Suppress unit printing and parsing for Dimensionless

[averbraeck]

The Dimensionless scalar, vector and matrix print a unit called unit under certain circumstances, and 1 as a unit under other circumstances. unit is not a sensible indication of the unit for a dimensionless number. 1 is not a valid SI-unit. Probably the best solution is to completely suppress the unit printing and parsing for the Dimensionless scalar, vector and matrix.

[OTSim]

I believe the word "unit" (that gets printed under some circumstances) comes from quantity/Quantity.java line 26.
In a way, this code contradicts the comment on line 24 of that same file.

[averbraeck]

The problem is that in the current setup, a unit's display string cannot have a length of 0. I have to allow that first before I can see whether all methods of the Dimensionless class such as of(String) still work. I will write some extra unit tests for the Dimensionless class to test whether this is working as intended.

[averbraeck]

probably, also the SIUnit class needs to be updated that is does not display a unit when the result is dimensionless. I believe it now prints 1 as the unit, which is clearly wrong.

[OTSim]

A unit of "1" is not totally bad. But a unit named ""unit"" is wrong. The evaluator (under development as part of djutils-eval) evaluates "PI()" to "3.14159265 unit" but "PI()/PI()" to "1.00000000 1".

[averbraeck]

A unit's name and id can now be blank, as long as the SI units show that the unit is dimensionless, and the scale is the IdentityScale. In the Unit constructor, this is checked as follows:

Throw.when(
  builder.getId().length() == 0
    && !(builder.getSiPrefixes().equals(SIPrefixes.NONE) 
    && builder.getScale().equals(IdentityScale.SCALE)),
  UnitRuntimeException.class, "Constructing unit %s: id.length cannot be 0", cName);

and

Throw.when(
  builder.getName().length() == 0
    && !(builder.getSiPrefixes().equals(SIPrefixes.NONE) 
    && builder.getScale().equals(IdentityScale.SCALE)),
  UnitRuntimeException.class, "Constructing unit %s.%s: name.length cannot be 0", cName, unitId);

[averbraeck]

The Dimensionless class has been updated and functions as expected. All unit tests pass again.

[averbraeck]

For the handling of a dimensionless SIScalar, a new issue has been opened as Issue #14.

[averbraeck]

The Dimensionless and FloatDimensionless classes seem to work as intended. All unit tests pass.

[OTSim]

Things are somewhat improved but there are still some inconsistencies. To reproduce the problem run:
System.out.println(new Dimensionless(10, DimensionlessUnit.SI).divide(new Dimensionless(3, DimensionlessUnit.SI)));
System.out.println(((DoubleScalarRel)new Dimensionless(10, DimensionlessUnit.SI)).divide(new Dimensionless(3, DimensionlessUnit.SI)));
This prints
3.33333333
3.33333333 1
N.B.: I propose to add a unit test that checks that the two divide operations in this demonstration code return identical results.

In the first case the division is carried out by Dimensionless.java(391) which simple sets the target unit to DimensionlessUnit.SI
In the second case the division is carried out by DoubleScalarRel.java(100) which does things "the hard way":
1: It calls DoubleScalar.divide to do the real work.
2: DoubleScalar.divide lives in file DoubleScalar.java at line 545. It calls Unit.lookupOrCreateUnitWithSIDimensions to determine the SIUnit of the result. This method construct an SIUnit that has id "1" and name "1". I presume that this is wrong.

By the way: DoubleScalarRel.java(100) is a divide method that returns SIScalar, but the javadoc says that it returns DoubleScalar<?>.

[averbraeck]

Issue #14 has taken care of the "1" that is not printed anymore.

[averbraeck]

So indeed, Issue #12 was not completely solved and needed Issue #14 to get rid of all the "1" units when the Dimensionless is expressed as a SIUnit scalar. I will close this issue again after all unit tests involving the "1" unit pass as well.

[averbraeck]

Issue #14 is now (almost) completed, which should solve the remaining problems of this issue, so I close it again.