WHERE NOT EXISTS work not correctly

[warfollowmy_ver]

1) There are several QVD files from one structure.

2) It is necessary to download all the files through the '*.qvd (qvd)' because the file becomes larger with different names.

3) It is necessary to download everything but without 'C'.

Examples of the attachment

Example 'LOAD ALL' load all. OK.

Example 'LOAD EXISTS' load only listed with 'A' and 'B'. OK.

Example 'LOAD NOT EXISTS' load everything but without 'C'. ??? Who knows what's going on? 😃

[tresesco]

Not exists seems to be working fine. What is your expected out put?

[warfollowmy_ver]

You are joking? 😃

Obviously must be load all values without  'C' but was be loaded not all values off A and B.

Load only first rows A and B.

Please look QVD files or second and first example, look number of rows and values of A and B...

[swuehl]

Are you looking for

t:

NoConcatenate

LOAD * INLINE [

'GroupFilter'

'C'

];

t0:

LOAD * FROM *.qvd (qvd) WHERE NOT EXISTS (GroupFilter,Group);

[swuehl]

As soon as the first value e.g. of Group 'A' is loaded from T1.qvd, it does exist in the symbol table...

[tresesco]

I am certainly not joking. Exists works like that. It just checks if the value is loaded so far. That is, is value 'A' is read in the first record, it would not load from subsequent rows, because it is already loaded (or, exists). It is kind of distinct load for a field value.

Hope this helps.

[warfollowmy_ver]

Sorry. Now I understand.

[Anonymous]

try where not match() vs where not exist()

[johnw]

To get your intended behavior, what I typically do is load an extra copy of whatever the field is. So in this case, I would load an [Exclude Group] field. Then I'd load where not exists([Exclude Group],[Group]). Then I'd drop the extra field I'd created.

Edit: Which is to say I'd do it just like Stephan said. Never mind!

[johnw]

That works, at least if the list is small (as it is here), but it also breaks the optimized QVD load, so it will go much more slowly if these are large QVDs.