Hi Lisa,

you have understood my thought.


I would disagree *slightly* with you here; I can't imagine a case where the ADDITIVE keyword for APPEND MEMO would make a lot of sense for blobs although it is obviously quite useful for memo fields <s>.


If i define a my BINARY file format, and this have not a header or a footer,
then to add two blob have sense.

Another issue is this, APPEND MEMO is more powerfull of REPLACE or UPDATE.

Preface:
APPEND MEMO fieldMemoOrBlob FROM Filename

is theoretically equal to

REPLACE fieldMemoOrBlob WITH fieldMemoOrBlob+FILETOSTR(Filename)
or
UPDATE ... SET fieldMemoOrBlob = fieldMemoOrBlob+FILETOSTR(Filename)

but !! this is not true in practical:
* THIS IS ALLOWED

APPEND MEMO fieldMemoOrBlob  FROM FileWithSizeMore16M OVERWRITE
REPLACE fieldMemoOrBlob WITH FILETOSTR(FileWithSizeMore16M)

* THIS IS ALLOWED

APPEND MEMO fieldMemoOrBlob  FROM FileWithAnySize OVERWRITE
APPEND MEMO fieldMemoOrBlob  FROM FileWithASizeThatDoFieldMemoOrBlobMore16M
....

* THIS IS NOT ALLOWED ( overflow string lengh error is fired )

REPLACE fieldMemoOrBlob WITH FILETOSTR(FileWithAnySize)
REPLACE fieldMemoOrBlob WITH fieldMemoOrBlob+FILETOSTR(FileWithASizeThatDoFieldMemoOrBlobMore16M)

I have no knowledge about how this was implemented. Given Aleksey's response, however, I'm thinking that they leveraged the current memo-handling that occurs when you use NOCPTRANS (memo binary). So that would mean things like SET BLOCKSIZE and PACK MEMO might indeed impact blobs. But I am guessing.

I agree with you.

My message to Aleksey is for remove my guess.

Fabio