pike.git/ lib/ modules/ Sql.pmod/ mysql.pike

2006-09-15

2006-09-15 13:08:50 by Martin Stjernholm <mast@lysator.liu.se>

• e7bdd005dc243731cbe727752a1019c16ef04b8b (80 lines) (+40/-40) ^{[ Show | Annotate ]}
  Branch: 7.9

Reverted automatic switching of character_set_connection since it
isn't side effect free.

E.g. the expression 'ü' < 'v' would evaluate differently depending on
whether the query is sent as latin1 or utf-8.

Another example:

insert into foo values ('<text data>', '<binary data>');

Here <binary data> would be clobbered if <text data> happen to contain
wide chars.

So it's simply not possible to get transparent unicode support in
MySQL. Deal with it. :\

Also tuned the MYSQL_CHARSET_DEBUG a little.

Rev: lib/modules/Sql.pmod/mysql.pike:1.31

1:   /* -  * $Id: mysql.pike,v 1.30 2006/08/22 11:27:08 grubba Exp $ +  * $Id: mysql.pike,v 1.31 2006/09/15 13:08:50 mast Exp $    *    * Glue for the Mysql-module    */
16:   #define UNICODE_DECODE_MODE 1 // Unicode decode mode   #define LATIN1_UNICODE_ENCODE_MODE 2 // Unicode encode mode with latin1 charset   #define UTF8_UNICODE_ENCODE_MODE 4 // Unicode encode mode with utf8 charset - #define BINARY_LATIN1_MODE 8 // Don't special-case latin1 control chars +       #ifdef MYSQL_CHARSET_DEBUG   #define CH_DEBUG(X...) werror("Sql.mysql: " + X)
24:   #define CH_DEBUG(X...)   #endif    - // Set to the above if the connection is in utf8-mode. Enable latin1 - // unicode encode mode by default; it should be compatible with - // earlier pike versions. + // Set to the above if the connection is requested to be in one of the + // unicode modes. latin1 unicode encode mode is enabled by default; it + // should be compatible with earlier pike versions.   static int utf8_mode;      // The charset, either "latin1" or "utf8", currently assigned to - // character_set_client and character_set_connection when unicode - // encode mode is enabled. Zero when the connection charset has been - // set to something else than "latin1" or "unicode". + // character_set_client when unicode encode mode is enabled. Zero when + // the connection charset has been set to something else than "latin1" + // or "unicode".   static string send_charset;      static void update_unicode_encode_mode_from_charset (string charset)
42:    utf8_mode |= LATIN1_UNICODE_ENCODE_MODE;    utf8_mode &= ~UTF8_UNICODE_ENCODE_MODE;    send_charset = "latin1"; -  CH_DEBUG("Entering latin1 mode.\n"); +  CH_DEBUG ("Entering latin1 encode mode.\n");    break;    case "unicode":    utf8_mode |= UTF8_UNICODE_ENCODE_MODE;    utf8_mode &= ~LATIN1_UNICODE_ENCODE_MODE;    send_charset = "utf8"; -  CH_DEBUG("Entering utf8 mode.\n"); +  CH_DEBUG ("Entering unicode encode mode.\n");    break;    default:    // Wrong charset - the mode can't be used.    utf8_mode |= LATIN1_UNICODE_ENCODE_MODE|UTF8_UNICODE_ENCODE_MODE;    send_charset = 0; -  CH_DEBUG("Entering other mode.\n"); +  CH_DEBUG ("Not entering latin1/unicode encode mode " +  "due to incompatible charset %O.\n", charset);    break;    }   }
68:   //!   //! Unicode encode mode works as follows: Eight bit strings are sent   //! as @expr{latin1@} and wide strings are sent using @expr{utf8@}. - //! @[big_query] sends @expr{SET character_set_client@} and @expr{SET - //! character_set_connection@} statements as necessary to update the - //! charset on the server side. If the server doesn't support that - //! then it fails, but the wide string query would fail anyway. + //! @[big_query] sends @expr{SET character_set_client@} statements as + //! necessary to update the charset on the server side. If the server + //! doesn't support that then it fails, but the wide string query + //! would fail anyway.   //!   //! To make this transparent, string literals with introducers (e.g.   //! @expr{_binary 'foo'@}) are excluded from the UTF-8 encoding. This
105:   //! will get UTF-8 encoded by the server.   //!   //! @note - //! When unicode encode mode is enabled, the connection charset - //! will mirror the client charset. This is necessary for unicode - //! characters to survive for wide queries, and for binary data - //! to survive for narrow queries in a transparent manner. - //! - //! @note +    //! When unicode encode mode is enabled and the connection charset   //! is @expr{latin1@}, the charset accepted by @[big_query] is not   //! quite Unicode since @expr{latin1@} is based on @expr{cp1252@}.
123:   //! @seealso   //! @[set_unicode_decode_mode], @[set_charset]   { -  if (enable) { -  CH_DEBUG("Enabling unicode encode mode.\n"); +  if (enable)    update_unicode_encode_mode_from_charset (lower_case (get_charset())); -  } else { +  else {    utf8_mode &= ~(LATIN1_UNICODE_ENCODE_MODE|UTF8_UNICODE_ENCODE_MODE);    send_charset = 0;    CH_DEBUG("Disabling unicode encode mode.\n");
334:   string latin1_to_utf8 (string s)   //! Converts a string in MySQL @expr{latin1@} format to UTF-8.   { -  CH_DEBUG("Converting latin1 query to utf8.\n"); +     return string_to_utf8 (replace (s, ([    "\x80": "\u20AC", /*"\x81": "\u0081",*/ "\x82": "\u201A", "\x83": "\u0192",    "\x84": "\u201E", "\x85": "\u2026", "\x86": "\u2020", "\x87": "\u2021",
569:    string restore_charset; \    if (charset) { \    restore_charset = send_charset || get_charset(); \ -  CH_DEBUG("Restore charset is %O.\n", restore_charset); \ +     if (charset != restore_charset) { \ -  ::big_query("SET character_set_client=" + charset); \ -  ::big_query("SET character_set_connection=" + charset); \ +  CH_DEBUG ("Switching charset from %O to %O (due to charset arg).\n", \ +  restore_charset, charset); \ +  ::big_query ("SET character_set_client=" + charset); \ +  /* Can't be changed automatically - has side effects. /mast */ \ +  /* ::big_query("SET character_set_connection=" + charset); */ \    } else \    restore_charset = 0; \    } \
584:    if (String.width (query) == 8) \    new_send_charset = "latin1"; \    else { \ +  CH_DEBUG ("Converting (mysql-)latin1 query to utf8.\n"); \    query = utf8_encode_query (query, latin1_to_utf8); \    new_send_charset = "utf8"; \    } \
593:    if (_can_send_as_latin1 (query)) \    new_send_charset = "latin1"; \    else { \ +  CH_DEBUG ("Converting query to utf8.\n"); \    query = utf8_encode_query (query, string_to_utf8); \    new_send_charset = "utf8"; \    } \    } \ -  CH_DEBUG("New send charset is %O.\n", new_send_charset); \ +     \    if (new_send_charset != send_charset) { \ -  CH_DEBUG("Send charset was %O.\n", send_charset); \ +  CH_DEBUG ("Switching charset from %O to %O.\n", \ +  send_charset, new_send_charset); \    if (mixed err = catch { \ -  ::big_query("SET character_set_client=" + new_send_charset); \ -  ::big_query("SET character_set_connection=" + \ -  new_send_charset); \ +  ::big_query ("SET character_set_client=" + new_send_charset); \ +  /* Can't be changed automatically - has side effects. /mast */ \ +  /* ::big_query("SET character_set_connection=" + \ +  new_send_charset); */ \    }) { \    if (new_send_charset == "utf8") \    predef::error ("The query is a wide string " \    "and the MySQL server doesn't support UTF-8: %s\n", \ -  describe_error(err)); \ +  describe_error (err)); \    else \ -  throw(err); \ +  throw err; \    } \    send_charset = new_send_charset; \    } \    } \    \ -  CH_DEBUG("Sending query %O.\n", query); \ +  CH_DEBUG ("Sending query with charset %O: %O.\n", \ +  charset || send_charset, query); \    \    int|object res = ::do_query(query); \    \
625:    if (send_charset && (<"latin1", "utf8">)[charset]) \    send_charset = charset; \    else { \ -  CH_DEBUG("Restoring charset to %O.\n", restore_charset); \ -  ::big_query("SET character_set_client=" + restore_charset); \ -  ::big_query("SET character_set_connection=" + restore_charset); \ +  CH_DEBUG ("Restoring charset %O.\n", restore_charset); \ +  ::big_query ("SET character_set_client=" + restore_charset); \ +  /* Can't be changed automatically - has side effects. /mast */ \ +  /* ::big_query("SET character_set_connection=" + restore_charset); */ \    } \    } \    \    if (!objectp(res)) return res; \    \    if (utf8_mode & UNICODE_DECODE_MODE) { \ -  CH_DEBUG("Adding UnicodeWrapper.\n"); \ +  CH_DEBUG ("Using UnicodeWrapper for result.\n"); \    return .sql_util.UnicodeWrapper(res); \    } \    return res;