import copy import json import sys import tempfile import threading import unittest from pathlib import Path from urllib.parse import parse_qs, urlsplit from unittest.mock import patch sys.path.insert(0, str(Path(__file__).resolve().parents[1] / 'scripts')) from nao_quality_tools import QualityTools, QualityToolError from test_nao_quality_tools import FakeApi, snapshot_period from nao_agent_task import normalize_task, bind_task from test_nao_agent_task import task_value from analysis_visuals import build_visuals from nao_evidence_finalizer import coverage_sufficient from nao_protocol_execution import execute from nao_weight_ranking import rank_response, ranking_chart, ranking_report, missing_ranking_visuals QUESTION = '做几张26年的亮片生产重量柏拉图,产品级、产业部级' ACTOR='11111111-1111-4111-8111-111111111111' SESSION='22222222-2222-4222-8222-222222222222' TURN='55555555-5555-4555-8555-555555555555' def signed_response(positive_rows=872, unavailable_rows=167, *, top_weight=None, metric='totalWeight', stage='black'): period={'year':2026,'monthStart':1,'monthEnd':9} total=sum([top_weight] if top_weight is not None else []) rows=[] for index in range(positive_rows): weight=(top_weight if index==0 and top_weight is not None else 1) total+=weight if not (index==0 and top_weight is not None) else 0 rows.append({'key':f'product-{index:04d}','label':f'规格{index:04d}', stage:{metric:weight},'coverage':{stage+'Rows':1}}) rows.append({'key':'mimo-negative','label':'mimo-negative',stage:{metric:-10.05},'coverage':{stage+'Rows':1}}) for index in range(unavailable_rows): rows.append({'key':f'no-stage-{index:04d}','label':f'未覆盖{index:04d}',stage:{metric:None}, 'coverage':{stage+'Rows':0}}) total-=10.05 months=[{'year':2026,'month':month,'blackRows':1 if stage=='black' else 0, 'brightRows':1 if stage=='bright' else 0} for month in range(1,10)] summary={stage:{metric:total}} return {'rows':rows,'context':{'rowCount':len(rows)},'summary':summary, 'coverage':{'rawRows':len(rows),'includedRows':len(rows),'sourceMonths':months}},period class WeightApi(FakeApi): def __init__(self, *, partial_year=True, hole=False, mismatch=False): super().__init__() self.partial_year, self.hole, self.mismatch = partial_year, hole, mismatch self.metric_query = 0 def call(self, path, payload=None, method=None): result = super().call(path, payload, method) if path == '/api/analysis/tools/catalog': result['sourceFields'] = [{'key':'customer_code','label':'客户编码','type':'source_value', 'stages':['bright'],'columns':{'bright':'客户编码'},'operators':['in']}] if path.startswith('/api/metrics?'): self.metric_query += 1 params = parse_qs(urlsplit(path).query) for month in result['coverage']['sourceMonths']:month['brightRows']=12 rows = [] for label, weight, scrap in [('scrap-leader', 10, 9), ('production-leader', 80, 1), ('third', 60, 2)]: row = self._product_row(label, label, 1, scrap) row['bright']['totalWeight'] = weight rows.append(row) result['rows'] = rows result['context']['rowCount'] = 3 result['summary']['bright']['totalWeight'] = 151 if self.mismatch else 150 result['summary']['bright']['scrapWeight'] = 12 if self.partial_year: result['coverage']['sourceMonths'] = result['coverage']['sourceMonths'][:9] result['coverage']['missingStages'] = [{'year':2026,'month':m,'stage':'bright'} for m in (10,11,12)] if self.hole: result['coverage']['sourceMonths'] = [m for m in result['coverage']['sourceMonths'] if m['month'] != 5] if 'monthStart' in params: result['context']['monthStart'] = int(params['monthStart'][0]) result['context']['monthEnd'] = int(params['monthEnd'][0]) if 'columnFilters' in params: result['filters'] = {'columnFilters':json.loads(params['columnFilters'][0])} result['protocolEvidence'] = {'queryId':f'rank-query-{self.metric_query}', 'attemptId':'attempt','sessionId':'session','turnId':'turn','revision':1, 'executionId':'execution','responseHash':'a'*64} return result class NegativeWeightApi(WeightApi): def call(self, path, payload=None, method=None): result=super().call(path,payload,method) if path.startswith('/api/metrics?'): params=parse_qs(urlsplit(path).query) if params.get('view',[''])[0]=='product': rows=[] for key,label,weight in (('large','大重量规格',100),('small','小重量规格',1),('mimo','mimo修正',-10.05)): row=self._product_row(key,label,1,1) row['black']['totalWeight']=weight rows.append(row) result['rows']=rows;result['context']['rowCount']=3 result['summary']['black']['totalWeight']=90.95 for month in result['coverage']['sourceMonths']:month['blackRows']=1 result['protocolEvidence']={'attemptId':SESSION,'sessionId':ACTOR,'turnId':TURN, 'revision':3,'executionId':TURN,'queryId':'signed-rank-query','responseHash':'b'*64} return result def planned(**updates): value={'dimensions':['product','department'],'stages':['bright'],'fields':['totalWeight'], 'period':{'year':2026},'visualPreference':'required','visualType':'pareto','periodCoverage':'available'} value.update(updates) return normalize_task(task_value(**value), QUESTION) class WeightRankingTests(unittest.TestCase): def run_rank(self, metric='totalWeight', **options): api = WeightApi(**options) with tempfile.TemporaryDirectory() as folder: QualityTools(api_factory=lambda:api, evidence_dir=Path(folder)).quality_data({ 'mode':'rank','releaseId':'release-a','period':{'year':2026}, 'dimension':'product','stage':'bright','metric':metric,'limit':2}) doc = json.loads(next(Path(folder).glob('nao-quality-*.json')).read_text()) return doc, api def test_ranks_all_products_by_requested_stage_weight_before_top_cut(self): doc, api = self.run_rank() result = doc['result'] self.assertEqual([row['label'] for row in result['rows']], ['production-leader','third']) self.assertEqual(result['totalWeight'], 150) self.assertEqual(result['rows'][0]['share'], 80/150) self.assertAlmostEqual(result['rows'][-1]['cumulativeShare'], 140/150) self.assertLess(result['rows'][-1]['cumulativeShare'], 1) self.assertEqual(result['rankingCoverage']['groups'], 3) self.assertFalse(any('rowLimit=' in path for _,path,_ in api.calls if path.startswith('/api/metrics'))) def test_black_total_weight_signed_reconciliation_keeps_all_absent_groups_auditable(self): response,period=signed_response() result=rank_response(response,{'id':'release-a','name':'A','status':'PUBLISHED'},period,{}, 'product','black','totalWeight',10) audit=result['rankingCoverage'] self.assertEqual(response['context']['rowCount'],1040) self.assertEqual(audit['sourceGroups'],1040) self.assertEqual(audit['groups'],873) self.assertEqual(audit['unavailableGroups'],167) self.assertEqual(audit['positiveGroups'],872) self.assertEqual(audit['negativeGroups'],1) self.assertEqual(audit['negativeWeight'],-10.05) self.assertEqual(audit['negativeWeightMagnitude'],10.05) self.assertEqual(result['totalWeight'],861.95) self.assertEqual(result['rows'][0]['weight'],1) self.assertTrue(all(row['weight']>0 for row in result['rows'])) self.assertAlmostEqual(result['rows'][0]['share'],1/861.95) doc={'evidenceRef':'QSIGNED10','result':result} task=planned(dimensions=['product'],stages=['black'],fields=['totalWeight'],period=period) chart=ranking_chart(doc,'signed-black-total',task) self.assertIsNotNone(chart) self.assertIn('1项负向投入重量修正',chart['note']) bad=copy.deepcopy(response);bad['summary']['black']['totalWeight']+=.1 with self.assertRaisesRegex(QualityToolError,'合计与同范围正式总量不一致'): rank_response(bad,{'id':'release-a'},period,{},'product','black','totalWeight',10) def test_negative_weights_remain_rejected_outside_black_total_weight(self): for stage,metric in (('black','scrapWeight'),('bright','totalWeight'),('bright','scrapWeight')): with self.subTest(stage=stage,metric=metric): response={'rows':[{'key':'negative','label':'负值',stage:{metric:-1},'coverage':{stage+'Rows':1}}], 'context':{'rowCount':1},'summary':{stage:{metric:10}}, 'coverage':{'sourceMonths':[{'year':2026,'month':month,'blackRows':1,'brightRows':1} for month in range(1,10)]}} with self.assertRaisesRegex(QualityToolError,'重量不能为负'): rank_response(response,{'id':'release-a'},{'year':2026,'monthStart':1,'monthEnd':9}, {},'product',stage,metric,10) def test_signed_black_total_reports_correction_and_blocks_over_100_chart_as_partial(self): response,period=signed_response(positive_rows=2,unavailable_rows=0,top_weight=100) result=rank_response(response,{'id':'release-a','name':'A','status':'PUBLISHED'},period,{}, 'product','black','totalWeight',10) self.assertEqual(result['totalWeight'],90.95) self.assertEqual([row['label'] for row in result['rows']],['规格0000','规格0001']) self.assertGreater(result['rows'][-1]['cumulativeShare'],1) document={'evidenceRef':'QNEGATIVE1','result':result} task=planned(dimensions=['product'],stages=['black'],fields=['totalWeight'],period=period, visualPreference='required') context=bind_task({'releaseId':'release-a'},task) self.assertIsNone(ranking_chart(document,'negative-correction',task)) gap=missing_ranking_visuals(task,[],[document]) self.assertEqual(len(gap),1) self.assertIn('负向修正使正重量Top合计超过黑片净总重',gap[0]) report=ranking_report([document],task,context) self.assertIn('1项负向投入重量修正,净额-10.05 kg仍计入正式净总重分母',report['facts'][0]['text']) section=report['sections'][0] self.assertGreater(section.index('排名表只展示正重量Top'),section.index('| 1 |')) def test_protocol_execution_keeps_signed_denominator_but_returns_partial_without_over_100_chart(self): t=normalize_task(task_value(kind='analysis',dimensions=['product'],stages=['black'], dimensionStages={'product':['black']},fields=['totalWeight'],period={'year':2026}, trendRequired=False,visualPreference='auto',visualType='pareto'), '按产品查看黑片投入重量排名',{'_protocol':{'version':1}}) if t['filters'].get('columnFilters')==[]:t['filters'].pop('columnFilters') api=NegativeWeightApi() with tempfile.TemporaryDirectory() as folder,patch('nao_evidence_finalizer.synthesize',return_value={ 'answer':'已核验黑片修正重量排名','evidenceSufficient':True,'missingEvidence':[],'usage':{}}): root=Path(folder);(root/'evidence').mkdir();(root/'private').mkdir() result=execute(request={'releaseId':'release-a','turnId':TURN,'question':t['question']},task=t, authority={'attemptId':SESSION,'sessionId':ACTOR,'turnId':TURN,'revision':3,'executionId':TURN}, actor_root=root,delegated=api,connection=('offline','synthetic','model'), cancel_event=threading.Event(),emit=lambda _:None) doc=json.loads(next((root/'evidence').glob('nao-quality-*.json')).read_text()) self.assertEqual(doc['result']['totalWeight'],90.95) self.assertEqual(doc['result']['rankingCoverage']['negativeGroups'],1) self.assertEqual(doc['result']['rankingCoverage']['negativeWeight'],-10.05) self.assertEqual(result['status'],'PARTIAL') self.assertTrue(any('负向修正使正重量Top合计超过黑片净总重' in item for item in result['evidenceCompletion']['missingEvidence'])) self.assertFalse(result['visuals']) def test_rank_retains_column_filter_month_range_and_execution_proof(self): api=WeightApi() rule={'field':'customer_code','stages':['bright'],'operator':'in','values':['0012']} with tempfile.TemporaryDirectory() as folder: QualityTools(api_factory=lambda:api,evidence_dir=Path(folder)).quality_data({ 'mode':'rank','releaseId':'release-a','period':{'year':2026,'monthStart':1,'monthEnd':9}, 'filters':{},'columnFilters':[rule],'dimension':'product','stage':'bright', 'metric':'totalWeight','limit':2}) doc=json.loads(next(Path(folder).glob('nao-quality-*.json')).read_text()) result=doc['result'] self.assertEqual(result['filters']['columnFilters'],[rule]) self.assertEqual(result['period'],{'year':2026,'monthStart':1,'monthEnd':9}) self.assertEqual(result['periodCoverage']['availableMonths'],list(range(1,10))) self.assertTrue(result['periodCoverage']['complete']) self.assertEqual(result['protocolEvidence']['queryId'],'rank-query-1') metrics_paths=[path for _,path,_ in api.calls if path.startswith('/api/metrics')] self.assertEqual(len(metrics_paths),1) self.assertIn('columnFilters=',metrics_paths[0]) self.assertNotIn('rowLimit=',metrics_paths[0]) self.assertNotIn('rowOffset=',metrics_paths[0]) def test_rank_only_allows_column_filter_parameter_and_other_modes_reject_it(self): api=WeightApi() with tempfile.TemporaryDirectory() as folder: tools=QualityTools(api_factory=lambda:api,evidence_dir=Path(folder)) with self.assertRaisesRegex(QualityToolError,'query/compare/rank'): tools.quality_data({'mode':'changes','releaseId':'release-a','period':{'year':2026}, 'comparisonPeriod':{'year':2025},'dimension':'group','columnFilters':[]}) self.assertEqual(api.calls,[]) def test_ranking_chart_scope_preserves_exact_continuous_month_range(self): from nao_weight_ranking import ranking_chart doc,_=self.run_rank() doc['result']['period']={'year':2026,'monthStart':3,'monthEnd':5} t=planned();t['period']=doc['result']['period'] chart=ranking_chart(doc,'range-ranking-turn',t) self.assertIsNotNone(chart) self.assertEqual({key:chart['scope'][key] for key in ('year','monthStart','monthEnd')},t['period']) def test_pareto_keeps_top_rows_but_binds_ratio_to_full_population(self): doc, _ = self.run_rank() task = planned() context = bind_task({'releaseId':'release-a'},task) output = build_visuals([doc],QUESTION,'turn-ranking',context=context) self.assertEqual(output['visualState'],'ready') chart = output['visuals'][0] self.assertEqual(chart['type'],'pareto') self.assertEqual(chart['coverage']['denominator'],150) self.assertEqual(chart['coverage']['shareBasis'],'all_scope_weight') self.assertIn('1—9月',chart['subtitle']) self.assertEqual(len(chart['categories']),2) department=copy.deepcopy(doc);department['result']['dimension']='department';department['evidenceRef']='Q00000002' output=build_visuals([doc,department],QUESTION,'turn-ranking',context=context) self.assertEqual({v['dimension'] for v in output['visuals']},{'product','department'}) from nao_weight_ranking import missing_ranking_visuals self.assertEqual(missing_ranking_visuals(task,output['visuals']),[]) self.assertTrue(missing_ranking_visuals(task,output['visuals'][:1])) def test_pareto_emits_each_requested_weight_field_for_the_same_stage(self): total_doc,_=self.run_rank() scrap_doc,_=self.run_rank(metric='scrapWeight') task=planned(fields=['totalWeight','scrapWeight'],dimensions=['product'],visualPreference='auto') context=bind_task({'releaseId':'release-a'},task) output=build_visuals([total_doc,scrap_doc],QUESTION,'two-field-ranking',context=context) self.assertEqual(output['visualState'],'ready') self.assertEqual({visual['metric'] for visual in output['visuals']}, {'bright_total_weight','bright_scrap_weight'}) from nao_weight_ranking import missing_ranking_visuals self.assertEqual(missing_ranking_visuals(task,output['visuals']),[]) def test_missing_dimension_is_not_completion_and_year_data_is_not_called_absent(self): doc, _ = self.run_rank() task = planned() context = bind_task({'releaseId':'release-a'},task) self.assertFalse(coverage_sufficient([doc],QUESTION,context)) department = copy.deepcopy(doc) department['result']['dimension'] = 'department' department['evidenceRef'] = 'Q00000002' self.assertTrue(coverage_sufficient([doc,department],QUESTION,context)) complete = planned(periodCoverage='complete') self.assertFalse(coverage_sufficient([doc,department],QUESTION,bind_task({'releaseId':'release-a'},complete))) def test_internal_month_holes_and_total_mismatch_do_not_pass_as_complete(self): doc, _ = self.run_rank(hole=True) context = bind_task({'releaseId':'release-a'},planned()) self.assertFalse(coverage_sufficient([doc],QUESTION,context)) with self.assertRaises(QualityToolError):self.run_rank(mismatch=True) def test_enum_strings_are_normalized_without_guessing_business_intent(self): task = normalize_task(task_value(dimensions='product,department',stages='bright',fields='totalWeight'),QUESTION) self.assertEqual(task['dimensions'],['product','department']) self.assertEqual(task['stages'],['bright']) def test_each_ranking_dimension_uses_only_its_declared_stage(self): doc, _ = self.run_rank() department = copy.deepcopy(doc) department['result'].update(dimension='department', stage='black') department['evidenceRef'] = 'Q00000002' task = planned() task['stages'] = ['black', 'bright'] task['dimensionStages'] = {'product': ['bright'], 'department': ['black']} context = bind_task({'releaseId': 'release-a'}, task) self.assertTrue(coverage_sufficient([doc, department], QUESTION, context)) charts = build_visuals([doc, department], QUESTION, 'mixed-stage-ranking', context=context)['visuals'] self.assertEqual({(v['dimension'], v['stage']) for v in charts}, {('product', 'bright'), ('department', 'black')}) self.assertTrue(all(v['coverage']['denominator'] == 150 for v in charts)) from nao_weight_ranking import missing_ranking_visuals self.assertEqual(missing_ranking_visuals(task, charts), []) wrong = copy.deepcopy(department); wrong['result']['stage'] = 'bright' self.assertFalse(coverage_sufficient([doc, wrong], QUESTION, context)) if __name__ == '__main__':unittest.main()