Released

I was able to find a solution:

1) I created a new CustomConverter implementation which maps the raw fields' bytes into the desirable string using a configurable encoding. This works out of the box with the BinlogReader since the BinlogReader returns the fields' raw bytes. 

2) I configured my connector with "database.characterSetResults": "NULL" to override Debezium's default UTF-8 character_set_results because I want the raw bytes from the field. 

3) I patched (https://github.com/debezium/debezium/pull/1909) SnapshotReader's readField to return the raw bytes for char-type fields.

1) Actual data is lost because the CHAR field value is encoded using the wrong charset – presumably the default JDBC URL charset setting. Still investigating that.

2) that sounds like a good idea. I can take a look at that.

Hi,

1) What kind of information is lost? Data or metadata?
2) RelationaColumn could be extended to provide data from io.debezium.relational.Column.charsetName()

J.

jpechane I like the idea and I think that generally that's the way to go.

Couple of problems I encountered while testing it out: 

1) The SnapshotReader passes the field's value to the converter as a String (ignoring the column's charset altogether from what I can tell) with some of the information lost.

2) The BinlogReader passes the raw bytes of the field to the converter which is great but I don't see a way to get the column's original character set. I tried RelationalColumn#typeExpression which according to the JavaDocs should give me the character set as well but I'm only getting the type name (CHAR).

Any pointers? 

Hi,

this sounds like a perfect fit for https://debezium.io/documentation/reference/1.3/development/converters.html