Hi Jan,
yes, this is a bug.
What is missing IMO is calling reset() in the getter of DBClobData.getReader()
e.g. like this:
/**
* Get the Reader for the large string
*
* @return Returns the reader with the character data for the CLOB
*/
public Reader getReader()
{
try
{
reader.reset();
return reader;
}
catch (IOException e)
{
throw new InternalException(e);
}
}
Would you like to create a JIRA issue for that?
Regards,
Rainer
> from: [email protected] [mailto:[email protected]]
> to: [email protected]
> subject: Re-usable DBCommand with DBClobData
>
> Hello,
>
> I used to write an importer in Empire-db that has a CLOB (LONGTEXT)
> column. I wanted to update existing and insert new items, so I did:
>
> cmd.set(DB.col1.to(col1));
> cmd.set(DB.col2.to(col2));
> // ...
> cmd.set(DB.clob_col.to(string));
>
> // try update
> int updateCount = this.db.executeUpdate(cmd, this.conn);
>
> if (updateCount < 1) {
> // not found, insert
> cmd.set(DB.id.to(id));
> this.db.executeInsert(cmd, this.conn);
> }
>
> which is, afaik, the way to do it.
>
> Turns out: on the first run clob_col is always empty. I debugged and
> everything looked fine. The same code is used for INSERT and UPDATE, only
> the generated statement is different.
>
> Now I tracked it down to DBClobData, which uses a java.io.Reader internally.
> That reader is consumed on the first try (UPDATE) so its empty on the second
> try (UPDATE).
>
> AFAIK the DBCommand is designed to be used that way. So I think this is a
> bug. What do you think?
>
> - jan
