LabVIEW

cancel
Showing results for 
Search instead for 
Did you mean: 

Darren's Occasional Nugget 09/26/2022


@Ettepet wrote:

There are still flaws to this way of reading an Excel file.

 

Besides getting the internal data in USA-format there are also rounding errors to deal with.  Exact dotted numbers in a number-field are retreived in an imprecise way:  "10.7" becomes "10.700000000000001" if you define the number field with "Decimal places: 1".

 

Reading this through the regular Labview-mechanism gives "10.7", or in my local format: "10,7". (like it shows in my local format Excel file)


Note that 10.7 is not exactly representable in binary floating point.

"If you weren't supposed to push it, it wouldn't be a button."
0 Kudos
Message 21 of 53
(1,838 Views)

@Ettepet wrote:

There are still flaws to this way of reading an Excel file.

 

Besides getting the internal data in USA-format there are also rounding errors to deal with.  Exact dotted numbers in a number-field are retreived in an imprecise way:  "10.7" becomes "10.700000000000001" if you define the number field with "Decimal places: 1".

 

Reading this through the regular Labview-mechanism gives "10.7", or in my local format: "10,7". (like it shows in my local format Excel file)


Is there something in the .xlsx file we can use to determine whether a value is a numeric or a string?

"If you weren't supposed to push it, it wouldn't be a button."
0 Kudos
Message 22 of 53
(1,820 Views)

Hi..

can someone convert to lv17

0 Kudos
Message 23 of 53
(1,793 Views)

@Ettepet wrote:

There are still flaws to this way of reading an Excel file.

 

Besides getting the internal data in USA-format there are also rounding errors to deal with.  Exact dotted numbers in a number-field are retreived in an imprecise way:  "10.7" becomes "10.700000000000001" if you define the number field with "Decimal places: 1".

 

Reading this through the regular Labview-mechanism gives "10.7", or in my local format: "10,7". (like it shows in my local format Excel file)


I did some checking in the unpacked xml-files and discovered that the imprecision was introduced by a colleague, not by some inherent Excel-problem. I have asked him to import only rounded values in future.

 

NB: I would have edited this to my original post but editing older posts seems to be restricted.

0 Kudos
Message 24 of 53
(1,764 Views)

What do you mean "rounded values"?  Rounded values will produce the same issues as long as you are representing them as floating point numbers.

Bill
CLD
(Mid-Level minion.)
My support system ensures that I don't look totally incompetent.
Proud to say that I've progressed beyond knowing just enough to be dangerous. I now know enough to know that I have no clue about anything at all.
Humble author of the CLAD Nugget.
0 Kudos
Message 25 of 53
(1,714 Views)

@gkrn wrote:

Hi..

can someone convert to lv17


Sure, here you go. A couple of notes:

1) I only tested this with one simple .xlsx file. It worked, but I didn't do the extensive testing on this one that I did on the original. That being said, I can't think of a reason why this 2017 version wouldn't work just like the original does.

2) LabVIEW 2017 doesn't have NXG Style Controls, which I used for the front panel of the original. So the controls look "weird" in 2017.

0 Kudos
Message 26 of 53
(1,716 Views)

@ooth wrote:

@Frozen

My  test Excel file didn't have any empty cells before the data, like the one you provided does. So I made a fix for that a little different than how Darren checked. See if this works for you. Might be faster than Darren's.


There were several problems/bugs in this code (empty-field detection criterium bug, escape sequences not implemented, decimal point/komma, etc.).

 

Here a fully working version in Labview 2018.  It is in my own compact format, so a little harder to read.

(wondering if it remains as fast as yours)

 

 

0 Kudos
Message 27 of 53
(1,632 Views)

Hi Ettepet,

I'm sure it depends on the Excel file that you use. I tried your code and it took 21.1sec and mine took 0.3sec. But I'm sure my code probably does have the problems you mentioned (empty-field detection criterium bug, escape sequences not implemented, decimal point/komma, etc.). I really just took Darren's original and found where it was taking longer to process.

Message 28 of 53
(1,611 Views)

Here is a much faster version.  I just inlined some code.  Thanks for the feedback!

Message 29 of 53
(1,593 Views)

That code is a winner, Ettepet! It took 0.3sec to open my Excel file.

Thanks.

Message 30 of 53
(1,588 Views)