def check_export(tmpdir, ds, variant_name, csv_fname):
if OVERWRITE_EXPECTED_FILES:
out_fname = csv_fname
else:
out_fname = os.path.join(tmpdir, '%s.csv' % variant_name)
variant = ds._get_export_variant(variant_name)
assert variant
ds.export(out_fname, '*.csv', variant, MockWorker())
csv = load_raw_csv(csv_fname)
expected_cols = csv_columns(csv[0], csv[1:])
csv = load_raw_csv(out_fname)
cols = csv_columns(csv[0], csv[1:])
keys_expected = set(expected_cols.keys())
keys = set(cols.keys())
common_keys = keys_expected & keys
all_keys = keys_expected | keys
assert 2 * len(common_keys) >= len(all_keys) # at least 1/2 common keys
# run with pytest -s instead
#print(common_keys)
for key in [
'subject'
] + sorted(common_keys): # sort to make error locations deterministic
assert len(expected_cols[key]) == len(cols[key]), key
assert expected_cols[key] == cols[key], key
def test_consistency_analysis(tmpdir, name, alts, subj_count):
rows = load_raw_csv('docs/src/_static/examples/%s.csv' % name)
ds = ExperimentalData.from_csv('dataset', rows[1:], (0, 1, None, 2))
assert ds.alternatives == alts.split()
assert len(ds.subjects) == subj_count
dsc = ds.analysis_consistency(MockWorker(), None)
assert len(dsc.subjects) == len(ds.subjects)
check_export(tmpdir, dsc, 'summary',
'gui/test/expected/%s-cons-summary.csv' % name)
check_export(tmpdir, dsc, 'congruence violations (wide)',
'gui/test/expected/%s-cons-garp.csv' % name)
check_export(tmpdir, dsc, 'strict general cycles (wide)',
'gui/test/expected/%s-cons-sarp.csv' % name)
check_export(tmpdir, dsc, 'strict binary cycles (wide)',
'gui/test/expected/%s-cons-sarp-bin.csv' % name)
check_export(tmpdir, dsc, 'binary cycles (wide)',
'gui/test/expected/%s-cons-garp-bin.csv' % name)
dst_menus = ds.analysis_tuple_intrans_menus(MockWorker(), None)
dst_alts = ds.analysis_tuple_intrans_alts(MockWorker(), None)
assert len(dst_menus.subjects) == len(ds.subjects)
assert len(dst_alts.subjects) == len(ds.subjects)
def load_from_csv(fname : str) -> Budgetary:
lines = dataset.load_raw_csv(fname)
if not lines:
raise BudgetaryError("the CSV file is empty")
header, *rows = lines
if (len(header)-1) % 2 != 0:
raise BudgetaryError("budgetary datasets should have an even number of numeric columns")
n_alts = (len(header)-1) // 2
alternatives = [f'{i+1}' for i in range(n_alts)]
subjects : Dict[str,Tuple[List[np.array],List[np.array]]] = dict() # ordered
for line_no, row in enumerate(rows, start=2):
if len(row) != len(header):
raise BudgetaryError(f'{fname}, line {line_no}: incorrect number of columns')
subj_name, *cols = row
if subj_name not in subjects:
subjects[subj_name] = ([], [])
prices, amounts = subjects[subj_name]
prices.append(np.array([float(x) for x in cols[:n_alts]], dtype=np.float32))
amounts.append(np.array([float(x) for x in cols[n_alts:]], dtype=np.float32))
ds = Budgetary(os.path.basename(fname), alternatives)
ds.subjects = [
Subject(name=n, prices=np.vstack(ps), amounts=np.vstack(ams))
for (n,(ps,ams)) in subjects.items()
]
ds.update_nr_observations()
return ds
def run(self):
fname: Optional[str] = None
def work():
assert fname is not None
ds = self.make_dataset(name=os.path.basename(fname))
self.main_win.add_dataset(ds)
fname, _something = QFileDialog.getOpenFileName(
self, "Import CSV", filter="CSV files (*.csv)")
if not fname:
return
rows = dataset.load_raw_csv(fname)
if not rows:
QMessageBox.warning(
self,
"CSV import",
"The input file seems to be empty",
)
return
self.fill_rows(rows)
self.accepted.connect(self.catch_exc(work))
self.exec_()
def test_integrity(tmpdir):
rows = load_raw_csv('docs/src/_static/examples/integrity.csv')
ds = ExperimentalData.from_csv('dataset', rows[1:], (0, 1, None, 2))
nds = ds.analysis_integrity_check(MockWorker(), None)
assert isinstance(nds, dataset.integrity_check.IntegrityCheck)
assert len(nds.subjects) == 1
assert nds.subjects[0].name == 'a'
assert nds.subjects[0].issues == [
dataset.integrity_check.RepeatedMenu(menu={0, 1}, ),
dataset.integrity_check.ChoiceNotInMenu(
menu={0, 1},
choice=2,
),
]
def estimate(args):
rows = load_raw_csv(args.fname_in)
ds = ExperimentalData.from_csv('dataset', rows[1:], (0, 1, None, 2))
AVAILABLE_MODELS = [
preorder(strict=True, total=True),
preorder(strict=False, total=True),
unattractive(strict=True, total=True),
unattractive(strict=False, total=True),
preorder(strict=True, total=False),
preorder(strict=False, total=False),
UndominatedChoice(strict=True),
UndominatedChoice(strict=False),
PartiallyDominantChoice(fc=True),
PartiallyDominantChoice(fc=False),
Overload(PreorderParams(strict=True, total=True)),
Overload(PreorderParams(strict=False, total=True)),
StatusQuoUndominatedChoice(),
TopTwo(),
SequentiallyRationalizableChoice(),
Swaps(),
]
if not args.models:
print('Please specify a model using -m:')
for m in AVAILABLE_MODELS:
print(' ' + str(m))
sys.exit(1)
if args.models == 'all':
models = AVAILABLE_MODELS
else:
models = [m for m in AVAILABLE_MODELS if str(m) in args.models]
if not models:
raise Exception('bad model spec')
dsm = ds.analysis_estimation(
ProgressWorker(),
EstimationOpts(
models=models,
disable_parallelism=args.sequential,
disregard_deferrals=args.disregard_deferrals,
))
variant = dsm._get_export_variant(args.export_variant)
dsm.export(args.fname_out, '*.csv', variant, MockWorker())
def test_model_estimation(tmpdir, name, alts, subj_count):
indices: Tuple[Optional[int], ...]
if name in ('status-quo', ):
indices = (0, 1, 2, 3)
else:
indices = (0, 1, None, 2)
rows = load_raw_csv('docs/src/_static/examples/%s.csv' % name)
ds = ExperimentalData.from_csv('aug', rows[1:], indices)
models = [
preorder(strict=True, total=True),
preorder(strict=False, total=True),
unattractive(strict=True, total=True),
unattractive(strict=False, total=True),
preorder(strict=True, total=False),
preorder(strict=False, total=False),
UndominatedChoice(strict=True),
UndominatedChoice(strict=False),
PartiallyDominantChoice(fc=True),
PartiallyDominantChoice(fc=False),
Overload(PreorderParams(strict=True, total=True)),
Overload(PreorderParams(strict=False, total=True)),
]
if all(cr.default is not None
for subj in map(SubjectC.decode_from_memory, ds.subjects)
for cr in subj.choices):
models.append(StatusQuoUndominatedChoice())
dsm = ds.analysis_estimation(
MockWorker(),
EstimationOpts(models,
disable_parallelism=False,
disregard_deferrals=False))
check_export(tmpdir, dsm, 'compact (human-friendly)',
'gui/test/expected/%s-models-compact.csv' % name)
check_export(tmpdir, dsm, 'detailed (machine-friendly)',
'gui/test/expected/%s-models-detailed.csv' % name)
def consistency(args):
rows = load_raw_csv(args.fname_in)
ds = ExperimentalData.from_csv('dataset', rows[1:], (0, 1, None, 2))
dsm = ds.analysis_consistency(ProgressWorker(), None)
variant = dsm._get_export_variant(args.export_variant)
dsm.export(args.fname_out, '*.csv', variant, MockWorker())
