AnswerHub Upload Issue
Our AnswerHub community currently has a bug that prevents you from attaching your model or other files to your questions. AnswerHub is aware of this issue and they are working on a fix. Until we receive their patch, you can upload your models and other files to FlexSim's file share site. Include the file name in your question so that we know which upload to reference.

question

Daniel Pavan avatar image
1 Like"
Daniel Pavan asked Ralf Gruber commented

Incorrect calculation values

Please run the following code in a script window with the output window open:

int test1 = 138;

int test2 = 50;

double test3 = test1/test2;

pr();pt(\"dp number testx\");pt(numtostring(test3, 54, 50));

return test3;

The answer provided in the script window appears to be correct (2.7600) but the answer shown in the output windows shows the calulated value(2.75999999999999980). Why is the actual calculated value slightly wrong?

FlexSim (other - please specify)
FlexSim (other)
calculationdecimal precisionfloating pointcomparisonaccuracy
5 |100000

Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.

1 Answer

·
Phil BoBo avatar image
2 Likes"
Phil BoBo answered Ralf Gruber commented

Danny,

That is the nature of how computers handle decimal precision values, particularly how C++ handles a division operation. There are lots of complicated explanations and papers about why it is done this way.

A relatively short explanation can be found on wikipedia: http://en.wikipedia.org/wiki/Floating_point

There is a section entitled "Accuracy problems" near the bottom of that article that explains it a bit.

Usually it doesn't matter that the precision is slightly off, but sometimes it messes with your logic when doing comparison operations as shown below.

The main issue is when you think it is going to get into one of the simple cases (==, <=, >=, <, >, or !) but it doesn't. There are a few things you can do to see if it is "close enough" though to avoid such precision comparison issues. I've included some of those in the following code as well. It happens that your example actually does work for the == test although ofttimes such division calculations wouldn't. You have to be careful in C++ with == when you are comparing to a double value that has been assigned based on a division or multiplication operation.

int test1 = 138.0;
int test2 = 50.0;
double test3 = test1/test2;

pt("dp number testx");pt(numtostring(test3, 54, 50));
pr();
if(test3 == 2.76)
{
 pt("exactly 2.76");
 pr();
}
if(fabs(test3 - 2.76) < 0.0001)
{
 pt("close enough to equal 2.76");
 pr();
}
if(test3 <= 2.76)
{
 pt("<= 2.76");
 pr();
}
if(test3-0.0001 <= 2.76)
{
 pt("close enough to <= 2.76");
 pr();
}
if(test3 >= 2.76)
{
 pt(">= 2.76");
 pr();
}
if(test3+0.0001 >= 2.76)
{
 pt("close enough to >= 2.76");
 pr();
}
return test3;

I hope this helps. Good luck.

· 14
5 |100000

Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.

fabs(test3)-2.76 

Did you mean:

fabs(test3-2.76)

?

0 Likes 0 ·

Yes, thanks. I've updated it.

0 Likes 0 ·

@phil.bobo

Need a solution for my problem with pulling values from internal list near 0.00001.

Before I pull a request number from a internal list I would like to ask for the labelvalue on the list.

For Gobal List I do this with a Tablequery for example:

Table T_result = Table.query("Select * From internalList WHERE Material == 'Tape'");

double value = T_result[1][x];

double request;

if ( request < 0.0000001)

{request = value;}

return request;

But how can I do this for internal lists because Tablequery is not possible?

Thanks for you help!

0 Likes 0 ·

Using Pull From List, pass 0 for the request number and require number if you want to query a list without removing any entries.

See Pull From List: Request Number and Require Number

"If the request number is zero, then the pull operation will become a mere querying operation. The pull will retrieve all objects on the list that meet the query filter, but it will not remove any values from the list."

"If the require number is zero then the activity will always finish immediately, returning the amount that it successfully pulled. In this case no back order will be created."

--

Also, the code you posted doesn't make sense. You aren't initializing the request variable so it will always be 0 and always be < 0.0000001. All that code snippet is doing is returning T_result[1][x].

0 Likes 0 ·

c-problem-0-example.fsm

@phil.bobo

sorry, my english is not so good, I thing I have not described my problem in a good way.

enclosed you will find a examplemodel, maybe my problem will be a little bit clearly. With the global List the model runs good. If you change the global List with the internel List the Model stoped at the last pallet!

0 Likes 0 ·

@phil.bobo

I do this in the second example (using Internal List without Tablequery). You will see that the token at the last run (number 5) stocks in the Pull from List activity and not in the End of Production activity. That´s my problem. I have a lot of internal list like this exampl in my real model (round about 350 pull activity in my model).

c-problem-0-example-intlist.fsm

0 Likes 0 ·

@jordan.johnson

I think what @christoph gruber is actually looking for has already been asked for in this thread:

https://answers.flexsim.com/questions/57299/sql-query-on-internal-list.html

Looking at the Release Notes, this has not yet been addressed, is that correct?

Thanks

Ralf

0 Likes 0 ·

That is correct, we have not added that feature yet.

0 Likes 0 ·
Show more comments

Write an Answer

Hint: Notify or tag a user in this post by typing @username.

Up to 12 attachments (including images) can be used with a maximum of 23.8 MiB each and 47.7 MiB total.