Float vs Double

CNEMCSE 3 days ago

Does Connect IQ actually support Double values? as far as I can see, the accuracy of all values seems to be limited to float, no matter how you declare them

Take the following code for example:

var myval1 = 1.123456789123456789 as Double;

var myval2 = 1.123456789123456789 as Float;

var valString1 = myval1.toString();

var valString2 = myval2.toString();

System.println("Double = " + valString1 + " Float = " + valString2);

The resulting output from this is:

Double = 1.123457 Float = 1.123457

Which would indicate to me that 'Double' is really just a Float...

FlowState 3 days ago

- 1.123456789123456789 is a float literal, as is any numerical literal in the form x.y. You need to add the d suffix to make it a double: 1.123456789123456789d

It would be nice if there were a compiler warning to tell you that your literal will lose precision because it's not specified as a double (or that the literal has too much precision even for a double), but there is none.

- The casts "as Double" and "as Float" do nothing to affect the value / type of the casted values at runtime. The only serve to override the compile-time type checker's determination of casted values' types (just like TypeScript). I would avoid casts as much as possible, as they defeat the purpose of type checking and arguably give the dev a false sense of security. (e.g. it's possible to "cast away" a type error a compile-time, which could lead to a crash at runtime)

- If you use Float/Double.toString() (or just concatenate a string with float or double, which has the same effect), you will get a default conversion with limited precision (6 decimal places). For additional precision, you need to use Float/Double.format()

Test code:

var doubleVal = 1.123456789123456789d;
var floatVal = 1.123456789123456789;

// doubleVal.toString() is implicitly called due to concatenation with a string System.println("doubleVal (toString) = " + doubleVal); // floatVal.toString() is implicitly called due to concatenation with a string
System.println("floatVal (toString) = " + floatVal);

System.println("doubleVal.format(\"%.20f\") = " + doubleVal.format("%.20f"));
System.println("floatVal.format(\"%.20f\") = " + floatVal.format("%.20f"));

// determine runtime types:
System.println("doubleVal instanceof Double: " + (doubleVal instanceof Double));
System.println("doubleVal instanceof Float: " + (doubleVal instanceof Float));
System.println("floatVal instanceof Double: " + (floatVal instanceof Double));
System.println("floatVal instanceof Float: " + (floatVal instanceof Float));

Output:

doubleVal (toString) = 1.123457 floatVal (toString) = 1.123457 doubleVal.format(\"%.20f\") = 1.12345678912345681155 floatVal.format(\"%.20f\") = 1.12345683574676513672 doubleVal instanceof Double: true
doubleVal instanceof Float: false
floatVal instanceof Double: false
floatVal instanceof Float: true

CNEMCSE 3 days ago in reply to FlowState

Ah! Excellent, so there is a trick I missed then... thanks for that!

jim_m_58 3 days ago

See https://developer.garmin.com/connect-iq/monkey-c/functions/

"Declaring Variables"

the "as Double" and "as Float" is only for the Type Checker.