TM1 Forum

Hi All,
Good day.
I have some doubts related to Public view with private subset. Tried to google and search the forum but didn't find any related thread.
Hope someone can share some pointers with me.

For example, I have a TI which working with a view A (it can be used for purpose of copying or zerorizing).
For this view A, some of the dimension is saved without giving a subset name and shown as private (A save is done when creating the view without assign subset to the dimension). I have uploaded an example for the view in property window.

For this type of views, I don't think it is a good practice to be used in TI development.
May I know what could be the impact for TI that having this kind of views?
The most possible impact come across my mind is the result of "thrashing" for parallel interaction.
Any input is welcome. Thanks in advance for the sharing. =)

Thanks and regards,
CJ

Hi CJ,

I think the icons are a bit misleading here. The three All subsets shown are not so much private subsets, but in fact implicit unnamed subsets (instead of explicit named subsets) which only exist in the context of that view. It would not be possible to save a public view using named private subsets.
If you indeed need all elements selected for the regarding dimensions, then in my opinion there's nothing wrong with saving the view like that.

Michel

cjsee wrote:For this type of views, I don't think it is a good practice to be used in TI development.
May I know what could be the impact for TI that having this kind of views?
The most possible impact come across my mind is the result of "thrashing" for parallel interaction.
Any input is welcome. Thanks in advance for the sharing. =)

Michel has already given you the correct answer, "All" doesn't refer to a private subset but to the implicit TM1SubsetAll. So I can't give you an answer to that question, but I will give you an opinion on "what could be the impact for TI that having this kind of views". I wouldn't be so concerned about thrashing/contention/rollback/locking but about seeing incorrect and unexpected results from the view or subsets being unwittingly or unintentionally edited. For this reason I would never use a persistent view as either a data source or zero out view in a production system but would always go with views generated by TI code using a string filter of the dimensions/elements to include. Bedrock can be used to do all the heavy lifting for such view creation so can be implemented without needing much technical knowledge. It is a much better option (in my opinion) than a manually created view.

I would agree, as TI views have to be public and normally cover large amounts of data a user could try to open them which could lead to memory issues on their machine. Not only should you use TI code to create the views / subsets, you should always remove them when the process is finished with them. I'm not sure if Bedrock covers that as well?

Thanks all for the replies and your interpretation is right. They are unnamed subset when saving the public view.
Sorry for the misleading screen capture. I think I should call them as "internal subset" instead.

Yes, I am totally agree with lotsaram that this should not be done in TI development.
However, it seems that it doesn't cause much issues to TM1 in CX9.0, so this was used quite commonly in a lot of TI in my productions.
I think apart from the unintentional edit for the persistent view, there could have some unanticipated behavior because of this.
As from my checking, there are cases that data not being zero-rised properly and I can't really find any concrete reason for this by tracing the server/transactions logs. The most obvious thing I able to notice is "thrashing" could happen as I can find from log that sometimes TI has been triggered more than the finished count.

Again, many thanks to all! =)

Cheers,
CJ