Request: round tag

[Octal450]

I'm submitting a ...

• feature request

Describe the request

Inside <function>, currently I am using the integer tag with a summation of 0.5 to round values. It would be great to have a tag that works like this:

<round>
    <property>myprop</property>
    <unit>5</unit>
</round>

In this case, rounding myprop to 5.

Which support rounding to any arbitrary number (10, 5, 12.5, etc) including decimal like 0.1, 0.2, etc.

What is the motivation / use case for changing the behavior?
Not needing to do this all over my code :D

Thanks,
Josh

[bcoconni]

You could define a function template (note the hashtag # that will be replaced by the actual property when the template function will be applied later):

<function name="my-correct-round" type="template">
  <quotient>
    <integer>
      <sum>
        <product>
          <property> # </property>
          <value> 10.0 </value>
        </product>
        <value> 0.5 </value>
      </sum>
    </integer>
    <value> 10.0 </value>
  </quotient>
</function>

then apply it to the property hydraulics/stabilizer/final-deg using the apply="" attribute:

<sum>
  <property apply="my-correct-round"> hydraulics/stabilizer/final-deg </property>
  <property> some/other/property </property>
</sum>

[seanmcleod]

Ah, I had forgotten about the support for template functions, even though I added a bunch a number of years ago 😉

Add a set of template functions for unit conversions to SI units - cfd181b

[seanmcleod]

Although I'm sure the next request will be to allow parameters to be passed to the template functions 😉

[Octal450]

I didn't realize you could do that. But I would have to copy it thru all the planes. Hence the request. :)

-J

[seanmcleod]

Ideally you'd stick this in a single rounding.xml file and then include/reference it in each aircraft's FDM that requires it.

jsbsim/scripts/737_cruise.xml

Lines 78 to 83 in 57361d0

	<output name="737_cruise_output.csv" type="CSV" rate="10" file="unitconversions.xml">
	<property apply="convert-lbs-To-N" caption="propulsion/engine[0]/thrust-N">propulsion/engine[0]/thrust-lbs</property>
	<property apply="convert-lbs-To-N" caption="propulsion/engine[1]/thrust-N">propulsion/engine[1]/thrust-lbs</property>
	<property>attitude/theta-rad</property>
	<property apply="convert-rad-To-deg" caption="attitude/theta-deg">attitude/theta-rad</property>
	</output>

[seanmcleod]

So I tried a quick test with the aerodynamics section.

<aerodynamics file="unitconversions">
......
......
        <axis name="DRAG">

            <function name="aero/coefficient/CD0">
                <description>Drag_at_zero_lift</description>
                <product>
                    <property apply="convert-R-to-K">aero/qbar-psf</property>
                    <property>metrics/Sw-sqft</property>
                      <table>
                          <independentVar>aero/alpha-rad</independentVar>
                          <tableData>
                             -1.57	1.5000
                             -0.26	0.0420
                              0.00	0.0210
                              0.26	0.0420
                              1.57	1.5000
                          </tableData>
                      </table>
                </product>
            </function>

However I got the following error. Basically none of the functions defined within the aerodynamics section but outside of a specific axis were found.

In file aircraft/737/737.xml: line 585
Property aero/function/kCDge does not exist
FATAL ERROR: JSBSim terminated with an exception.
The message was: FGPropertyValue::GetValue() The property aero/function/kCDge does not exist.

jsbsim/aircraft/737/737.xml

Lines 451 to 516 in 57361d0

	<aerodynamics>
	<function name="aero/function/kCDge">
	<description>Change_in_drag_due_to_ground_effect</description>
	<table>
	<independentVar>aero/h_b-mac-ft</independentVar>
	<tableData>
	0.0000 0.0480
	0.1000 0.5150
	0.1500 0.6290
	0.2000 0.7090
	0.3000 0.8150
	0.4000 0.8820
	0.5000 0.9280
	0.6000 0.9620
	0.7000 0.9880
	0.8000 1.0000
	</tableData>
	</table>
	</function>
	<function name="aero/function/kCLge">
	<description>Change_in_lift_due_to_ground_effect</description>
	<table>
	<independentVar>aero/h_b-mac-ft</independentVar>
	<tableData>
	0.0000 1.2030
	0.1000 1.1270
	0.1500 1.0900
	0.2000 1.0730
	0.3000 1.0460
	0.4000 1.0280
	0.5000 1.0190
	0.6000 1.0130
	0.7000 1.0080
	0.8000 1.0060
	0.9000 1.0030
	1.0000 1.0020
	1.1000 1.0000
	</tableData>
	</table>
	</function>
	<function name="aero/function/kCLsb">
	<description>Change_in_lift_due_to_speed_brake</description>
	<table>
	<independentVar>fcs/speedbrake-pos-norm</independentVar>
	<tableData>
	0.0000 1.0
	0.1000 0.85
	</tableData>
	</table>
	</function>
	<function name="aero/function/kCLsp">
	<description>Change_in_lift_due_to_spoilers</description>
	<table>
	<independentVar>fcs/spoiler-pos-norm</independentVar>
	<tableData>
	0.0000 1.0
	0.1000 0.6
	</tableData>
	</table>
	</function>
	<axis name="DRAG">

I then commented out any references to these functions and then the script succeeded and I noticed a difference in thrust required etc., i.e. my template function was being applied. But it appears that by 'including' my template functions they effectively replace any other functions defined in the aerodynamics section outside a specific axis as opposed to ending up with a union of the functions.

I haven't looked at the source code yet. Not sure if this is intentional.

[bcoconni]

But I would have to copy it thru all the planes. Hence the request. :)

Similar requests (refer to issues #94, #126) have been noted, and JSBSim cannot accommodate every conceivable mathematical function in C++. Doing so would increase the maintenance burden for specialized use cases and contradict the purpose of employing XML to circumvent hardcoded C++ functionalities. The introduction of template functions was intended to address this very issue.

[bcoconni]

However I got the following error. Basically none of the functions defined within the aerodynamics section but outside of a specific axis were found.

This is most likely due to this code since an <output> section is being loaded in an <aerodynamics> section.

jsbsim/src/input_output/FGModelLoader.cpp

Lines 78 to 81 in 57361d0

	if (document->GetName() != el->GetName()) {
	document->SetParent(el);
	el->AddChildElement(document);
	}

I would have expected the code from <output> to be ignored by <aerodynamics> but that is obviously not the case. This looks very much like a bug…

[seanmcleod]

Ah, sorry, I forgot to mention I had actually edited my local copy of unitconversions.xml, to:

<?xml version="1.0"?>
<aerodynamics>
  
  <!-- 
  JSBSim exported non-SI units:
  
  - Angles
  rad

Before I made the change I got this error:

In file aircraft/737/737.xml: line 494
 Read model 'output' while expecting model 'aerodynamics'

In file aircraft/737/737.xml: line 494

Aircraft aerodynamics element has problems in file Path "aircraft/737/737.xml"
Script file Path "./scripts/737_cruise.xml" was not successfully loaded
(base) C:\source\jsbsim>

[Octal450]

@bcoconni
I completely understand that, and it make sense for what was mention in that issue.

However, I feel that rounding is one of the "basic" operations that should be implemented, similar to how ceiling, floor was added.

With rounding, it is not a specific use case, but a very common need in all kinds of things. Hence I find that it should be included.

But I completely understand why you had said about template for the others, which are more complex and specific operations.

Kind Regards,
Josh

[bcoconni]

With rounding, it is not a specific use case, but a very common need in all kinds of things. Hence I find that it should be included.

JSBSim already includes the standard rounding functions floor, ceil, and integer, which suffice for the great majority of use cases. Your particular requirement (to multiply by 10, add 0.5, round to an integer, and then divide by 10) is indeed specific and does not fall into the 'common need' category.

jsbsim/src/math/FGFunction.cpp

Lines 520 to 523 in 57361d0

	} else if (operation == "floor") {
	Parameters.push_back(make_MathFn(floor, fdmex, element, Prefix, var));
	} else if (operation == "ceil") {
	Parameters.push_back(make_MathFn(ceil, fdmex, element, Prefix, var));

jsbsim/src/math/FGFunction.cpp

Lines 546 to 552 in 57361d0

	} else if (operation == "integer") {
	auto f = [](const decltype(Parameters)& p)->double {
	double result;
	modf(p[0]->GetValue(), &result);
	return result;
	};
	Parameters.push_back(new aFunc<decltype(f), 1>(f, fdmex, element, Prefix, var));

[Octal450]

Hi @bcoconni
I don't understand that. floor/ceil and integer do not round the number. Hence why I am doing that thing with 10 and 0.5

Actually, you don't need the * 10 thing. that's just because I'm rounding to the 0.1s place not the 1s place. It not a specific.

It's just rounding to a place.

By rounding I mean
0.5 -> 1
0.4999 -> 0
0.8 -> 1
0.2 -> 0

Behavior with floor:
All 0.x goes to 0

Behavior with ceiling:
All 0.x goes to 1

Integer = Floor. Actually I don't understand why Floor was added, is the same as integer, no?

Kind Regards,
Josh

[Octal450]

My example in the first post, rounds the stabilizer to the 0.1's place.

In the above post, I gave examples for the 1's place.

The equivalent current code would be:

<integer> <!-- or floor -->
    <sum>
        <property>hydraulics/stabilizer/final-deg</property>
        <value>0.5</value>
    </sum>
</integer>

Bests,
Josh

[seanmcleod]

Integer = Floor. Actually I don't understand why Floor was added, is the same as integer, no?

floor() rounds down. int() truncates. The difference is clear when you use negative numbers:

>>> import math
>>> math.floor(-3.5)
-4
>>> int(-3.5)
-3

<integer> <!-- or floor --> - No, you need floor here.

[Octal450]

I see. I'll have to go fix that.

Thanks.

I maintain my request though

Regards,
Josh