Files
whetstone_DSL/editor/tests/step1940_test.cpp
Bill 84708da611 Sprint 282: Polyglot test harness (steps 1938–1942)
PolyglotSpecRunner, BehavioralParityChecker, ParityRegressionGuard,
PolyglotSuiteOrchestrator — 26/26 tests pass.
Integration step 1942: poly-sort Python+Rust+Go parity verified.
Metrics: 7 whetstone calls, 8855 tokens.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-01 22:21:40 -07:00

101 lines
3.2 KiB
C++

// Step 1940: ParityRegressionGuard
//
// t1: approve and isApproved/approvedCount
// t2: guard — all match baseline → passed=true, empty regressions/newLanguages
// t3: guard — output changed → regression detected
// t4: guard — unapproved language in results → newLanguages
// t5: updateBaseline changes expected output
#include "ParityRegressionGuard.h"
#include <iostream>
namespace ws = whetstone;
static int p=0,f=0;
#define T(n) { std::cout<<" "<<#n<<"... "; }
#define P() { std::cout<<"PASS\n"; ++p; }
#define F(m) { std::cout<<"FAIL: "<<m<<"\n"; ++f; }
#define C(c,m) if(!(c)){F(m);return;}
void t1(){
T(approve_and_metadata);
ws::ParityRegressionGuard g;
g.approve("sort","Python","1 2 3\n");
g.approve("sort","Rust", "1 2 3\n");
C( g.isApproved("sort","Python"), "Python approved");
C( g.isApproved("sort","Rust"), "Rust approved");
C(!g.isApproved("sort","Go"), "Go not approved");
C( g.approvedCount("sort") == 2, "two approved");
C( g.approvedCount("other") == 0, "other zero");
P();
}
void t2(){
T(guard_all_match_passes);
ws::ParityRegressionGuard g;
g.approve("sort","Python","1 2 3\n");
g.approve("sort","Rust", "1 2 3\n");
std::vector<ws::RunResult> results = {
{"Python","1 2 3\n",0,true},
{"Rust", "1 2 3\n",0,true}
};
auto r = g.guard("sort", results);
C(r.passed, "passed");
C(r.regressions.empty(), "no regressions");
C(r.newLanguages.empty(),"no new languages");
P();
}
void t3(){
T(guard_detects_regression);
ws::ParityRegressionGuard g;
g.approve("sort","Python","1 2 3\n");
g.approve("sort","Rust", "1 2 3\n");
std::vector<ws::RunResult> results = {
{"Python","1 2 3\n",0,true},
{"Rust", "3 2 1\n",0,true} // regression!
};
auto r = g.guard("sort", results);
C(!r.passed, "not passed");
C(r.regressions.size()==1, "one regression");
C(r.regressions[0]=="Rust","Rust regressed");
P();
}
void t4(){
T(guard_new_language_not_approved);
ws::ParityRegressionGuard g;
g.approve("sort","Python","ok\n");
std::vector<ws::RunResult> results = {
{"Python","ok\n",0,true},
{"Go", "ok\n",0,true} // Go not approved
};
auto r = g.guard("sort", results);
C(r.passed, "passed (no regressions)");
C(r.newLanguages.size()==1,"one new language");
C(r.newLanguages[0]=="Go", "Go is new");
P();
}
void t5(){
T(updateBaseline_changes_expected);
ws::ParityRegressionGuard g;
g.approve("sort","Rust","1 2 3\n");
g.updateBaseline("sort","Rust","sorted\n");
// old output now regresses
std::vector<ws::RunResult> results = {{"Rust","1 2 3\n",0,true}};
auto r1 = g.guard("sort", results);
C(!r1.passed, "old output is now regression");
// new output passes
std::vector<ws::RunResult> results2 = {{"Rust","sorted\n",0,true}};
auto r2 = g.guard("sort", results2);
C(r2.passed, "new baseline passes");
P();
}
int main(){
std::cout << "Step 1940: ParityRegressionGuard\n";
t1(); t2(); t3(); t4(); t5();
std::cout << "\n" << p << "/" << (p+f) << " passed\n";
return f > 0 ? 1 : 0;
}