LabVIEW Idea Exchange

It would be useful if LabVIEW offered the option to use an Error Collector Node inside structures.

The Error Collector Node would collect or capture any/all unhandled errors that occurred inside the structure that the node is part of. If one or more unhandled errors occurred, the Error Collector Node would output the first error that occurred (in chronological error). If no errors occurred, the node would output "No Error".

The following annotated screenshot shows how the Error Collector Node would help reduce the number of error wire segments and of Merge Errors nodes.

The following screenshot shows a second example. Notice that using the Error Collector Node would result in a significant reduction of block diagram items: 10 fewer wire segments and 1 fewer Merge Error Node. The reduction in block diagram items could be even larger if we consider that the other cases of the case structure (potentially dozens of cases) also benefit from the same Error Collector Node being placed on the border of the case structure.

The two screenshots above are examples created specifically for the purpose of posting this idea. The following screenshot is a real-world example (taken from production code) of a VI that could benefit from the Error Collector Node, which would remove the need for numerous error wire segments and Merge Error nodes.

Notes

• The Error Collector Node was prototyped as a hexagonal output tunnel in the screenshots above. I would, of course, be happy if a different glyph or shape is chosen to represent this new type of tunnel.
• Error Collector Node would essentially behave similarly to a localised Automatic Error Handling functionality. It would collect or capture any unhandled errors inside its area of responsibility (the structure), and would convert those errors back into "manually-handled" errors - i.e. errors than are passed downstream via an error wire.
• The Error Collector Node would be useful especially in situations where many code branches execute in parallel as it eliminates the need for lots of Merge Error nodes.
• The first two screenshots mention the words "approximately equivalent to". "Approximately" because, if multiple errors occur inside the structure, the Error Collector Node does not guarantee which of those errors are output in the same way that the Merge Errors node does. For example, in the first screenshot, if all three VIs (VI A, VI B, and VI C) experience an error, there is no guarantee as to which of those errors the Error Collector Node would output. The node would output the first error that occurred (in chronological order), so it would depend on which VI finished execution first. This could change from execution to execution, and from machine to machine. Whereas the right-hand-side version, which uses Merge Errors, would always output the error generated by VI A.
  
  • This would usually not be an issue in practice. If more determinism is needed, the programmer could, of course, fall back on manually wiring the errors to define an exact error behaviour.
• It should be possible to add a maximum of one Error Collector Node to each structure: Flat Sequence Structures, Case Structures, For Loops, While Loops, etc.
• It would be useful if the Error Collector Node could be used outside of structures (directly on the outermost region of a block diagram). Again, enforcing a maximum of one Error Collector Node per outer block diagram would make sense. The Error Collector Node would execute after all other block diagram nodes and structures (would be the last thing to execute). The output of the Error Collector node could then be fed directly to an "Error Out" block diagram terminal. This would remove the need to wire most error wires inside the VI, while ensuring that no error goes uncaptured.
• If the Error Collector Node exist only as a structure output tunnel, and not as a stand-alone node outside of structures, then a better name for it might be the Error Collector Output Tunnel.
• The behaviour of the Error Collector Node would be unaffected by Automatic Error Handling being enabled or disabled in that VI.
• Using Error Collector Nodes would benefit programmers in the following ways:
  • Would reduce the amount of "click-work" that programmers currently need to do (the number of wire segments and Merge Error nodes that need to be created), while ensuring that all unhandled errors are captured.
  • Would reduce the amount of block diagram "clutter". This "clutter" is apparent in the third screenshot, which shows many criss-crossing error and DAQmx wires.
  • Would decrease the size on disk of VIs thanks to fewer block diagram items needing to be represented in the VI file. This would help towards making git repositories a little bit smaller, and loading VIs into memory a little bit quicker.
• Informally, using Error Collector Nodes would sit in-between the strictness of manually wiring all error outputs, and the looseness of relying solely on Automatic Error Handling. The error handling gold-standard would remain manually wiring all error outputs, but using Error Collector Nodes might be "good enough" in many situations, if used judiciously.

Problem: Currently wiring an error wire to a structure input tunnel that does not continue as a wire clears the error that exists on the wire.

Happy case: When running the VI shown below, Automatic Error Handling correctly detects that the error out terminal of Error Cluster From Error Code.vi is unwired, and handles the error (displays the error as a dialogue window).

Unhappy case: Wiring the error wire from the error out terminal of Error Cluster From Error Code.vi to a structure input tunnel clears the error. Automatic Error Handling does not detect or handle the error.

In my opinion this was simply an unfortunate design decision (can happen to all of us) back when it was made, decades ago. IMO there is no logical argument to support this behaviour. The fact that the error wire is wired to an input tunnel does not mean that the error was handled. At best, when a programmer intentionally used this technique, it represents a non-self-documenting coding practice (why not use the self-documenting Clear Errors.vi?). At worst, it means clearing errors simply because the programmer forgot to wire the wire through the structure. It means clearing errors when the programmer did not explicitly ask for this. It means "sweeping errors under the carpet", and can result in overly "optimistic" applications (apps that seemingly execute without error when in fact unhandled errors are being generated).

Please note that even though the screenshot above shows a Flat Sequence Structure input tunnel, the behaviour applies to every structure (case structure, for loop, while loop, etc).

To summarise, the problem is that the screenshot above is functionally equivalent to explicitly using the Clear Errors.vi, as seen below.

Clear Errors.vi is of course the self-documenting, recommend method of clearing errors. It should also be the only method of clearing errors.

Ironically, Clear Errors.vi itself uses the "clear error by wiring it to input tunnel" technique inside its "0" case, as seen below.

To its credit, Clear Errors.vi uses a correct technique for clearing errors inside its other, default case.

Another example found "in the wild" of a VI using the "clear error by wiring it to input tunnel" technique. This VI ships with LabVIEW and is found at <LabVIEW installation folder>\vi.lib\Utility\EditLVProj\Identify VIs Among Project Items.vi.

Solution: Disable the "clear error by wiring it to input tunnel" behaviour. This would fix what IMO is an incorrect design decision. Unfortunately, fixing this decision now would result in VIs that use the "clear error by wiring it to input tunnel" technique to start throwing unhandled errors if AEH is enabled. This is not ideal, but it might be worth accepting this short-term drawback for long-term gain.

Moreover, it may be useful to introduce a Clear Errors node (primitive function). The Clear Errors.vi could then make use of the Clear Errors node inside both of its cases. Alternatively, the Clear Errors node could simply replace and supersede the Clear Errors.vi.

Automatic Error Handling (AEH) is a useful feature. It captures errors that were otherwise left unwired by the programmer (intentionally or accidentally). It represents a "safety net" that can make the programmer aware of errors that they may otherwise remain unaware of.

Problem: Currently AEH functionality is only available in the development environment. It is not available in built executables. Even when all VIs in a project have AEH enabled, once built into an EXE, all VIs behave as if AEH was disabled.

Solution: It should be possible to honour each VI's AEH setting in built EXEs too, not just in DevEnv. The EXE build specification could contain a setting named "Honour each VI's Automatic Error Handling setting in EXE". When ticked (enabled), any VIs for which AEH was enabled in the development environment will continue to benefit from AEH behaviour in the EXE. Any VIs for which AEH was disabled will continue to have it disabled in the EXE. This means that, from an error handling/error manipulation point of view, the application would behave identically when being run as an EXE as when being run in Development Environment. This is more consistent, and can be helpful.

The current behaviour (forcibly removing AEH in EXE) means that EXEs are prone to having errors that were not discovered during DevEnv testing being "swept under the carpet". In other words, currently EXEs are overly "optimistic" - they can make the programmer believe that everything is ok when in fact one or multiple unhandled errors are occurring, errors that would have been visible in DevEnv. This is particularly relevant to apps that run for long periods of time (e.g. life cycle testers) that may encounter errors that were simply unforeseen or untested in DevEnv (e.g. error after one month of continuous running due to running out of disk space when saving measurements log file to disk).

The screenshot below shows how the new setting could look like in the Advanced page of the EXE build spec.

Notes

• To be absolutely clear: I am not asking for AEH to be enabled by default in all EXEs. I am also not asking for AEH to become enabled in all VIs when the new build spec setting is ticked.  This would override the AEH setting of all VIs - I am not asking for this.
• The default value of the new setting should be False (unticked). When False, the built EXE would behave exactly as it does now - AEH would be disabled in all VIs. This would maintain the current behaviour as default.
• The new setting would give the programmer more control - it would allow the programmer to decide whether they want AEH or not in their EXE. Currently AEH is taken away in EXE, even when we (professional LabVIEW programmers) might want it enabled.
• I would be happy if the new setting was available only for desktop, non-real-time applications, and not on Real-Time targets.

I'm proposing a new certification, one certification to rule them all, the Certified LabVIEW Gangster Group Developer.

Or simply a fully functioning CLA that requires 3 people to take the test.

Architect

Developer

Embedded Developer

It should function using an FPGA, or a simulated interface.