description * Given an input text descriptor, * return TRUE if the data pointed to represents * international text (subject to user defined or non-binary * collation or comparison). * **************************************/ fb_assert(pText != NULL); if (!IS_TEXT(pText)) return FALSE; if (!INTERNAL_TTYPE(pText)) return TRUE; return FALSE; } int INTL_data_or_binary(const dsc* pText) { /************************************** * * I N T L _ d a t a _ o r _ b i n a r y * ************************************** * * Functional description * **************************************/ return (INTL_data(pText) || (pText->dsc_ttype() == ttype_binary)); } int INTL_defined_type(thread_db* tdbb, ISC_STATUS * status, SSHORT t_type) { /************************************** * * I N T L _ d e f i n e d _ t y p e * ************************************** * * Functional description * Is (t_type) a known text type? * Return: * FALSE type is not defined. * TRUE type is defined * status set to gds_status codes to describe any error. * * Note: * Due to cleanup that must happen in DFW, this routine * must return, and not call ERR directly. * **************************************/ SET_TDBB(tdbb); if (status) status[0] = isc_arg_end; TextType obj = INTL_texttype_lookup(tdbb, t_type, NULL, status); if (obj == NULL) return FALSE; return TRUE; } UCS2_CHAR INTL_getch(thread_db* tdbb, TextType* obj, SSHORT t_type, const UCHAR** ptr, USHORT* count) { /************************************** * * I N T L _ g e t c h * ************************************** * * Functional description * Get next character from a buffer. * **************************************/ SET_TDBB(tdbb); fb_assert(obj); fb_assert(ptr); if (*obj == NULL) { *obj = INTL_texttype_lookup(tdbb, t_type, ERR_post, NULL); fb_assert(*obj != NULL); } UCS2_CHAR wc; const SSHORT used = obj->mbtowc(&wc, *ptr, *count); if (used == -1) return 0; *ptr += used; *count -= used; return wc; } void INTL_init(thread_db* tdbb) { /************************************** * * I N T L _ i n i t * ************************************** * * Functional description * **************************************/ } USHORT INTL_key_length(thread_db* tdbb, USHORT idxType, USHORT iLength) { /************************************** * * I N T L _ k e y _ l e n g t h * ************************************** * * Functional description * Given an index type, and a maximum length (iLength) * return the length of the byte string key descriptor to * use when collating text of this type. * **************************************/ SET_TDBB(tdbb); fb_assert(idxType >= idx_first_intl_string); const SSHORT ttype = INTL_INDEX_TO_TEXT(idxType); USHORT key_length; if (ttype >= 0 && ttype <= ttype_last_internal) key_length = iLength; else { TextType obj = INTL_texttype_lookup(tdbb, ttype, ERR_post, NULL); key_length = obj.key_length(iLength); } /* Validity checks on the computed key_length */ if (key_length > MAX_KEY) key_length = MAX_KEY; if (key_length < iLength) key_length = iLength; return (key_length); } CharSet INTL_charset_lookup(thread_db* tdbb, SSHORT parm1, ISC_STATUS* status) { /************************************** * * I N T L _ c h a r s e t _ l o o k u p * ************************************** * * Functional description * * Lookup a character set descriptor. * * First, search the appropriate vector that hangs * off the dbb. If not found, then call the lower * level lookup routine to allocate it, or return * null if we don't know about the charset. * * Returns: * *charset - if no errors; * - if error & err non NULL * NULL - if error & err NULL * **************************************/ CharSetContainer *cs = CharSetContainer::lookupCharset(tdbb, parm1, status); if (!cs) return NULL; return cs->getCharSet(); } TextType INTL_texttype_lookup(thread_db* tdbb, SSHORT parm1, FPTR_ERROR err, ISC_STATUS* status) { /************************************** * * I N T L _ t e x t t y p e _ l o o k u p * ************************************** * * Functional description * * Lookup either a character set descriptor or * texttype descriptor object. * * First, search the appropriate vector that hangs * off the dbb. If not found, then call the lower * level lookup routine to find it in the libraries. * * Returns: * *object - if no errors; * - if error & err non NULL * NULL - if error & err NULL * **************************************/ SET_TDBB(tdbb); Database* dbb = tdbb->tdbb_database; if (parm1 == ttype_dynamic) parm1 = MAP_CHARSET_TO_TTYPE(tdbb->tdbb_attachment->att_charset); CharSetContainer* csc = CharSetContainer::lookupCharset(tdbb, parm1, status); if (!csc) return NULL; return csc->lookupCollation(tdbb, parm1); } void INTL_pad_spaces(thread_db* tdbb, DSC * type, UCHAR * string, USHORT length) { /************************************** * * I N T L _ p a d _ s p a c e s * ************************************** * * Functional description * Pad a buffer with spaces, using the character * set's defined space character. * **************************************/ SET_TDBB(tdbb); fb_assert(type != NULL); fb_assert(IS_TEXT(type)); fb_assert(string != NULL); const USHORT charset = INTL_charset(tdbb, type->dsc_ttype(), NULL); pad_spaces(tdbb, charset, string, length); } USHORT INTL_string_to_key(thread_db* tdbb, USHORT idxType, const dsc* pString, DSC* pByte, bool partial) { /************************************** * * I N T L _ s t r i n g _ t o _ k e y * ************************************** * * Functional description * Given an input string, convert it to a byte string * that will collate naturally (byte order). * * Return the length of the resulting byte string. * **************************************/ UCHAR pad_char; SSHORT ttype; SET_TDBB(tdbb); fb_assert(idxType >= idx_first_intl_string || idxType == idx_string || idxType == idx_byte_array || idxType == idx_metadata); fb_assert(pString != NULL); fb_assert(pByte != NULL); fb_assert(pString->dsc_address != NULL); fb_assert(pByte->dsc_address != NULL); fb_assert(pByte->dsc_dtype == dtype_text); switch (idxType) { case idx_string: pad_char = ' '; ttype = ttype_none; break; case idx_byte_array: pad_char = 0; ttype = ttype_binary; break; case idx_metadata: pad_char = ' '; ttype = ttype_metadata; break; default: pad_char = 0; ttype = INTL_INDEX_TO_TEXT(idxType); break; } /* Make a string into the proper type of text */ UCHAR buffer[MAX_KEY]; const char* src; USHORT len = CVT_make_string(pString, ttype, &src, reinterpret_cast(buffer), sizeof(buffer), ERR_post); USHORT outlen; char* dest = reinterpret_cast(pByte->dsc_address); switch (ttype) { case ttype_metadata: case ttype_binary: case ttype_ascii: case ttype_none: while (len--) *dest++ = *src++; /* strip off ending pad characters */ while (dest > (const char*)pByte->dsc_address) { if (*(dest - 1) == pad_char) dest--; else break; } outlen = (dest - (const char*)pByte->dsc_address); break; default: TextType obj = INTL_texttype_lookup(tdbb, ttype, ERR_post, NULL); outlen = obj.string_to_key(len, reinterpret_cast(src), pByte->dsc_length, reinterpret_cast(dest), partial ? TRUE : FALSE); break; } return (outlen); } int INTL_str_to_upper(thread_db* tdbb, DSC * pString) { /************************************** * * I N T L _ s t r _ t o _ u p p e r * ************************************** * * Functional description * Given an input string, convert it to uppercase * **************************************/ SET_TDBB(tdbb); fb_assert(pString != NULL); fb_assert(pString->dsc_address != NULL); UCHAR* src; UCHAR buffer[MAX_KEY]; USHORT ttype; USHORT len = CVT_get_string_ptr(pString, &ttype, &src, reinterpret_cast(buffer), sizeof(buffer), ERR_post); UCHAR* dest; switch (ttype) { case ttype_binary: /* cannot uppercase binary strings */ break; case ttype_none: case ttype_ascii: case ttype_unicode_fss: dest = src; while (len--) { *dest++ = UPPER7(*src); src++; } break; default: TextType obj = INTL_texttype_lookup(tdbb, ttype, ERR_post, NULL); obj.str_to_upper(len, src, len, src); break; } /* * Added to remove compiler errors. Callers are not checking * the return code from this function 4/5/95. */ return (0); } UCHAR INTL_upper(thread_db* tdbb, USHORT ttype, UCHAR ch) { /************************************** * * I N T L _ u p p e r * ************************************** * * Functional description * Given an input character, convert it to uppercase * **************************************/ SET_TDBB(tdbb); switch (ttype) { case ttype_binary: /* cannot uppercase binary strings */ return (ch); case ttype_none: case ttype_ascii: case ttype_unicode_fss: return (UPPER7(ch)); default: TextType obj = INTL_texttype_lookup(tdbb, ttype, ERR_post, NULL); return obj.to_upper(ch); } } static bool all_spaces( thread_db* tdbb, CHARSET_ID charset, const BYTE* ptr, USHORT len, USHORT offset) { /************************************** * * a l l _ s p a c e s * ************************************** * * Functional description * determine if the string at ptr[offset] ... ptr[len] is entirely * spaces, as per the space definition of (charset). * The binary representation of a Space is character-set dependent. * (0x20 for Ascii, 0x0020 for Unicode, 0x20 for SJIS, but must watch for * 0x??20, which is NOT a space. **************************************/ SET_TDBB(tdbb); fb_assert(ptr != NULL); CharSet obj = INTL_charset_lookup(tdbb, charset, NULL); fb_assert(obj != NULL); /* * We are assuming offset points to the first byte which was not * consumed in a conversion. And that offset is pointing * to a character boundary */ // Single-octet character sets are optimized here if (obj.getSpaceLength() == 1) { const BYTE* p = &ptr[offset]; const BYTE* const end = &ptr[len]; while (p < end) { if (*p++ != *obj.getSpace()) return false; } return true; } else { const BYTE* p = &ptr[offset]; const BYTE* const end = &ptr[len]; const unsigned char* space = obj.getSpace(); const unsigned char* const end_space = &space[obj.getSpaceLength()]; while (p < end) { space = obj.getSpace(); while (p < end && space < end_space) { if (*p++ != *space++) return false; } } return true; } } static USHORT nc_to_wc(TEXTTYPE obj, UCS2_CHAR* pWide, USHORT nWide, /* byte count */ const UCHAR* pNarrow, USHORT nNarrow, /* byte count */ SSHORT* err_code, USHORT* err_position) { /************************************** * * n c _ t o _ w c * ************************************** * * Functional description * Copies narrow chars buffer into wide chars buffer for charset NONE * **************************************/ fb_assert(obj != NULL); fb_assert((pNarrow != NULL) || (pWide == NULL)); fb_assert(err_code != NULL); fb_assert(err_position != NULL); *err_code = 0; if (pWide == NULL) return (sizeof(UCS2_CHAR) * nNarrow); /* all cases */ const UCS2_CHAR* const pStart = pWide; const UCHAR* const pNarrowStart = pNarrow; while (nWide-- > 1 && nNarrow) { /* YYY - Byte order issues here */ *pWide++ = (UCS2_CHAR) *pNarrow++; nWide--; nNarrow--; } if (!*err_code && nNarrow) { *err_code = CS_TRUNCATION_ERROR; } *err_position = (pNarrow - pNarrowStart) * sizeof(*pNarrow); return ((pWide - pStart) * sizeof(*pWide)); } static USHORT wc_to_wc(TEXTTYPE obj, UCS2_CHAR* pDest, USHORT nDest, /* byte count */ const UCS2_CHAR* pSrc, USHORT nSrc, /* byte count */ SSHORT* err_code, USHORT* err_position) { /************************************** * * w c _ t o _ w c * ************************************** * * Functional description * *************************************/ fb_assert(obj != NULL); fb_assert((pSrc != NULL) || (pDest == NULL)); fb_assert(err_code != NULL); fb_assert(err_position != NULL); *err_code = 0; if (pDest == NULL) /* length estimate needed? */ return (nSrc); const UCS2_CHAR* const pStart = pDest; const UCS2_CHAR* const pStart_src = pSrc; while (nDest > 1 && nSrc > 1) { *pDest++ = *pSrc++; nDest -= 2; nSrc -= 2; } if (!*err_code && nSrc) { *err_code = CS_TRUNCATION_ERROR; } *err_position = (pSrc - pStart_src) * sizeof(*pSrc); return ((pDest - pStart) * sizeof(*pDest)); } static void pad_spaces(thread_db* tdbb, CHARSET_ID charset, BYTE* ptr, USHORT len) { /* byte count */ /************************************** * * p a d _ s p a c e s * ************************************** * * Functional description * Pad a buffer with the character set defined space character. * **************************************/ SET_TDBB(tdbb); fb_assert(ptr != NULL); CharSet obj = INTL_charset_lookup(tdbb, charset, NULL); fb_assert(obj != NULL); /* Single-octet character sets are optimized here */ if (obj.getSpaceLength() == 1) { const BYTE* const end = &ptr[len]; while (ptr < end) *ptr++ = *obj.getSpace(); } else { const BYTE* const end = &ptr[len]; const UCHAR* space = obj.getSpace(); const UCHAR* const end_space = &space[obj.getSpaceLength()]; while (ptr < end) { space = obj.getSpace(); while (ptr < end && space < end_space) { *ptr++ = *space++; } /* This fb_assert is checking that we didn't have a buffer-end * in the middle of a space character */ fb_assert(!(ptr == end) || (space == end_space)); } } }