ToCompare vs ApproxCompare

[aligator]

Dear Support,

NT8 uses ApproxCompare in default Stochastics indicator to ensure a series is not divided by zero. However, the ApproxCompare dose not seem to be supported in NT8 since it will genrate an error if used for a custom script in a similar manner.

The closest supported feature I have found is "CompareTo." However it is not clear if it functions the same as ApproxCompare. So, I am trying to find the equivalent supported NT8 code for the following line from default Stochastics.

		protected override void OnBarUpdate()
		{
			double min0 = min[0];
			nom[0]		= Close[0] - min0;
			den[0]		= max[0] - min0;

[COLOR="Magenta"]			if (den[0].ApproxCompare(0) == 0)[/COLOR]
				fastK[0] = CurrentBar == 0 ? 50 : fastK[1];
			else
				fastK[0] = Math.Min(100, Math.Max(0, 100 * nom[0] / den[0]));

			// Slow %K == Fast %D
			K[0] = smaFastK[0];
			D[0] = smaK[0];
		}

All I really want is to make sure den[0] is a non-zero value.

The "if (den[0].CompareTo(0) == 0)" does not work.

Many Thanks.

[bltdavid]

All I really want is to make sure den[0] is a non-zero value.

The "if (den[0].ToCompare(0) == 0)" does not work.

If that's all you want, why is this not sufficient?

if (den[0] != 0)

[koganam]

If that's all you want, why is this not sufficient?

if (den[0] != 0)

You cannot compare doubles that way. Rounding considerations mean that you have to compare doubles by reference to a delta.

[aligator]

You cannot compare doubles that way. Rounding considerations mean that you have to compare doubles by reference to a delta.

Thanks,

Yes, I know I could use " if (den[0] != 0)" .

However, I am tying to understand why and how Stochastic indicator uses ApproxCompare without throwing an error.

Thanks.

[-=Edge=-]

However, I am tying to understand why and how Stochastic indicator uses ApproxCompare without throwing an error..

Make sure you are including this using statement


using NinjaTrader.Core.FloatingPoint;


Documentation for ApproxCompare in Help Guide Best Practices Here


https://ninjatrader.com/support/help..._practices.htm


1st thing under Misc ...

[aligator]

Make sure you are including this using statement


using NinjaTrader.Core.FloatingPoint;


Documentation for ApproxCompare in Help Guide Best Practices Here


https://ninjatrader.com/support/help..._practices.htm


1st thing under Misc ...

Thank you Edge,

I had noticed in best practices that ApproxCompare() is a Custom Extension Method. However, any reference to "Using declaration" was missing, and that was the key.

I had assumed that the CompareTo() is the equivalent method but there was no explanations.

Thanks a bunch.

[-=Edge=-]

I had assumed that the CompareTo() is the equivalent method but there was no explanations..

This might also help..

I have this link tagged as very informative information on working with floating point numbers..

https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/

[bltdavid]

You cannot compare doubles that way. Rounding considerations mean that you have to compare doubles by reference to a delta.

Of course you can. Maybe not always. It depends upon what you're doing.

(I know what you're referring to -- subtracting and comparing to double.epilson.
For example, see here.)

If you really want to know if a double value is not zero (because, say, it was manually
set to zero to indicate a starting value) this works fine.

In cases where you have a calculated value, this may or may not work. You have to understand the calculations being performed before reference to a delta becomes truly necessary.

[aligator]

If you really want to know if a double value is not zero

Thanks,

This was exactly what I wanted to do, dividing a double value by the difference between two other double series and wanted to avoid dividing by a possible zero.

Many thanks.

[koganam]

Of course you can. Maybe not always. It depends upon what you're doing.

(I know what you're referring to -- subtracting and comparing to double.epilson.
For example, see here.)

If you really want to know if a double value is not zero (because, say, it was manually
set to zero to indicate a starting value) this works fine.

In cases where you have a calculated value, this may or may not work. You have to understand the calculations being performed before reference to a delta becomes truly necessary.

"cannot" is not being used in the sense of impossibility: it is being used in the sense of not being defensive coding practice. Defensive coding requires using a delta to compare doubles, precisely because it obviates all issues, while also handling the particular instance that you cite of checking a value that was previously set, an instance where it is admittedly unnecessary to use a delta comparison.

[NinjaTrader_PatrickH]

Hello aligator,

Thank you for your post.

The 'using NinjaTrader.Core.FloatingPoint;' statement should be included by default when creating an indicator through the NinjaScript Editor. Do you have a use case where this using statement was not included?

Thank you Edge,

I had noticed in best practices that ApproxCompare() is a Custom Extension Method. However, any reference to "Using declaration" was missing, and that was the key.

I had assumed that the CompareTo() is the equivalent method but there was no explanations.

Thanks a bunch.

Information on CompareTo() can be found at the following link: https://docs.microsoft.com/en-us/dot...ramework-4.7.2

Please let me know if you have any questions.