Datasets:
Modalities:
Text
Formats:
json
Languages:
English
Size:
< 1K
Tags:
mathematics
formal-verification
symbolic-algebra
theorem-proving
multi-agent-systems
benchmark
Download tests/test_zero_exponent.py from tailot/lingologos-math: direct link, hf CLI and curl.
- Browser
- Download file 3.34 kB
-
https://huggingface.co/datasets/tailot/lingologos-math/resolve/main/tests/test_zero_exponent.py
- Command line
-
hf download hf://datasets/tailot/lingologos-math/tests/test_zero_exponent.py
-
curl -L -o test_zero_exponent.py https://huggingface.co/datasets/tailot/lingologos-math/resolve/main/tests/test_zero_exponent.py
3.34 kB
| """The zero exponent: inside the trusted base x^0 is the constant 1, also when the base is 0. | |
| The trusted base reads a power with exponent 0 as the empty product, so it is 1 for every base, which | |
| is what eval_term and the polynomial builder already did. Z3 5.x leaves (^ x 0.0) unconstrained where | |
| the base is 0, so the two directions of the verification disagreed on exactly that point and every | |
| problem containing a zero exponent ended as UNKNOWN. Both SMT emitters therefore substitute the | |
| constant 1. The tests pin the convention and the emitters; they do not pin what a Z3 release does with | |
| an undefined power, which is why that behaviour is not asserted anywhere. | |
| """ | |
| from fractions import Fraction as F | |
| import z3 | |
| from lgmath import exact, pipeline, smt | |
| def power(base, exponent): | |
| return ("pow", base, ("num", F(exponent))) | |
| def test_the_exact_reading_takes_the_zero_exponent_as_one(): | |
| for base in (("var", "x"), ("num", F(0)), ("add", ("num", F(1)), ("neg", ("num", F(1))))): | |
| node = power(base, 0) | |
| for x in (F(0), F(5), F(-7, 3)): | |
| assert exact.eval_formula(("eq", node, ("num", F(1))), {"x": x}), base | |
| assert not exact.eval_formula(("ne", node, ("num", F(1))), {"x": x}), base | |
| assert exact.eval_formula(("eq", power(("var", "x"), 0), ("num", F(1))), {"x": F(0)}) | |
| assert not exact.eval_formula(("eq", power(("var", "x"), 0), ("num", F(0))), {"x": F(0)}) | |
| def test_both_smt_emitters_write_the_constant_one(): | |
| for base in (("var", "x"), ("num", F(0))): | |
| node = power(base, 0) | |
| text = smt.term_smt(node) | |
| assert text == "1.0", text | |
| assert "^" not in text, text | |
| value = smt.z3_term(node, {}) | |
| solver = z3.Solver() | |
| solver.add(value != z3.RealVal("1/1")) | |
| assert solver.check() == z3.unsat, base | |
| def test_a_non_zero_exponent_is_still_a_power(): | |
| text = smt.term_smt(power(("var", "x"), 3)) | |
| assert "^" in text or "*" in text, text | |
| assert smt.term_smt(power(("var", "x"), 0)) != text | |
| env = {"x": z3.Real("x")} | |
| solver = z3.Solver() | |
| solver.add(smt.z3_term(power(("var", "x"), 3), env) != env["x"] ** 3) | |
| assert solver.check() == z3.unsat | |
| def test_problems_with_a_zero_exponent_reach_a_verdict(): | |
| proved = pipeline.run("prove x^0 = 1")["final"] | |
| assert proved["status"] in ("VERIFIED", "SINGLE_PROVER"), proved["reasons"] | |
| assert proved["answer"]["verdict"] == "proved" | |
| identity = pipeline.run("solve x^0 = 1 for x")["final"] | |
| assert identity["status"] in ("VERIFIED", "SINGLE_PROVER"), identity["reasons"] | |
| assert identity["answer_text"] == "(-inf, +inf)", identity["answer_text"] | |
| single = pipeline.run("solve x^0 = x for x")["final"] | |
| assert single["answer_text"] == "{1}", single["answer_text"] | |
| numeric = pipeline.run("evaluate 0^0")["final"] | |
| assert numeric["answer_text"] == "1", numeric["answer_text"] | |
| def test_the_zero_exponent_does_not_hide_a_real_condition(): | |
| shifted = pipeline.run("solve (x - 1)^0 = x for x")["final"] | |
| assert shifted["status"] in ("VERIFIED", "SINGLE_PROVER"), shifted["reasons"] | |
| assert shifted["answer_text"] == "{1}", shifted["answer_text"] | |
| false_on_one_point = pipeline.run("prove forall x. x^0 != 0")["final"] | |
| assert false_on_one_point["answer"]["verdict"] == "proved", false_on_one_point["answer_text"] | |