Custom Structures

[InfiniteNothing]

To further expand generic programming, we should be able to specify custom patterns as a custom structure. For example, if you very often write "retry once" code you could write retry2x.

I think if you expand this concept you can accomplish other requests on this board like case structures that respond to error wires with warnings on them. If you expand it farther you could have "express" structures that have a configuration interface.

[tst]

I don't see how this code is readable and debugable, which is a problem.

LabVIEW already has the option to place a piece of code in the palettes and then drop that actual piece of code (see here, for instance). Does that provide what you want, at least in the sense of not requiring you to rewrite the same piece of code?

[InfiniteNothing]

Ooops, I tried to edit the last post but it created a new one. See comment below

[InfiniteNothing]

I think it's as debugable as a subVI. You'd double click on the structure and it would take you to the VI in the lower half of the image. I'm not sure I made that clear. There are two advantages of this idea over the palette drop.

• The top level VI (the top half of my image) would have less clutter because more of the code would be in the subVI (below)
• If you later decided that you want all your "retry" loops to have a 250 ms period between retries or if you decide that it should only wait on the second iteration, you wouldn't have to go to every loop and change the code, you'd just change the subVI and the change would propagate everywhere. This lowers code maintenance and aids abstraction.

[smercurio_fc]

Sorry, but I still don't understand what you're asking for.

[InfiniteNothing]

Essentially, case structures with built in logic defined by a subVI. In my example, Retry2x.vi is the subVI. The code at the top half of  the image is calling the structure with built in logic by wrapping the subvi structure around the file open code.

Maybe subVI is the wrong term to use here. It might need its own term like substructure. 

[GregR]

One issue is that there is nothing in the subVI/subdiagram/substructure to represent the enclosed code. If the diagram must contain one and only one structure, then you can define that the referencer's code is inserted into that structure. I think you will very quickly come up with cases where you need multiple structures (either nested or not) and then you need a way to say where that code is inserted. There is also the issue of specifying tunnels. In your example you have one normal tunnel and one that looks like a sequence local. The custom structure would need to be able to define arbitrary visuals for an arbitrary set of terminals. This quickly becomes a very complicated set of configuration and would make this quite complex to create.

For the specific case of "retry" it seems like there are probably not that many operations that you commonly want to retry so making a subVI for each operation would probably be reasonable. I have a hard time seeing how we would generalize this to a wide set of use cases without making it very complicated.

[MichaelAivaliotis]

Why not encapsulate the whole retry action inside a SubVI?

[InfiniteNothing]

Because you might want to retry all sorts of things. TCP connections, database queries, etc. It's a very common way to handle errors. Currently, you'd have to create a subVI for each. If you change your mind on the implementation, maybe you want to enqueue a flag so you know a retry was needed, you have to apply that change throughout. 

[Darren]

Any idea that has not received any kudos within a year after posting will be automatically declined.