Arc-created TI process - Websheet Action Button - "Process Failed"

[AlanKirk]

I don’t think that this is a bug in Arc, it’s more likely to be to do with the API.

I created a process in Arc using a cube view as the source. I tested it out, running it from Arc. It successfully completed and did everything that it was supposed to do.

I created a websheet and put an action button on it, then selected all of the parameters using what IBM passes off as a range picker, but let’s not get started on that.

I click the button and… “Process Failed”.

OK, whhhyyyy, exactly?

I simulate running the process by copying all of the parameters, pasting them into Arc, and… it ran fine.

Went back to the Action Button… “Process Failed”.

Long story short, I opened the process in Perspectives and looked at the parameters, then closed it down. At that point Perspectives asked if I wanted to save changes, except… WHAT changes? I didn’t touch the thing. Didn’t visit the Variables tab, didn’t change a line of code. And I had made sure that I had saved the code (and that the save succeeded) in Arc before I opened Perspectives.

Nonetheless I said yes, went back to the websheet and THIS time the action button ran as expected.

So presumably Perspectives is still making some change that Arc / the API isn’t (in some circumstances at least), but having burnt way too much time on this already I don’t currently have the time to chase down the root cause of this.

In case anyone else encounters this, though, at least there is a workaround.

[tryan]

Hi @AlanKirk,

What version of Arc are you using?

[AlanKirk]

1.3.1.

Now I can guess what you’re going to say; you’re going to say that the current version is 1.4.1 and I should upgrade. And I shall, when I get a moment. That moment is not now since I have worked 272.57 hrs in the last 4 weeks to Friday afternoon, not counting this morning, and the time for upgrading is zero.

But I’ve been through your Milestones list for the 1.4 series and I don’t see anything remotely similar to this.

Also, and I suspect that this is related… this morning I created a new process, completely in Arc. (When I say “completely” it was created by doing a Save As from the earlier process, which it is closely related to.)

I ran the second process from Arc when I was done and got some odd results… as if the data source (a cube view which is created in the Prolog) had not changed.

Which indeed it hadn’t. I threw a Process Quit into the mix right after the creation of the data source, and went to take a look in the cube.

None of the custom subsets had been created, despite the Debug log showing no issues. The custom view had not been created, despite the Debug log showing no issues.

So I threw a copy of the database into another development instance. I cracked open the process in Perspectives TI and ran it from there. This time it worked.

I therefore loaded to two process files up in WinMerge to see whether there were any differences. There was one notable one.

The “Purely Arc” process shows 7 variables. The one that I recompiled in Perspectives TI showed 10; the 7 user defined ones and the three implicit ones which relate to a cube view. That is, Value_Is_String, NValue and SValue.

If I had to bet, I’d say that there’s an issue with TM1’s API functions not creating the data source properly when the data source is a cube view, even when the TI code itself explicitly sets its data source to a view.

The code that powers the “classic” editor is probably doing something upon process creation that the RESTful API isn’t.

I can’t prove it but that’s a heck of a coincidence if that’s not it.

[tryan]

Hi @AlanKirk,

Thanks for the update, can you please create a ticket at Issues · Cubewise CODE / Arc and Slice · GitLab so we can look into it further.