Hi, Jan!
On Mar 23, Jan Lindström wrote:
> Hi,
>
> All below is correct and naturally you may re-implement duplicate search on
> different index type. I think there is more code that needs change as this
> new index type would contain a column (hash value of the blob field) that
> is not on stored on base table, right ? Indexed columns are also stored
> inside a InnoDB data dictionary persistently, so that part would also need
> change (dict/dict0*.cc files). Actually, you could store blob has also to
> table, it could make things easier. Secondly, remember that unique keys can
> be used inside a InnoDB as foreign keys, this is again a design question,
> do you allow blobs to be foreign keys or not. Finally, unique key with NOT
> NULL can be used as primary key i.e. clustered key on InnoDB, using blobs
> hash on that might be out of reach on this timetable.
>
> R: Jan Lindström
> Principal Engineer
> InnoDB
Agree, blobs (even UNIQUE NOT NULL) cannot be used as primary keys.
For simplicity I would not support foreign keys either, this can be
added later.