module/nominatim.c

   1 #include "postgres.h"
   2 #include "fmgr.h"
   3 #include "mb/pg_wchar.h"
   4 #include <utfasciitable.h>
   5
   6 #ifdef PG_MODULE_MAGIC
   7 PG_MODULE_MAGIC;
   8 #endif
   9
  10 Datum transliteration( PG_FUNCTION_ARGS );
  11 Datum gettokenstring( PG_FUNCTION_ARGS );
  12 void str_replace(char* buffer, int* len, int* changes, char* from, int fromlen, char* to, int tolen, int);
  13 void str_dupspaces(char* buffer);
  14
  15 PG_FUNCTION_INFO_V1( transliteration );
  16 Datum
  17 transliteration( PG_FUNCTION_ARGS )
  18 {
  19         static char * ascii = UTFASCII;
  20         static uint16 asciilookup[65536] = UTFASCIILOOKUP;
  21         char * asciipos;
  22
  23         text *source;
  24         unsigned char *sourcedata;
  25         int sourcedatalength;
  26
  27         unsigned int c1,c2,c3,c4;
  28         unsigned int * wchardata;
  29         unsigned int * wchardatastart;
  30
  31         text *result;
  32         unsigned char *resultdata;
  33         int resultdatalength;
  34         int iLen;
  35
  36         if (GetDatabaseEncoding() != PG_UTF8)
  37         {
  38                 ereport(ERROR,
  39                                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
  40                                          errmsg("requires UTF8 database encoding")));
  41         }
  42
  43         if (PG_ARGISNULL(0))
  44         {
  45                 PG_RETURN_NULL();
  46         }
  47
  48         // The original string
  49         source = PG_GETARG_TEXT_P(0);
  50         sourcedata = (unsigned char *)VARDATA(source);
  51         sourcedatalength = VARSIZE(source) - VARHDRSZ;
  52
  53         // Intermediate wchar version of string
  54         wchardatastart = wchardata = (unsigned int *)palloc((sourcedatalength+1)*sizeof(int));
  55
  56         // Based on pg_utf2wchar_with_len from wchar.c
  57         while (sourcedatalength > 0 && *sourcedata)
  58         {
  59                 if ((*sourcedata & 0x80) == 0)
  60                 {
  61                         *wchardata = *sourcedata++;
  62                         wchardata++;
  63                         sourcedatalength--;
  64                 }
  65                 else if ((*sourcedata & 0xe0) == 0xc0)
  66                 {
  67                         if (sourcedatalength < 2) break;
  68                         c1 = *sourcedata++ & 0x1f;
  69                         c2 = *sourcedata++ & 0x3f;
  70                         *wchardata = (c1 << 6) | c2;
  71                         wchardata++;
  72                         sourcedatalength -= 2;
  73                 }
  74                 else if ((*sourcedata & 0xf0) == 0xe0)
  75                 {
  76                         if (sourcedatalength < 3) break;
  77                         c1 = *sourcedata++ & 0x0f;
  78                         c2 = *sourcedata++ & 0x3f;
  79                         c3 = *sourcedata++ & 0x3f;
  80                         *wchardata = (c1 << 12) | (c2 << 6) | c3;
  81                         wchardata++;
  82                         sourcedatalength -= 3;
  83                 }
  84                 else if ((*sourcedata & 0xf8) == 0xf0)
  85                 {
  86                         if (sourcedatalength < 4) break;
  87                         c1 = *sourcedata++ & 0x07;
  88                         c2 = *sourcedata++ & 0x3f;
  89                         c3 = *sourcedata++ & 0x3f;
  90                         c4 = *sourcedata++ & 0x3f;
  91                         *wchardata = (c1 << 18) | (c2 << 12) | (c3 << 6) | c4;
  92                         wchardata++;
  93                         sourcedatalength -= 4;
  94                 }
  95                 else if ((*sourcedata & 0xfc) == 0xf8)
  96                 {
  97                         // table does not extend beyond 4 char long, just skip
  98                         if (sourcedatalength < 5) break;
  99                         sourcedatalength -= 5;
 100                 }
 101                 else if ((*sourcedata & 0xfe) == 0xfc)
 102                 {
 103                         // table does not extend beyond 4 char long, just skip
 104                         if (sourcedatalength < 6) break;
 105                         sourcedatalength -= 6;
 106                 }
 107                 else
 108                 {
 109                         // assume lenngth 1, silently drop bogus characters
 110                         sourcedatalength--;
 111                 }
 112         }
 113         *wchardata = 0;
 114
 115         // calc the length of transliteration string
 116         resultdatalength = 0;
 117         wchardata = wchardatastart;
 118         while(*wchardata)
 119         {
 120                 if (*(asciilookup + *wchardata) > 0) resultdatalength += *(ascii + *(asciilookup + *wchardata));
 121                 wchardata++;
 122         }
 123
 124         // allocate & create the result
 125         result = (text *)palloc(resultdatalength + VARHDRSZ);
 126         SET_VARSIZE(result, resultdatalength + VARHDRSZ);
 127         resultdata = (unsigned char *)VARDATA(result);
 128
 129         wchardata = wchardatastart;
 130         while(*wchardata)
 131         {
 132                 if (*(asciilookup + *wchardata) > 0)
 133                 {
 134                         asciipos = ascii + *(asciilookup + *wchardata);
 135                         for(iLen = *asciipos; iLen > 0; iLen--)
 136                         {
 137                                 asciipos++;
 138                                 *resultdata = *asciipos;
 139                                 resultdata++;
 140                         }
 141                 }
 142                 else
 143                 {
 144                         ereport( WARNING, ( errcode( ERRCODE_SUCCESSFUL_COMPLETION ),
 145                               errmsg( "missing char: %i\n", *wchardata )));
 146
 147                 }
 148                 wchardata++;
 149         }
 150
 151         pfree(wchardatastart);
 152
 153         PG_RETURN_TEXT_P(result);
 154 }
 155
 156 void str_replace(char* buffer, int* len, int* changes, char* from, int fromlen, char* to, int tolen, int isspace)
 157 {
 158         char *p;
 159
 160         // Search string is too long to be pressent
 161         if (fromlen > *len) return;
 162
 163         p = strstr(buffer, from);
 164         while(p)
 165         {
 166                 if (!isspace || *(p-1) != ' ')
 167                 {
 168                         (*changes)++;
 169                         if (tolen != fromlen) memmove(p+tolen, p+fromlen, *len-(p-buffer)+1);
 170                         memcpy(p, to, tolen);
 171                         *len += tolen - fromlen;
 172                 }
 173                 p = strstr(p+1, from);
 174         }
 175 }
 176
 177 void str_dupspaces(char* buffer)
 178 {
 179         char *out;
 180         int wasspace;
 181
 182         out = buffer;
 183         wasspace = 0;
 184         while(*buffer)
 185         {
 186                 if (wasspace && *buffer != ' ') wasspace = 0;
 187                 if (!wasspace)
 188                 {
 189                         *out = *buffer;
 190                         out++;
 191                         wasspace = (*buffer == ' ');
 192                 }
 193                 buffer++;
 194         }
 195         *out = 0;
 196 }
 197
 198 PG_FUNCTION_INFO_V1( gettokenstring );
 199 Datum
 200 gettokenstring( PG_FUNCTION_ARGS )
 201 {
 202         text *source;
 203         unsigned char *sourcedata;
 204         int sourcedatalength;
 205
 206         char * buffer;
 207         int len;
 208         int changes;
 209
 210         text *result;
 211
 212         if (GetDatabaseEncoding() != PG_UTF8)
 213         {
 214                 ereport(ERROR,
 215                                         (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
 216                                          errmsg("requires UTF8 database encoding")));
 217         }
 218
 219         if (PG_ARGISNULL(0))
 220         {
 221                 PG_RETURN_NULL();
 222         }
 223
 224         // The original string
 225         source = PG_GETARG_TEXT_P(0);
 226         sourcedata = (unsigned char *)VARDATA(source);
 227         sourcedatalength = VARSIZE(source) - VARHDRSZ;
 228
 229         // Buffer for doing the replace in - string could get slightly longer (double is mastive overkill)
 230         buffer = (char *)palloc((sourcedatalength*2)*sizeof(char));
 231         memcpy(buffer+1, sourcedata, sourcedatalength);
 232         buffer[0] = 32;
 233         buffer[sourcedatalength+1] = 32;
 234         buffer[sourcedatalength+2] = 0;
 235         len = sourcedatalength+3;
 236
 237         changes = 1;
 238         str_dupspaces(buffer);
 239         while(changes)
 240         {
 241                 changes = 0;
 242                 #include <tokenstringreplacements.inc>
 243                 str_dupspaces(buffer);
 244         }
 245
 246         // 'and' in various languages
 247         str_replace(buffer, &len, &changes, " and ", 5, " ", 1, 0);
 248         str_replace(buffer, &len, &changes, " und ", 5, " ", 1, 0);
 249         str_replace(buffer, &len, &changes, " en ", 4, " ", 1, 0);
 250         str_replace(buffer, &len, &changes, " et ", 4, " ", 1, 0);
 251         str_replace(buffer, &len, &changes, " e ", 3, " ", 1, 0);
 252         str_replace(buffer, &len, &changes, " y ", 3, " ", 1, 0);
 253
 254         // 'the' (and similar)
 255         str_replace(buffer, &len, &changes, " the ", 5, " ", 1, 0);
 256         str_replace(buffer, &len, &changes, " der ", 5, " ", 1, 0);
 257         str_replace(buffer, &len, &changes, " den ", 5, " ", 1, 0);
 258         str_replace(buffer, &len, &changes, " die ", 5, " ", 1, 0);
 259         str_replace(buffer, &len, &changes, " das ", 5, " ", 1, 0);
 260         str_replace(buffer, &len, &changes, " la ", 4, " ", 1, 0);
 261         str_replace(buffer, &len, &changes, " le ", 4, " ", 1, 0);
 262         str_replace(buffer, &len, &changes, " el ", 4, " ", 1, 0);
 263         str_replace(buffer, &len, &changes, " il ", 4, " ", 1, 0);
 264
 265         // german
 266         str_replace(buffer, &len, &changes, "ae", 2, "a", 1, 0);
 267         str_replace(buffer, &len, &changes, "oe", 2, "o", 1, 0);
 268         str_replace(buffer, &len, &changes, "ue", 2, "u", 1, 0);
 269         str_replace(buffer, &len, &changes, "sss", 3, "ss", 2, 0);
 270         str_replace(buffer, &len, &changes, "ih", 2, "i", 1, 0);
 271         str_replace(buffer, &len, &changes, "eh", 2, "e", 1, 0);
 272
 273         // russian
 274         str_replace(buffer, &len, &changes, "ie", 2, "i", 1, 0);
 275         str_replace(buffer, &len, &changes, "yi", 2, "i", 1, 0);
 276
 277         // allocate & create the result
 278         len--;// Drop the terminating zero
 279         result = (text *)palloc(len + VARHDRSZ);
 280         SET_VARSIZE(result, len + VARHDRSZ);
 281         memcpy(VARDATA(result), buffer, len);
 282
 283         pfree(buffer);
 284
 285         PG_RETURN_TEXT_P(result);
 286 }
 287