PA Workspace - what can it do?

[ultrakocka]

Hello,

we're starting a new project and it's been decided to do it only in PAW. My problem is I just don't see it happening as it just doesn't have many features needed to make a functional application. I've been reading through other older topics and PA Workspace documentation and just wanted to confirm that I'm right on this (PA Workspace 2.0.57).

1) It's not possible to show custom error messages after running process with active button.
2) It's not possible to directly pass a selected cell/element in dimension to a process as parameter.

Thanks!
Simon

[Steve Rowe]

Hi Simon,

Not natively no but it's easy to build your own solution for both these issues.

If you have a cube that is dimensioned by }Users and a Measure dimension containing all the references for interaction with the front end then.

For 1 you can set-up picklists in this cube that are driven by the dimension / subset of interest.
The view of the front end cube can be limited to the user viewing it using an MDX subset.
Then in the TI you can determine the user calling the TI and read the value from your front end cube.

For 2 you can write the message back to the cell in the front end cube and display this in PAW.

I find this approach much easier than setting up TIs with loads of parameters, it's much easier to write the script to read the "parameters" and then assign these values to a global variable, its much more flexible.

Should be possible in both TM1Web and PAW and don't forget web sheets can be published through PAW too.

[MarenC]

Hi,

Unless I am missing something I would have to slightly disagree with Steve here.

Yes you can set up cubes etc, but there are times when you want to pass parameters to processes.

For example we have an employee move process and when the user navigates to the Move process tab it ask=s them to provide the
employee parameter. This is so we can run some pre-processes and add the employee as an attribute to the client.

Having another cube to handle all this is just messy for the end user.

Not having the ability to pass parameters from cell references is a limitation of workspace and hampers a fully integrated planning solution as
far as I am concerned. It is also annoying that you can't pass title elements when navigating etc.

I admit you can still use websheets for some of these issues but I think workspace needs more improvement in these areas rather than any new fancy front end dashboarding that the sales guys just love!

Maren

[Elessar]

Hi,

The button parameters synchronization was well discussed here

To be short, there are 2(3) options now:

1. Use tiny websheet only with active button in it. Active button takes parameters from SUBNMs, who are hidden and can be synchronized with PAW filters.
2. Picklists and view with MDX pointing on them
3. Wait . Here IBM says that "This one of the highest priority items in our feature backlog. We are working to bring it forward"

[gtonkin]

...
1) It's not possible to show custom error messages after running process with active button.
...

For 1) above, I write various values to a process log cube. You can add the cube view to your canvass, right click on the cell containing the message then add that to the canvass by clicking "Add to sheet".

There are obviously nuances and limitations - cell is now a static reference like using a DBRW with hard-coded values for each dimension but depends on your cube design.

Check out the Ideas site to see what people are asking for and vote to get functionality we are all looking for included:
References of Cells added to sheets to be maintainable
PAW Action Button ability to get parameters from selectors
PAW Synchronise Action Button parameters with Sheet Selections

[luizg2019]

1) It's not possible to show custom error messages after running process with active button.

Since PAW 2.0.49 is possible to display custom error messages after executing a process via button.

[Steve Rowe]

Hi,
Good shout on the error message luzig2019, hadn't picked this up.

On the "passing parameters" piece. I'm not going to pretend that what we have now is perfect but working with the tools we do have I would argue that we can get to 95% of where we need to be, which is normally good enough.

I'd also argue that, for the developer, taking the approach of using picklists and global variables instead of parameters has significant advantages.

• Action buttons are much more stable since they don't have parameter references. If I need to add a parameter to a TI I don't need to update all the action buttons that reference that TI. It is also much easier to set the action buttons up in the first place.
• There is a permanent record at the user level of what the TI was called with, the last run will show in the pick list cube and if you have logging on you'll see the history in the transaction log. This is very very useful when it comes to tracking of issues and audit trail.
• Setting parameters up in a TI is a manual process, if I have the same parameter being used and passed down a chain of TIs then this is a PITA to set-up. With the picklist / global vars approach this isn't required and the script can be copied and pasted between TIs leading to development time savings, quality of life improvements and reduction of opportunity for error.
• Picklists have error checking built in versus allowing manual entry of element names or other object references. This saves time on the scripting side as we don't need error checking in each TI.
• Consistency of variable names becomes much easier to achieve / enforce with the use of Global Variables. It becomes somewhat like working with VBA with Option Explicit turned on, normally considered to be a good thing.

For example in the master TI for a call sequence I would have

StringGlobalVariable('sMonthToLoad');
sMonthToLoad=CellGetS('Front End Cube' , TM1User , 'Month To Load');

In all the downstream TIs where I need the month I put, IMO much easier than working with parameters.

StringGlobalVariable('sMonthToLoad');

So IMO parameters aren't the panacea that people imagine them to be, it is not 100% upside. They would be useful though, especially for simpler apps and in some circumstances would reduce the time to build. If anyone from IBM reads this, I'm not saying we don't need the functionality, I don't fancy waking up with a horses head in my bed!

Cheers,

[MarenC]

Hi Steve,

thanks for the further explanation. I would not like to accept or refute what you say without actually trying it!

But we would need error checking in each TI because the user might not update the cube and simply go ahead and run the process!?

Maren

[Steve Rowe]

But we would need error checking in each TI because the user might not update the cube and simply go ahead and run the process!?

Yeah that's a small wrinkle of TM1 picklists, is that a blank is allowed. Fairly easy to check for this single exception.

Interestingly when they were first implemented and for years afterwards a blank was not an allowed option in the picklist, they had to match exactly to the definition. "Everyone" used to moan loudly about this.

Eventually IBM added the possibility of selecting a blank reference (but crucially didn't allow the developer the option of including this or not in the definition of the picklist).

This weakened the quality of picklists for some use cases and strengthened it for others.

Just really illustrates that with most pieces of functionality few things are as clear cut as people imagine and we all have a tendency to think that our use case is the most special that absolutely most be supported and be a top priority for IBM (DBRH anyone???)

[ultrakocka]

Thanks all for valueable ideas!

- ItemReject indeed solves (somehow unexpectedly) the problem with custom messages which I had.

- Using "front-end" selection/picklist cube for passing parameters to process is .. well, why not, I can definitely imagine using it, but quite different way of doing things than I'm used to. We'll give it a try as it seems quite reasonable solution, just hope the overhead won't prove too large.

Best regards,
Simon

[luizg2019]

Hi,
Good shout on the error message luzig2019, hadn't picked this up.

It seems like barely anyone noticed this new feature
Should I make a topic about this on the "Useful code, tips and tricks" section?

[Wim Gielis]

Should I make a topic about this on the "Useful code, tips and tricks" section?

Sure, please do so

[Alan Kirk]

Hi,
Good shout on the error message luzig2019, hadn't picked this up.

It seems like barely anyone noticed this new feature
Should I make a topic about this on the "Useful code, tips and tricks" section?

I don't think you can; IIRC posting rights in that forum require at least MVP rank. But if you create one in the main forum one of the admins can move it for you.

[Emixam]

I tried luizg2019 solution in PAW but apparently only users with ADMIN access can see the custom error message. Normal users (non-admin) only see "Process completed with messages". Am I missing something here ?

[Mark RMBC]

Hi Emixam,

this is a bug that appeared around version 62 or so. Well the first bug was that it didn't return any message and then by fixing that they introduced the bug that only admin users see it.

I raised this on the IBM community forum and surprisingly got no response, see here:

https://community.ibm.com/community/use ... gestviewer

I also raised it with IBM support and got the following response:

"The issue has been reproduced in PAW 63 and I will log a defect investigation for it. I will keep you updated."

regards,

Mark

[gtonkin]

I have worked around this for now by having a check in Epilog to check if any errors were encountered in previous tabs e.g.
nErrorCount=PrologMinorErrorCount+MetadataMinorErrorCount+DataMinorErrorCount;

If we have errors, I then fire off an email process with the relevant log attached. Chances are that this will remain in place.

[Mark RMBC]

Hi George,

I really hope it doesn't stay in place. If a user process errors I want an instant message on screen for the user to view, it is far from ideal if they have to open an email which has been sent to them!

My reading of IBM's response is that they regard it as a defect, and so do I!

regards, Mark

[gtonkin]

Hi George,

I really hope it doesn't stay in place. If a user process errors I want an instant message on screen for the user to view, it is far from ideal if they have to open an email which has been sent to them!

My reading of IBM's response is that they regard it as a defect, and so do I!

regards, Mark

100% Mark - needs to be immediately visible to the users with all the expected detail. I will keep the email as an additional item and possibly only notify the custodian in the future as there may be something they need to address. Also quicker than trying to locate the log in PAW later.

I have also had mixed results in errors message handling when processes are run from the database pane vs action buttons. Will be checking those results again too to ensure behaviour is the same.

[Emixam]

My reading of IBM's response is that they regard it as a defect, and so do I!

Thanks for your answer Mark. So you already raised a ticket to IBM ?

[Mark RMBC]

So you already raised a ticket to IBM ?

Yes!