Error checking in Flexsim

Guys,

If you work for some time with a software product you get use to it and just work with it in your way. So it is always nice if you get new users which work the first time with the software. They try things which you never would try, because you already know some of the issues and you do not expect “special” functionality.
Anyway, I gave a training class some time ago and the main feedback was that Flexsim should have a better error checking. Not in the code, which is already quite good, but modeling or logical error checking.

- If Global Table not exists should come an error message.
- Empirical distributions should control to if they add up with 100%.
- Error checking should be much better (table names, label names treenode array, etc).
- If a parameter in a command is missing e.g. object pointer, Flexsim gives no error message.
- Would be nice to have a check if a pointer is wrong (no node). Okay, this can be done by using objectexists() but maybe a general switch would be nice to give a print out in the console or show a screen message.
- If in a task sequence an item pointer is wrong it would be nice if Flexsim gives an error message. There is an error message if an object is not connected to the network but not if the item does not exist. Usually Flexsim crashes in this case.

I am not sure if this will slow down Flexsim, but I am sure the experts will know.

In general it would be nice if Flexsim would give some more error messages. I know that there are more information in the system console but still worth to think about, right?

Take care
tom the (A)tom

I agree with Tom. It can take a lot time to find these mistakes in your model. Especially if you just start using Flexsim.
If the speed becomes an issue, then may be it is an idea to have some kind of debug-mode, where these checks will be done?

Guys,

You can use the Trace Debugger and the system console to get a lot of the error checking that you ar talking about.

I agree that it would be nice to have better error checking but you have to look at the magnitude of the task and the reality of the situation. The reason we simulate is the same reason that error checking is so hard; you just don't know what's going to happen until you run the simulation.

Take labels for example. If you are using more than one flowitem or creating your own and not using a Source, you don't know what flowitem is going to enter an object until it does. So, there is no way to check the labels until during the simulation. The system console will tell you that a label didn't exist at this point but the question is whether Flexsim should be more abrupt in telling you about the label. I say no because sometimes I have flowitems with the label and some without it. I do this because I don't want the extra overhead of another label duplicated thousands of times and I'm OK with the getlabelnum function returning a 0 when the label does not exist.

Another example is pointers. When exactly do we determine that you messed up? Many pointers are bad or NULL on creation and may never actually point to anything. Pointer can also be set to something that doesn't exist intentionally because we want to use the objectexists() command. So the question is, when should Flexsim check the pointer for good data and what should it do if it is bad? To do this check would require massive amounts of overhead and slow your model down so I would not want Flexsim to do that check.

The end result is that we can't just say that we want better error checking. We have to determine what error checks we want and when we want them and then ask for those specifically.

I am on your side here so please don't interpret this otherwise.

A useful tip that I've found the hard way: The development list is long and daunting and adding something new to it is like buying a lottery ticket. I've found that if I am willing to develop something to the 80% - 90% level that I have a much higher success rate of getting things put into the software.

Hi Tom,

I totally agree with you. And please see my other thread to address the similar issues:
http://www.flexsim.com/community/for...read.php?t=464

Alan

Hey Guys,
Yeah, actually, there is a lot of good error information that is already printed to the system console (and there's also a lot of bad information printed there as well), for example, bad label references and bad global table references are already printed to the system console in a pretty readable manner. I think we just need to clean up the system console prints so that there isn't so much junk printed there, and we need to have the system console open automatically when this error information is printed. We're going to work on that for this next release.

Is it good to have a separate console window to printout error informations for debugging purpose? I may imagine a lot of other information will still be print out to the system console as indicated by its name.

Honestly, right now the system console is mostly just junk (it's leftover engine debugging code). I don't see any reason why we shouldn't just clear out all the messages to the system console, and make it into a pure error console.

@Anthony
I like the idea to use the system console as an error console with useful information what is going on. I agree it should pop up if something is printed to it. If there would be a tick to control the pop up of this console I think we go in the right direction.

@Brandon
I did not use the Trace Debugger so far, so maybe that’s what I shall try to use. But so far I understood it is pretty slow and not really easy to handle.
Anyway, maybe this would be an idea for a session during the Advanced Training at the Flexsimposium. I would be happy to learn more about it, even if I check it out myself before.

@Alan
Yes, that issue you talk about in your thread is exactly what I am talking about. You could first check with objectexists() if the node exists, but than you need to do it manually. So if we would get a proper print out in the system console would be a great help.

Originally Posted by Tom David

@Brandon
I did not use the Trace Debugger so far, so maybe that’s what I shall try to use. But so far I understood it is pretty slow and not really easy to handle.
Anyway, maybe this would be an idea for a session during the Advanced Training at the Flexsimposium. I would be happy to learn more about it, even if I check it out myself before.

Is there any good documentation on how to use Trace Debugger? Or what is the right/supposed way to use it? Is there any issue need to know if using it? Frankly, I never used it and the description in the help is not quite useful.

Anthony it is a VERY good idea to make a true runtime error console. This will fit very nicely into the main objective of the October 4.5 release which is to improve the look, feel, and ease-of-use of Flexsim for new users. As part of cleaning up and improving the usefulness of the information displayed in the system console, I would like to suggest we find a better way of displaying the consoles in general. I think the consoles are hard to manage. They are always getting in the way, have to be cleared out, have to be resized to see everything, etc. I'm not sure what all would help this, but others might have some ideas. I think Phil's new tabbed windows he's developed for other things in 4.5 is one thing that would help. I personally would like to see consoles cleared out on model reset, but others may not, so it should be a setting.

Hi Cliff,

So I'm not the only one always fighting with the consoles.
I have a wide screen monitor so I always change the size from a lying rectangle to a standing one to the far right of the screen. And then at a certain moment the console needs to be gone so I close it and the next time I open it is again this lying rectangle that I have to move and resize again.

What I would like is that if you change the size and position, it opens again in that position. I think it is fine if that is only the case within 1 session, but it would make life a lot easier and I don't think it is hard to implement.

Steven