X-Git-Url: https://git.plutz.net/?a=blobdiff_plain;f=parsers%2F40_indexer.sh;h=e3065c14aa11b8c5c82b357188f008ce48223e3a;hb=d9a3dee90a669c88bd6b4b0fa553e01ed1e729a1;hp=ec7be4a7f4a6d9fa17da3b19ed762c814ce9b78b;hpb=0befccc1274549229b77d50c1b8b24d6c4cf5403;p=shellwiki

diff --git a/parsers/40_indexer.sh b/parsers/40_indexer.sh
index ec7be4a..e3065c1 100755
--- a/parsers/40_indexer.sh
+++ b/parsers/40_indexer.sh
@@ -34,7 +34,13 @@ mkdir -p "$I"
   printf '%s\n' "$line"
 done \
 | awk '
-  BEGIN { FS = "([] \t\n\r!\"#'\''()*+,./:;<=>?\\^_`{|}~[-]|%[0-9A-Fa-f]{2})+" }
+  BEGIN { # Field separator FS should include punctuation, including Unicode Block U+2000 - U+206F
+          if ( length("¡") == 1 )  # Utf-8 aware AWK
+          FS = "([] \t\n\r!\"#'\''()*+,./:;<=>?\\^_`{|}~[-]|%[0-9A-Fa-f]{2}|'"$(printf '[\342\200\200-\342\201\257]')"')+";
+          else                     # UTF-8 Hack
+          FS = "([] \t\n\r!\"#'\''()*+,./:;<=>?\\^_`{|}~[-]|%[0-9A-Fa-f]{2}|'"$(printf '\342\200[\200-\277]|\342\201[\201-\257]')"')+";
+          fi
+        }
         { for (n = 1; n <= NF; n++) {
             if ( $n != "" && length($n) <= 128 ) {
               words[tolower($n)]++; total++;