Oracle: Convert xml entities in a varchar2 field to utf-8 characters

Question

I have a field in a table which holds XML entities for special characters, since the table is in latin-1. E.g. "Hallöle slovenčina" (the "ö" is in latin-1, but the "č" in "slovenčina" had to be converted to an entity by some application that stores the values into the database)

Now I need to export the table into a utf-8 encoded file by converting the XML entities to their original characters.

Is there a function in Oracle that might handle this for me, or do I really need to create a huge key/value map for that?

Any help is greatly appreciated.

EDIT: I found the function DBMS_XMLGEN.convert, but it only works on <,> and &. Not on &#NNN; :-(

Answer 1

I believe the problem with dbms_xmlgen is that there are technically only five XML entities. Your example has a numeric HTML entity, which corresponds with Unicode:

http://theorem.ca/~mvcorks/cgi-bin/unicode.pl.cgi?start=0100&end=017F

Oracle has a function UNISTR which is helpful here:

select unistr('sloven\010dina') from dual;

I've converted 269 to its hex equivalent 010d in the example above (in Unicode it is U+010D). However, you could pass a decimal number and do a conversion like this:

select unistr('sloven\' || replace(to_char(269, 'xxx'), ' ', '0') || 'ina') from dual;

EDIT: The PL/SQL solution:

Here's an example I've rigged up for you. This should loop over and replace any occurrences for each row you select out of your table(s).

create table html_entities (
    id NUMBER(3),
    text_row VARCHAR2(100)
);

INSERT INTO html_entities 
VALUES (1, 'Hallöle slovenčina Ċ ú');

INSERT INTO html_entities 
VALUES (2, 'I like the letter Ċ');

INSERT INTO html_entities 
VALUES (3, 'Nothing to change here.');

DECLARE
    v_replace_str NVARCHAR2(1000);
    v_fh UTL_FILE.FILE_TYPE;       
BEGIN
    --v_fh := utl_file.fopen_nchar(LOCATION IN VARCHAR2, FILENAME IN VARCHAR2, OPEN_MODE IN VARCHAR2, MAX_LINESIZE IN BINARY_INTEGER);

    FOR v_rec IN (select id, text_row from html_entities) LOOP
        v_replace_str := v_rec.text_row;
        WHILE (REGEXP_INSTR(v_replace_str, '&#[0-9]+;') <> 0) LOOP
            v_replace_str := REGEXP_REPLACE(
                v_replace_str, 
                '&#([0-9]+);',
                unistr('\' || replace(to_char(to_number(regexp_replace(v_replace_str, '.*?&#([0-9]+);.*$', '\1')), 'xxx'), ' ', '0')),
                1,
                1
            );
        END LOOP;

        -- utl_file.put_line_nchar(v_fh, v_replace_str);
        dbms_output.put_line(v_replace_str);

    END LOOP;
    --utl_file.fclose(v_fh);
END;
/

Notice that I've stubbed in calls to the UTL_FILE function to write NVARCHAR lines (Oracle's extended character set) to a file on the database server. The dbms_output, while great for debugging, doesn't seem to support extended characters, but this shouldn't be a problem if you use UTL_FILE to write to a file. Here's the DBMS_OUTPUT:

Hallöle slovencina C ú
I like the letter C
Nothing to change here.

Answer 2

You can also just use the internationalization package :

UTL_I18N.unescape_reference ('text')

Works great in changing those html entities to normal characters (such as cleanup after moving a database from iso 8859P1 to UTF-8)