test: Cover float comparison ops and WhileStmt float condition by Jaskirat-s7 · Pull Request #239 · arxlang/irx

Jaskirat-s7 · 2026-03-18T09:49:36Z

PR Title

test(#38): cover float comparison operators and WhileStmt float condition

⚠️ Stacked PR — This branch builds on top of the Phase 2 PR (test/issue-38-uninit-var-fix) which is already under review. The Phase 2 changes (IfStmt float coverage + VariableDeclaration zero-init) are in that PR. Only the 2 commits below are new in Phase 3:

bf4c7aa — test_while_float_condition (WhileStmt float condition via fcmp_ordered)

e85c85f — Float comparison operators <=, >=, ==, != with Float32

PR Description

What this PR does

This PR adds 8 new tests across 2 test files that cover previously untested float-type code paths in LLVMLiteIRVisitor. No production code was changed — only test files.

Newly covered compiler paths

Test	Operator	Lines covered	What it proves
test_binary_op_float_le	`Float32 <=`	1115–1117	`fcmp_ordered <=` emitted correctly
test_binary_op_float_ge	`Float32 >=`	1125–1127	`fcmp_ordered >=` emitted correctly
test_binary_op_float_eq	`Float32 ==`	1165–1168	`fcmp_ordered ==` emitted correctly
test_binary_op_float_ne	`Float32 !=`	1188–1190	`fcmp_ordered !=` emitted correctly
test_while_float_condition	`WhileStmt + Float32`	1362–1363	Float loop condition cast to `i1` correctly

All 5 of these code paths previously had 0% coverage.

Why the overall percentage looks unchanged

The 86% figure is a rounded integer over 1,793 tracked lines. Each additional covered line moves the raw percentage by ~0.056%. Covering 5 new lines moves it from 85.96% → 86.24% — both display as 86%. The raw missed line count dropped from 211 → 208 in llvmliteir.py.

To be transparent: a visible percentage jump to 87% requires covering ~18 lines in one PR. These tests cover 5 lines but specifically target previously untested float comparison semantics.

Why these tests matter (not just coverage)

The original issue (#38) was partly about tests that only verified the compiler didn't crash, not that it produced correct output. Every test in this PR:

Compiles the AST → LLVM IR → native binary via Clang
Executes the binary and asserts the printed string matches the expected result
Tests both branches (e.g., 2.0 <= 3.0 → "le_true" AND 3.0 <= 2.0 → "le_false")

For example, test_binary_op_float_le proves that 2.0 <= 3.0 produces "le_true" at runtime — not just that it compiles without crashing.

Test count

Before: 210 tests
After: 218 tests (all passing)

Files changed

tests/test_binary_op.py — +211 lines (4 new test functions, 8 parametrized cases)
tests/test_while.py — +54 lines (1 new test function)

yuvimittal · 2026-03-19T08:35:06Z

tests/test_binary_op.py

+@pytest.mark.parametrize(
+    "a_val, b_val, op, expected",
+    [
+        (2.0, 3.0, "<=", "le_true"),


what is le here?

le is short for "less than or equal" (<=). le_true is printed when the condition holds (2.0 <= 3.0), and le_false
when it doesn't (3.0 <= 2.0). I am happy to rename to something clearer like "yes" / "no" or "lte_pass" / "lte_fail" if preferred.

yuvimittal · 2026-03-19T08:38:21Z

tests/test_while.py

+
+
+@pytest.mark.parametrize("builder_class", [LLVMLiteIR])
+def test_while_float_condition(


we can parametrize it with the basic while functionality, right?

I have added Float32 and Float64 to test_while_expr. One issue came up during implementation was that the original increment used UnaryOp("++"), which generates an integer add instruction in LLVM IR. When floats were added, this produced add double which is invalid — floats need fadd.
It was fixed by replacing the increment with VariableAssignment + BinaryOp("+"), which correctly dispatches to fadd for floats and add for integers. All 6 types now pass (Int8/Int16/Int32/Int64/Float32/Float64).

yuvimittal · 2026-03-19T08:47:11Z

tests/test_binary_op.py

+    module = builder.module()
+
+    cond = astx.BinaryOp(
+        op_code=">=",


i don't understand why are we hardcoding operator, op should be accepted as parameter

I have merged all 4 functions (test_binary_op_float_le, test_binary_op_float_ge, test_binary_op_float_eq, test_binary_op_float_ne) into a single test_binary_op_float_comparison that accepts op_code as a parameter. The parametrize list now covers all 4 float comparison operators (<=, >=, ==, !=) with both true and false branches , 8 cases total, all are passing.

yuvimittal · 2026-03-19T08:47:49Z

tests/test_binary_op.py

+
+
+@pytest.mark.parametrize(
+    "a_val, b_val, expected",


@pytest.mark.parametrize(
"a_val, b_val, op, true_label, false_label",
[
(2.0, 3.0, "<=", "le_true", "le_false"),
(3.0, 2.0, ">=", "ge_true", "ge_false"),
(2.0, 2.0, "==", "eq_true", "eq_false"),
(1.0, 2.0, "!=", "ne_true", "ne_false"),
],
)

i have updated to use your suggested structure exactly. 4 rows with a_val, b_val, op, true_label, false_label as separate parameters. true_label is used in the then block and asserted as output, false_label in the else
block. All 4 cases pass.

Jaskirat-s7 · 2026-03-19T19:45:02Z

@yuvimittal ,I have addressed all 4 review comments in the latest push:

1."what is le here?" — le/ ge/eq/ne are standard abbreviations for "less than or equal", "greater than or equal", etc. I have kept the labels as-is since they're consistent with operator naming conventions.
2."op should be accepted as parameter" — i have merged the 4 separate float comparison functions (float_le, float_ge, float_eq, float_ne) into a single test_binary_op_float_comparison accepting op as a parameter.
3.Suggested parametrize structure with true_label, false_label — i have updated to your exact suggestion: 4 rows with a_val, b_val, op, true_label, false_label as explicit parameters.
4."we can parametrize it with basic while functionality, right?" — i have added Float32/Float64 to test_while_expr. One issue: UnaryOp("++") generates add double (invalid for floats — needs fadd) , fixed by replacing the increment with VariableAssignment + BinaryOp("+"), which correctly dispatches fadd/add by type. All 6 types now pass (Int8/16/32/64 + Float32/64).

…n zero-init paths - Add Float32 to test_if_else_stmt and test_if_only_stmt parametrize lists - Add test_variable_declaration_no_initializer for Int32 and Float32 - Fix: VariableDeclaration visitor checked 'node.value is not None' but astx defaults value to astx.Undefined(); added isinstance check so uninitialized vars reach the zero-init path (llvmliteir.py lines 3023-3059)

…eedback

… review

…tion_no_initializer Add PrintExpr and expected_output assertion ('0' for Int32, '0.000000' for Float32/Float64) so the test proves default-to-zero behavior at runtime, not just that the build does not crash. Also add Float64 to the parametrize list.

… review - add PrintExpr + expected_output assertion ('0' for ints, '0.000000' for floats) to prove zero-default at runtime, not just build success - remove unused literal_type parameter from signature and parametrize - add Int8, Int16, Int64 alongside Int32/Float32/Float64 for consistency

Add test_while_float_condition to test_while.py which: - Initializes a Float32 variable to 3.0 - Uses it directly as the WhileStmt condition (triggers fcmp_ordered conversion) - Decrements it each iteration until 0.0 - Asserts 'done' is printed proving the loop ran and terminated Covers previously uncovered lines 1362-1363 in llvmliteir.py (fcmp_ordered float-to-i1 cast path inside WhileStmt visitor).

Add 4 parametrized tests to test_binary_op.py targeting previously uncovered float comparison paths in LLVMLiteIRVisitor: - test_binary_op_float_le -> covers line 1115 (fcmp_ordered <=) - test_binary_op_float_ge -> covers line 1125 (fcmp_ordered >=) - test_binary_op_float_eq -> covers line 1165 (fcmp_ordered ==) - test_binary_op_float_ne -> covers line 1188 (fcmp_ordered !=) Each test has two parametrized cases (true/false branches) with expected_output assertions proving semantically correct output.

…le tests - merge test_binary_op_float_le/ge/eq/ne into single parametrized test_binary_op_float_comparison accepting op, true_label, false_label - replace UnaryOp('++') with VariableAssignment+BinaryOp('+') in test_while_expr to support Float32/Float64 (++ generates 'add' not 'fadd' for floats causing invalid LLVM IR) - add Float32 and Float64 to test_while_expr parametrize list

Jaskirat-s7 force-pushed the feature-issue-38-phase3 branch from e85c85f to 65e4ce7 Compare March 18, 2026 09:59

Jaskirat-s7 mentioned this pull request Mar 18, 2026

test: Cover IfStmt float condition and VariableDeclaration zero-init paths #236

Open

yuvimittal requested changes Mar 19, 2026

View reviewed changes

Jaskirat-s7 force-pushed the feature-issue-38-phase3 branch 2 times, most recently from fe37ecd to 18f6ff6 Compare March 19, 2026 19:39

Jaskirat-s7 and others added 11 commits March 20, 2026 01:49

fix: add missing Type import in test_variable_assignment.py

0c79d66

test(arxlang#38): add Float64 to IfStmt parametrize list per review f…

e8df269

…eedback

Fix commented-out test case in variable assignment test

4f44a0b

Fix comment formatting in test_variable_assignment.py

910fa41

test(arxlang#38): add Float32/Float64 to test_variable_assignment per…

bd9e76f

… review

Jaskirat-s7 force-pushed the feature-issue-38-phase3 branch from 18f6ff6 to 8c9278a Compare March 19, 2026 20:19

Jaskirat-s7 requested a review from yuvimittal March 21, 2026 08:54

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test: Cover float comparison ops and WhileStmt float condition#239

test: Cover float comparison ops and WhileStmt float condition#239
Jaskirat-s7 wants to merge 11 commits intoarxlang:mainfrom
Jaskirat-s7:feature-issue-38-phase3

Jaskirat-s7 commented Mar 18, 2026

Uh oh!

yuvimittal Mar 19, 2026

Uh oh!

Jaskirat-s7 Mar 19, 2026

Uh oh!

yuvimittal Mar 19, 2026

Uh oh!

Jaskirat-s7 Mar 19, 2026

Uh oh!

yuvimittal Mar 19, 2026

Uh oh!

Jaskirat-s7 Mar 19, 2026

Uh oh!

yuvimittal Mar 19, 2026

Uh oh!

Jaskirat-s7 Mar 19, 2026

Uh oh!

Jaskirat-s7 commented Mar 19, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants



		@pytest.mark.parametrize("builder_class", [LLVMLiteIR])
		def test_while_float_condition(

Conversation

Jaskirat-s7 commented Mar 18, 2026

PR Title

PR Description

What this PR does

Newly covered compiler paths

Why the overall percentage looks unchanged

Why these tests matter (not just coverage)

Test count

Files changed

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Jaskirat-s7 commented Mar 19, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants