ST6RI-886 Evaluation of functions from additional model libraries #727

seidewitz · 2026-01-10T00:01:12Z

This PR implements the (non-model-level) evaluation of all the functions from the following Kernel Function Library packages that had not been previously implemented:

CollectionFunctions
ControlFunctions
SequenceFunctions

Implemented Functions

The following Kernel Functions Library functions are implemented in this PR.

CollectionFunctions

There are no LibraryFunction implementations of these functions, but they all have complete functional specifications in the library model, which evaluate correctly given the other updates in this PR (except for the '#' index functions, which already have model-level evaluation implementations).

ControlFunctions

exists
forAll
maximize
minimize
reduce
reject
selectOne

DataFunctions

These functions were implemented to support the implementation of maximize and minimize.

max
min

SequenceFunctions

excludingAt
excluding
head
includesOnly
includingAt
including
intersection
last
equals
same
subsequence
tail
union

Refactoring

The PR also includes some refactoring of the structure of packages related to library function implementations.

The existing model-level evaluable LibraryFunction implementations in the package org.omg.sysml.expressions.function have been re-organized into Java subpackages based on the Kernel Function Library package of the function being implemented.
- org.omg.sysml.expressions.function.base
- org.omg.sysml.expressions.function.bool
- org.omg.sysml.expressions.function.control
- org.omg.sysml.expressions.function.data
Existing and new non-model-level evaluable LibraryFunction implementations in the package org.omg.sysml.execution.expressions.function have been similarly re-organized into subpackages.
- org.omg.sysml.execution.expressions.function.data
- org.omg.sysml.execution.expressions.function.numerical
- org.omg.sysml.execution.expressions.function.sequence
- org.omg.sysml.execution.expressions.function.string

Other Changes

Update the implementation of model-level evaluation of equals and not-equals functions to handle Collections.
Fixed the utility method EvaluationUtil::expressionFor when the argument is a list or a body expression.
Removed old hand-written overrides of getInput and getOutput in ExpressionImpl that are no longer necessary.

Created Java packages under corresponding to Kernel Function Library packages.

- Also added tests to ExpressionEvaluationTest.

Functions are evaluated using modeled return expressions (except for functions that are evaluated as model-level evaluable library functions).

seidewitz · 2026-01-10T00:03:28Z

@himi @hpdekoning
If one or both of you could do at least a quick review of this in the next few days, I would appreciate it. But, if not, I will go ahead and merge it anyway based on my own testing.

- Also implemented DataFunctions::min and DataFunctions::max, used in the implementation of ControlFunctions::minimize and maximize. - Also removed old overrides of getInput and getOutput in ExpressionImpl.

hpdekoning · 2026-01-10T21:41:23Z

...ml.execution/src/org/omg/sysml/execution/expressions/functions/control/MinimizeFunction.java

+						Type maxFunction = SysMLLibraryUtil.getLibraryType(expr, MIN_FUNCTION);
+						InvocationExpression maxInvocation = EvaluationUtil.createInvocationOf(maxFunction, result, exprValue.get(0));
+						EList<Element> newResult = evaluator.evaluateInvocation(maxInvocation, target);
+						return newResult == null || newResult.size() != 1? null: newResult.get(0);


I am guessing you meant naming these minFunction, minInvocation. Looks like a copy/paste typo, does not change behavior & result.

Oops, you are right. I've fixed it.

- maxFunction -> minFunction, maxInvocation -> minInvocaion

hpdekoning · 2026-01-10T22:23:05Z

org.omg.sysml/src/org/omg/sysml/expressions/ModelLevelLibraryFunctionFactory.java

-
+		// ControlFunctions
 		put(new DotFunction());
 		put(new ConditionalFunction());


Should this be named ConditionalIfFunction to show correspondence to KerML if function?

This function corresponds to what is called a "conditional expression" in KerML and is described as a "conditional test" function. So I think ConditionalFunction is an OK name.

hpdekoning

It looks great. I could only do some rather random spot checks.

You may want to check that the copyright statement is updated for 2026 everywhere, but perhaps that is a global check just before release.

himi · 2026-01-10T22:29:36Z

Thank you. I want to try these functions tonight. But please go ahead if you want to finish this PR.

himi · 2026-01-11T03:11:53Z

I briefly tested these functions:

himi · 2026-01-11T03:15:37Z

I could not find a written specification on these functions in KerML spec. Is the unionResult is intended? While intersection captures the equivalence of numbers, union does not. I found includesOnly seems to check the equality in set semantics, and please check it is intended.

himi · 2026-01-11T03:18:52Z

I could not make an example to check the difference of == and === and therefore could not check the difference of same and equals. I tried the example below, but both eq2 and eq3 are false. Even if I tried bind ts1 = ts2;, they are false. But if I try timeslice ts2 = ts1;, both are true.

part TestEq {
    timeslice ts1;
    timeslice ts2 {
        :>> startShot = ts1.startShot;
    }
    
    attribute eq2 = ts1 == ts2;
    attribute dq3 = ts1 === ts2;
}

seidewitz · 2026-01-11T19:30:13Z

@himi

I could not find a written specification on these functions in KerML spec. Is the unionResult is intended? While intersection captures the equivalence of numbers, union does not. I found includesOnly seems to check the equality in set semantics, and please check it is intended.

In general, many of the library functions are really not well specified right now. However, many of the sequence functions actually have complete definitions in SequenceFunctions.kerml, and the Java implementations are intended to be conform to these definitions (except for excludes which is erroneous defined circularly in terms of itself right now!).

The definition of union is that it operates on ordered, non-unique sequences and that it is equivalent to using the , operator:

	function union{ in seq1: Anything[0..*] ordered nonunique; in seq2: Anything[0..*] ordered nonunique;
		return : Anything[0..*] ordered nonunique = (seq1, seq2);
	}

So, unionResult is as expected.

The definition of intersection is

	function intersection{ in seq1: Anything[0..*] ordered nonunique; in seq2: Anything[0..*] ordered nonunique;
		return : Anything[0..*] ordered nonunique = seq1->select {in x; seq2->includes(x)};
	}

So, it selects the values in seq1, in order, including duplicates, that are also included inseq2 (regardless of duplicates in seq2). So intersectionResult is also as expected.

The definition of includesOnly is

	function includesOnly{ in seq1: Anything[0..*] nonunique; in seq2: Anything[0..*] nonunique;
		return : Boolean[1] = seq1->includes(seq2) and seq2->includes(seq1);
	}

And this is exactly how it is implement, by checking inclusion of each sequence in the other. The intent of includesOnly is to essentially provide a set-based equality check, as opposed to equals, which checks value-by-value equality in order.

One could possibly propose changes in how these functions are defined, but that would be a specification issue, not an implementation issue.

seidewitz · 2026-01-11T19:33:37Z

I could not make an example to check the difference of == and === and therefore could not check the difference of same and equals. I tried the example below, but both eq2 and eq3 are false. Even if I tried bind ts1 = ts2;, they are false. But if I try timeslice ts2 = ts1;, both are true.

There currently is no difference in the implementation of == and === (or SequenceFunctions::equals and SequenceFunctions::same). These functions are the same, anyway, for data values. And the current evaluation implementation does no support the semantics of change over time, so it does not include the difference between them for occurrences, either.

himi

@seidewitz thank you for confirming it. To be honest, union() seems confusing to me because it is actually concat(). But I understand it is intended. I think we need a proper documentation on these. Also I think we need unique() for set semantics and we can implement it with collect and reduce but the complexity will be O^2.

seidewitz added 5 commits December 12, 2025 13:56

ST6RI-886 Reorganized package structure library function impl classes.

f99b275

Created Java packages under corresponding to Kernel Function Library packages.

ST6RI-886 Implemented all functions in SequenceFunctions package.

27a5698

- Also added tests to ExpressionEvaluationTest.

ST6RI-886 Added Junit tests for CollectionFunctions.

78247a3

Functions are evaluated using modeled return expressions (except for functions that are evaluated as model-level evaluable library functions).

ST6RI-886 Fixed EvaluationUtil::expressionFor for lists and body exprs.

d3c89be

ST6RI-886 Implemented model-level eval of equality ops on collections.

a83e872

seidewitz added this to the 2025-12 milestone Jan 10, 2026

seidewitz self-assigned this Jan 10, 2026

seidewitz requested review from himi and hpdekoning January 10, 2026 00:01

ST6RI-886 Implemented remaining functions in ControlFunctions package.

b6822f4

- Also implemented DataFunctions::min and DataFunctions::max, used in the implementation of ControlFunctions::minimize and maximize. - Also removed old overrides of getInput and getOutput in ExpressionImpl.

seidewitz force-pushed the ST6RI-886 branch from a3b846b to b6822f4 Compare January 10, 2026 00:12

hpdekoning reviewed Jan 10, 2026

View reviewed changes

ST6RI-886 Corrected local variable naming in MinimizeFunction.

d9bb651

- maxFunction -> minFunction, maxInvocation -> minInvocaion

hpdekoning reviewed Jan 10, 2026

View reviewed changes

hpdekoning approved these changes Jan 10, 2026

View reviewed changes

ST6RI-886 Updated copyright date on LibraryFunctionFactory.java.

177e6c3

himi approved these changes Jan 12, 2026

View reviewed changes

ST6RI-886 Evaluation of functions from additional model libraries #727

Are you sure you want to change the base?

ST6RI-886 Evaluation of functions from additional model libraries #727

Uh oh!

Conversation

seidewitz commented Jan 10, 2026

Uh oh!

seidewitz commented Jan 10, 2026

Uh oh!

hpdekoning Jan 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

seidewitz Jan 10, 2026

Choose a reason for hiding this comment

Uh oh!

hpdekoning Jan 10, 2026

Choose a reason for hiding this comment

Uh oh!

seidewitz Jan 10, 2026

Choose a reason for hiding this comment

Uh oh!

hpdekoning left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

himi commented Jan 10, 2026

Uh oh!

himi commented Jan 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

himi commented Jan 11, 2026

Uh oh!

himi commented Jan 11, 2026

Uh oh!

seidewitz commented Jan 11, 2026

Uh oh!

seidewitz commented Jan 11, 2026

Uh oh!

himi left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

hpdekoning Jan 10, 2026 •

edited

Loading

hpdekoning left a comment •

edited

Loading

himi commented Jan 11, 2026 •

edited

Loading