Dne 28.12.2011 17:21, Milan Babuskov napsal(a):

> Josef Kokeš wrote:
>> I am trying to convert a BLOB (unlimited size, theoretically) to a long
>> VARCHAR (say, VARCHAR(8192)). This should be easy enough to do:
>>
>> DECLARE VARIABLE b BLOB; /* input */
>
> Which subtype of BLOB? IIRC, BLOB subtype 1 (text) should support
> character sets, so you don't have to lose non-ASCII characters.
>
> If it's a binary blob (subtype 0) you can also try to dump the contents
> to a file (using FBExport or your code) and then import it back into a
> new subtype 1 blob.

I need to do it on the run, not once, so dump is not an option. With
subtype 0, too.

Anyway, it turned out the UDF approach works after all, though I really
don't see why the new version works while the version I posted in the
original question doesn't - as far as I can tell the only difference is
that I removed the initialization of the return field.

Pepak