[Rafael-SOWNet]

Part of #2195, item 3.

forall, exists, exists!, subset and card were refused by both compilers. Over a finite set, the set is known when the expression is compiled, so each one is a statement or a number about its members. CompiledMembership writes each out before either compiler sees it:

	becomes
forall, exists	the conjunction and the disjunction of the body at each member; True and False over the empty set
exists!	exactly one of them: the disjunction over i of P_i and not P_j for every other j, by Kleene's connectives
{ ... } subset S	each member a member of S, by #2223's membership rewrite
card({ ... })	the number of distinct values: each member counts, as an Iverson bracket, where it equals none before it

Each is capped at 64 members, since the cardinality and exists! grow with the square of the count. Past the cap, or over any set that is not finite, it is left as written, and neither compiler compiles it.

	2.6.0	now
"forall t in { 1, 2, 3 } : t < x".ToEntity().Compile("x").Call(4)	UncompilableNodeException	(1, 0)
"{ x } subset [0; 3]".ToEntity().Compile<double, bool>("x")(0.5)	UncompilableNodeException	True
"card({ 1, x })".ToEntity().Compile("x").Call(1)	UncompilableNodeException	(1, 0)

At NaN the compiled value is undecided, or the vacuous truth of a quantifier over the empty set. Evaluation differs there for two reasons:

• a finite set drops a member whose value is NaN, so { NaN } is { } and card({ 1, NaN }) is 1;
• a NaN anywhere in a quantifier's body makes the quantifier NaN.

Undecided answers nothing wrong, and the vacuous truth is the mathematics, so the difference is pinned by its own test rather than mirrored.

Tests. FiniteSetStatementCompilationTest compares both compilers with evaluation at 12 points, among them i, 1 + i and the infinities:

• nine statements, including nested quantifiers and quantifiers over the empty set;
• three cardinalities;
• the NaN cases;
• four refusals over sets that are not finite.

Full suite 18113/0, interactive trio, AOT smoke 73/0, all frameworks build.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.71%. Comparing base (fba9096) to head (fcda989).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2234      +/-   ##
==========================================
+ Coverage   86.68%   86.71%   +0.02%     
==========================================
  Files         310      310              
  Lines       67738    67766      +28     
  Branches    19649    19660      +11     
==========================================
+ Hits        58720    58760      +40     
+ Misses       4902     4894       -8     
+ Partials     4116     4112       -4     

Flag	Coverage Δ
cpp	37.77% <ø> (ø)
csharp	87.26% <100.00%> (+0.02%)	⬆️
fsharp	32.97% <ø> (ø)
interactive	40.72% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...riMath/Functions/Compilation/CompiledMembership.cs	97.87% <100.00%> (+3.13%)	⬆️

... and 10 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.