Script task

Really, only that the user can't directly interact with it.  Nothing with a saveInto will work.  Nothing only configured to work from within a saveInto will work.  

Am not getting you

Objects with a saveInto parameter won't work, for instance a!textField, or any form of link.  a!startProcess() (if I remember how to spell it) and a few other functions can only work from within a component's saveInto parameter.  Those will not work from a script task.

Anything that does not require user input is theoretically possible with a script task.

 thank you!

Something that isn't impossible but is pretty frustrating when you run into it -- the fact that you can't take one complex expression and save various elements of it each into their own PV.  This often means that you're either adding multiple outputs to the same script task (regardless of how redundant they are in functionality) to populate each PV, and/or chaining together mutliple script tasks to do consecutive-style processing.

Note that I don't exactly have a proposed solution in mind to this, but the fact that we still need to use such silly workarounds sometimes just underscores how dated and inflexible the functionality is.  That goes for the "output variable" system in general across most process model nodes, not just script tasks of course.

Something that isn't impossible but is pretty frustrating when you run into it -- the fact that you can't take one complex expression and save various elements of it each into their own PV.  This often means that you're either adding multiple outputs to the same script task (regardless of how redundant they are in functionality) to populate each PV, and/or chaining together mutliple script tasks to do consecutive-style processing.

Note that I don't exactly have a proposed solution in mind to this, but the fact that we still need to use such silly workarounds sometimes just underscores how dated and inflexible the functionality is.  That goes for the "output variable" system in general across most process model nodes, not just script tasks of course.

Mike Schmitt I assume you're aware that you can now use the Map type both as an output from a complex expression AND as a type natively within a Process. This means your Expression can be called once, return what is in effect an array of strongly typed elements, and get access to those elements directly in the process model. Or have I missed your point somehow?

Stewart Burchell - yes, in fact i mentioned this during an external conversation I had a bit later yesterday, referring to this thread (though it sorta slipped my mind when writing my above comment).  My mind is, however, still in the mode where I'd sometimes like to evaluate one expression and save different parts of the output into different PVs, without having to evaluate that expression multiple times or take multiple steps.  In such cases (admittedly pretty rare, when dealing with more complex process functionality) using a!map would work but would still be a workaround.

I should also mention that using a CDT would fulfil the same concept. If you can reference a pv! then it's just as easy to reference an element in a CDT that's in a pv!. The Map was supposed to get around having to define CDTs to make this pattern work as you can generate them at run-time. But you also need to be pretty darn sure you know what your Map contains so that you address its contents reliably.

Hi Mike,

If I understood you correctly, on saving different parts of the output to different PVs.

We can query the database in the input tab of Script task, and use that activity class param in the output tab.

So the query is executed only once and extract total count, data, and other fields to the different output PVs.