MCOL-3536 collation

dbcon/joblist/dictstep-jl.cpp

@@ -59,6 +59,7 @@ DictStepJL::DictStepJL(const pDictionaryStep& dict)
         filterString = dict.fFilterString;
 
     filterCount = dict.fFilterCount;
+    charsetNumber = dict.fColType.charsetNumber;
 }
 
 DictStepJL::~DictStepJL()
@@ -88,7 +89,7 @@ void DictStepJL::createCommand(ByteStream& bs) const
     }
     else
         bs << filterString;
-
+    bs << charsetNumber;
     CommandJL::createCommand(bs);
 }
 

dbcon/joblist/dictstep-jl.h

@@ -76,6 +76,7 @@ private:
     std::vector<std::string> eqFilter;
     bool hasEqFilter;
     uint8_t eqOp;    // COMPARE_EQ or COMPARE_NE
+    uint32_t charsetNumber;
 };
 
 };  // namespace

primitives/linux-port/dictionary.cpp

@@ -103,7 +103,7 @@ Notes:
 */
 
 void PrimitiveProcessor::p_TokenByScan(const TokenByScanRequestHeader* h,
-                                       TokenByScanResultHeader* ret, unsigned outSize, bool utf8,
+                                       TokenByScanResultHeader* ret, unsigned outSize,
                                        boost::shared_ptr<DictEqualityFilter> eqFilter)
 {
     const DataValue* args;
@@ -113,7 +113,6 @@ void PrimitiveProcessor::p_TokenByScan(const TokenByScanRequestHeader* h,
     int offsetIndex, argIndex, argsOffset;
     bool cmpResult = false;
     int tmp, i, err;
-
     const char* sig;
     uint16_t siglen;
 
@@ -191,6 +190,8 @@ void PrimitiveProcessor::p_TokenByScan(const TokenByScanRequestHeader* h,
         if (eqFilter)
         {
             // MCOL-1246 Trim whitespace before match
+            // TODO MCOL-3536 use CHARSET_INFO* cs for collation
+            // cs->hash_sort(hash_sort(const uchar *key, size_t len, ulong *nr1, ulong *nr2)) 
             string strData(sig, siglen);
             boost::trim_right_if(strData, boost::is_any_of(" "));
             bool gotIt = eqFilter->find(strData) != eqFilter->end();
@@ -214,15 +215,8 @@ void PrimitiveProcessor::p_TokenByScan(const TokenByScanRequestHeader* h,
                 cmpResult = !cmpResult;
         }
         else
-        {
-            if (utf8)
         {
             tmp = cs->strnncoll(sig, siglen, args->data, args->len);
-            }
-            else
-            {
-                tmp = strncmp(sig, args->data, std::min(siglen, args->len));
-            }
             cmpResult = compare(tmp, h->COP1, siglen, args->len);
         }
 
@@ -262,15 +256,8 @@ void PrimitiveProcessor::p_TokenByScan(const TokenByScanRequestHeader* h,
                 }
 
                 else
-                {
-                    if (utf8)
                 {
                     tmp = cs->strnncoll(sig, siglen, args->data, args->len);
-                    }
-                    else
-                    {
-                        tmp = strncmp(sig, args->data, std::min(siglen, args->len));
-                    }
                     cmpResult = compare(tmp, h->COP2, siglen, args->len);
                 }
 
@@ -297,15 +284,8 @@ void PrimitiveProcessor::p_TokenByScan(const TokenByScanRequestHeader* h,
                     }
 
                     else
-                    {
-                        if (utf8)
                     {
                         tmp = cs->strnncoll(sig, siglen, args->data, args->len);
-                        }
-                        else
-                        {
-                            tmp = strncmp(sig, args->data, std::min(siglen, args->len));
-                        }
                         cmpResult = compare(tmp, h->COP2, siglen, args->len);
                     }
 
@@ -667,8 +647,12 @@ PrimitiveProcessor::makeLikeFilter (const DictFilterElement* filterString, uint3
     return ret;
 }
 
-void PrimitiveProcessor::p_Dictionary(const DictInput* in, vector<uint8_t>* out, bool utf8,
+void PrimitiveProcessor::p_Dictionary(const DictInput* in, 
-                                      bool skipNulls, boost::shared_ptr<DictEqualityFilter> eqFilter, uint8_t eqOp)
+                                      vector<uint8_t>* out,
+                                      bool skipNulls, 
+                                      uint32_t charsetNumber,
+                                      boost::shared_ptr<DictEqualityFilter> eqFilter, 
+                                      uint8_t eqOp)
 {
     PrimToken* outToken;
     const DictFilterElement* filter = 0;
@@ -679,6 +663,7 @@ void PrimitiveProcessor::p_Dictionary(const DictInput* in, vector<uint8_t>* out,
     uint16_t aggCount;
     bool cmpResult;
     DictOutput header;
+    const CHARSET_INFO* cs = get_charset(charsetNumber, MYF(MY_WME));
 
     // default size of the ouput to something sufficiently large to prevent
     // excessive reallocation and copy when resizing
@@ -714,30 +699,13 @@ void PrimitiveProcessor::p_Dictionary(const DictInput* in, vector<uint8_t>* out,
             nextSig(in->NVALS, in->tokens, &sigptr, in->OutputType,
                     (in->InputFlags ? true : false), skipNulls))
     {
-
-        string sig_utf8;
-
-        if (utf8)
-        {
-            string tmpString((char*)sigptr.data, sigptr.len);
-            sig_utf8 = tmpString;
-        }
-
         // do aggregate processing
         if (in->OutputType & OT_AGGREGATE)
         {
             // len == 0 indicates this is the first pass
             if (max.len != 0)
             {
-                if (utf8 )
+                tmp = cs->strnncoll(sigptr.data, sigptr.len, max.data, max.len);
-                {
-                    string max_utf8((char*)max.data, max.len);
-                    tmp = utf8::idb_strcoll(sig_utf8.c_str(), max_utf8.c_str());
-                }
-                else
-                {
-                    tmp = strncmp((char*)sigptr.data, (char*)max.data, std::min(sigptr.len, max.len));
-                }
 
                 if (tmp > 0)
                     max = sigptr;
@@ -747,15 +715,7 @@ void PrimitiveProcessor::p_Dictionary(const DictInput* in, vector<uint8_t>* out,
 
             if (min.len != 0)
             {
-                if (utf8)
+                tmp = cs->strnncoll(sigptr.data, sigptr.len, min.data, min.len);
-                {
-                    string min_utf8((char*)min.data, min.len);
-                    tmp = utf8::idb_strcoll(sig_utf8.c_str(), min_utf8.c_str());
-                }
-                else
-                {
-                    tmp = strncmp((char*)sigptr.data, (char*)min.data, std::min(sigptr.len, min.len));
-                }
 
                 if (tmp < 0)
                     min = sigptr;
@@ -788,15 +748,6 @@ void PrimitiveProcessor::p_Dictionary(const DictInput* in, vector<uint8_t>* out,
         for (filterIndex = 0; filterIndex < in->NOPS; filterIndex++)
         {
             filter = reinterpret_cast<const DictFilterElement*>(&in8[filterOffset]);
-            string filt_utf8;
-            size_t filt_utf8_len = 0;
-
-            if (utf8)
-            {
-                string tmpString((const char*)filter->data, filter->len);
-                filt_utf8 = tmpString;
-                filt_utf8_len = filt_utf8.length();
-            }
 
             if (filter->COP & COMPARE_LIKE)
             {
@@ -807,18 +758,7 @@ void PrimitiveProcessor::p_Dictionary(const DictInput* in, vector<uint8_t>* out,
             }
             else
             {
-                if (utf8)
+                tmp = cs->strnncoll(sigptr.data, sigptr.len, filter->data, filter->len);
-                {
-                    size_t sig_utf8_len = sig_utf8.length();
-                    tmp = utf8::idb_strcoll(sig_utf8.c_str(), filt_utf8.c_str());
-                    cmpResult = compare(tmp, filter->COP, sig_utf8_len, filt_utf8_len);
-                }
-                else
-                {
-                    tmp = strncmp((const char*) sigptr.data, (const char*)filter->data,
-                                  std::min(sigptr.len, static_cast<int>(filter->len)));
-                }
-
                 cmpResult = compare(tmp, filter->COP, sigptr.len, filter->len);
             }
 

primitives/linux-port/primitiveprocessor.h

@@ -185,7 +185,7 @@ public:
      * @note Throws logic_error if the output buffer is too small for the result.
      */
     void p_TokenByScan(const TokenByScanRequestHeader* t,
-                       TokenByScanResultHeader* out, unsigned outSize, bool utf8,
+                       TokenByScanResultHeader* out, unsigned outSize,
                        boost::shared_ptr<DictEqualityFilter> eqFilter);
 
     /** @brief The p_IdxWalk primitive processor
@@ -264,8 +264,9 @@ public:
      */
 //	void p_ColAggregate(const NewColAggRequestHeader *in, NewColAggResultHeader *out);
 
-    void p_Dictionary(const DictInput* in, std::vector<uint8_t>* out, bool utf8,
+    void p_Dictionary(const DictInput* in, std::vector<uint8_t>* out,
-                      bool skipNulls, boost::shared_ptr<DictEqualityFilter> eqFilter,
+                      bool skipNulls, uint32_t charsetNumber, 
+                      boost::shared_ptr<DictEqualityFilter> eqFilter,
                       uint8_t eqOp);
 
     inline void setLogicalBlockMode(bool b)

primitives/primproc/dictstep.cpp

@@ -99,6 +99,7 @@ void DictStep::createCommand(ByteStream& bs)
     else
         bs >> filterString;
     
+    bs >> charsetNumber;
 #if 0
     cout << "see " << filterCount << " filters\n";
     DictFilterElement* filters = (DictFilterElement*) filterString.buf();
@@ -173,8 +174,7 @@ void DictStep::issuePrimitive(bool isFilter)
     }
 
     bpp->pp.setLikeFilter(likeFilter);
-    // MCOL-3536 We shouldn't need to pass in utf8 -- maybe??
+    bpp->pp.p_Dictionary(primMsg, &result, isFilter, charsetNumber, eqFilter, eqOp);
-    bpp->pp.p_Dictionary(primMsg, &result, true, isFilter, eqFilter, eqOp);
 }
 
 void DictStep::copyResultToTmpSpace(OrderedToken* ot)

primitives/primproc/dictstep.h

@@ -141,6 +141,7 @@ private:
     messageqcpp::ByteStream filterString;
     uint32_t filterCount;
     uint32_t bufferSize;
+    uint32_t charsetNumber;
     uint16_t inputRidCount;
     
     bool hasEqFilter;

primitives/primproc/primitiveserver.cpp

@@ -1154,7 +1154,6 @@ int DictScanJob::operator()()
     PrimitiveProcessor pproc(gDebugLevel);
     TokenByScanResultHeader* output;
     QueryContext verInfo;
-    bool bUtf8;
 
     try
     {
@@ -1166,25 +1165,6 @@ int DictScanJob::operator()()
         *fByteStream >> verInfo;
         cmd = (TokenByScanRequestHeader*) fByteStream->buf();
 
-        // If charset is one of those that can be representedby standard ascii,
-        // we can get a performance improvement by using strcmp rather than
-        // the full charset compare system.
-        switch (cmd->charsetNumber) 
-        {
-            case 8:  // latin1_swedish_ci
-            case 9:  // latin2_general_ci
-            case 11: // ascii_general_ci
-            case 47: // latin1_bin
-            case 48: // latin1_general_ci
-            case 49: // latin1_general_cs
-            case 65: // ascii_bin
-            case 77: // latin2_bin
-                bUtf8 = false;
-                break;
-            default:
-                bUtf8 = true;
-        }
-        
         session = cmd->Hdr.SessionID;
         uniqueId = cmd->Hdr.UniqueID;
         runCount = cmd->Count;
@@ -1229,8 +1209,7 @@ int DictScanJob::operator()()
                       fLBIDTraceOn,
                       session);
             pproc.setBlockPtr((int*) data);
-            // MCOL-3536 We shouldn't need to pass in utf8 -- maybe??
+            pproc.p_TokenByScan(cmd, output, output_buf_size, eqFilter);
-            pproc.p_TokenByScan(cmd, output, output_buf_size, bUtf8, eqFilter);
 
             if (wasBlockInCache)
                 output->CacheIO++;

utils/funcexp/func_between.cpp

@@ -22,6 +22,10 @@
 *
 ****************************************************************************/
 
+#include <mariadb.h>
+#undef set_bits  // mariadb.h defines set_bits, which is incompatible with boost
+#include <my_sys.h>
+
 #include <cstdlib>
 #include <string>
 using namespace std;
@@ -53,16 +57,16 @@ inline bool numericLE(result_t op1, result_t op2)
     return op1 <= op2;
 }
 
-inline bool strGE(const string& op1, const string& op2)
+inline bool strGE(uint32_t charsetNumber, const string& op1, const string& op2)
 {
-    //return strcoll(op1.c_str(), op2.c_str()) >= 0;
+    const CHARSET_INFO* cs = get_charset(charsetNumber, MYF(MY_WME));    
-    return utf8::idb_strcoll(op1.c_str(), op2.c_str()) >= 0;
+    return cs->strnncoll(op1.c_str(), op1.length(), op2.c_str(), op2.length()) >= 0;
 }
 
-inline bool strLE(const string& op1, const string& op2)
+inline bool strLE(uint32_t charsetNumber, const string& op1, const string& op2)
 {
-    //return strcoll(op1.c_str(), op2.c_str()) <= 0;
+    const CHARSET_INFO* cs = get_charset(charsetNumber, MYF(MY_WME));    
-    return utf8::idb_strcoll(op1.c_str(), op2.c_str()) <= 0;
+    return cs->strnncoll(op1.c_str(), op1.length(), op2.c_str(), op2.length()) <= 0;
 }
 
 inline bool getBool(rowgroup::Row& row,
@@ -256,16 +260,16 @@ inline bool getBool(rowgroup::Row& row,
 
             if (notBetween)
             {
-                if (!strGE(val, pm[1]->data()->getStrVal(row, isNull)) && !isNull)
+                if (!strGE(ct.charsetNumber, val, pm[1]->data()->getStrVal(row, isNull)) && !isNull)
                     return true;
 
                 isNull = false;
-                return (!strLE(val, pm[2]->data()->getStrVal(row, isNull)) && !isNull);
+                return (!strLE(ct.charsetNumber, val, pm[2]->data()->getStrVal(row, isNull)) && !isNull);
             }
             
             return !isNull &&
-                   strGE(val, pm[1]->data()->getStrVal(row, isNull)) &&
+                   strGE(ct.charsetNumber, val, pm[1]->data()->getStrVal(row, isNull)) &&
-                   strLE(val, pm[2]->data()->getStrVal(row, isNull));
+                   strLE(ct.charsetNumber, val, pm[2]->data()->getStrVal(row, isNull));
         }
 
         default:

utils/funcexp/func_char_length.cpp

@@ -21,6 +21,10 @@
 *
 ****************************************************************************/
 
+#include <mariadb.h>
+#undef set_bits  // mariadb.h defines set_bits, which is incompatible with boost
+#include <my_sys.h>
+
 #include <cstdlib>
 #include <string>
 #include <sstream>
@@ -78,15 +82,12 @@ int64_t Func_char_length::getIntVal(rowgroup::Row& row,
         case execplan::CalpontSystemCatalog::UDECIMAL:
         {
             const string& tstr = parm[0]->data()->getStrVal(row, isNull);
-
             if (isNull)
                 return 0;
-
+            const char* b = tstr.c_str();
-            size_t strwclen = utf8::idb_mbstowcs(0, tstr.c_str(), 0) + 1;
+            const char* e = tstr.c_str() + tstr.length();
-            wchar_t* wcbuf = new wchar_t[strwclen];
+            const CHARSET_INFO* cs = get_charset(parm[0]->data()->resultType().charsetNumber, MYF(MY_WME));
-            strwclen = utf8::idb_mbstowcs(wcbuf, tstr.c_str(), strwclen);
+            return (int64_t)cs->numchars(b, e);
-            delete [] wcbuf;
-            return (int64_t)strwclen;
         }
 
         case execplan::CalpontSystemCatalog::DATE:

utils/funcexp/func_concat_ws.cpp

@@ -54,7 +54,6 @@ string Func_concat_ws::getStrVal(Row& row,
     if (isNull)
         return "";
 
-#ifdef STRCOLL_ENH__
     wstring wstr;
     size_t strwclen = utf8::idb_mbstowcs(0, delim.c_str(), 0) + 1;
     wchar_t* wcbuf = new wchar_t[strwclen];
@@ -96,7 +95,7 @@ string Func_concat_ws::getStrVal(Row& row,
     delete [] wcbuf;
     return ret;
 
-#else
+#if 0
     string str;
     string tmp;
     for ( uint32_t i = 1 ; i < parm.size() ; i++)