wingfoil-python 8.0.0

python bindings for wingfoil - graph based stream processing framework
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270

"""Tests for the augurs Python bindings (stream.augurs_forecast / augurs_outlier).

augurs is a pure-Rust compute library with no external service, so these tests
run by default — there is no `requires_*` marker to gate. They exercise the
pyo3 marshaling glue end to end against deterministic historical-mode runs.
"""

import unittest

from wingfoil import ticker


def _ramp():
    """A rising stream of floats: n + sin(n/2)."""
    import math

    return ticker(1.0).count().map(lambda n: float(n) + math.sin(n * 0.5))


def _series():
    """Four readings per tick; series 3 diverges after tick 15."""
    import math

    def reading(n):
        base = 100.0 + math.sin(n * 0.4)
        diverging = base + 80.0 if n > 15 else base + 0.2
        return [base, base + 0.1, base - 0.1, diverging]

    return ticker(1.0).count().map(reading)


def _sine(period=12):
    """A clean sine wave with the given period."""
    import math

    return ticker(1.0).count().map(lambda n: math.sin(n * math.tau / period))


class TestForecast(unittest.TestCase):
    def test_forecast_shape(self):
        forecast = _ramp().augurs_forecast(48, 3)
        captured = forecast.collect()
        captured.run(realtime=False, cycles=30)
        last = captured.peek_value()[-1]
        self.assertIsInstance(last, dict)
        self.assertEqual(len(last["point"]), 3)

    def test_forecast_with_intervals(self):
        forecast = _ramp().augurs_forecast(48, 2, level=0.9)
        captured = forecast.collect()
        captured.run(realtime=False, cycles=30)
        last = captured.peek_value()[-1]
        self.assertEqual(len(last["lower"]), 2)
        self.assertEqual(len(last["upper"]), 2)
        for i in range(2):
            self.assertLessEqual(last["lower"][i], last["point"][i])
            self.assertGreaterEqual(last["upper"][i], last["point"][i])

    def test_forecast_waits_for_min_points(self):
        forecast = _ramp().augurs_forecast(64, 1, min_points=20)
        captured = forecast.collect()
        captured.run(realtime=False, cycles=10)
        # 10 cycles < 20 min_points → never ticked, so the collected stream
        # has no value (peeks as None rather than an empty list).
        self.assertFalse(captured.peek_value())


class TestOutlier(unittest.TestCase):
    def test_flags_diverging_series(self):
        outliers = _series().augurs_outlier(30, 0.5)
        captured = outliers.collect()
        captured.run(realtime=False, cycles=30)
        last = captured.peek_value()[-1]
        self.assertIn(3, last["outlying"])
        self.assertEqual(len(last["scores"]), 4)

    def test_aligned_series_quiet(self):
        aligned = ticker(1.0).count().map(lambda n: [50.0, 50.05, 49.95])
        outliers = aligned.augurs_outlier(20, 0.5)
        captured = outliers.collect()
        captured.run(realtime=False, cycles=20)
        self.assertEqual(captured.peek_value()[-1]["outlying"], [])


class TestForecastMstl(unittest.TestCase):
    def test_mstl_forecast_swings(self):
        import math

        seasonal = ticker(1.0).count().map(
            lambda n: 0.1 * n + 5.0 * math.sin(n * math.tau / 12.0)
        )
        forecast = seasonal.augurs_forecast(120, 12, periods=[12])
        captured = forecast.collect()
        captured.run(realtime=False, cycles=80)
        point = captured.peek_value()[-1]["point"]
        self.assertEqual(len(point), 12)
        self.assertGreater(max(point) - min(point), 2.0)


class TestOutlierDbscan(unittest.TestCase):
    def test_dbscan_flags_diverging(self):
        outliers = _series().augurs_outlier(40, 0.5, detector="dbscan")
        captured = outliers.collect()
        captured.run(realtime=False, cycles=30)
        self.assertIn(3, captured.peek_value()[-1]["outlying"])


class TestChangepoint(unittest.TestCase):
    def test_detects_level_shift(self):
        series = ticker(1.0).count().map(lambda n: 50.0 if n > 20 else 0.0)
        changes = series.augurs_changepoint(60)
        captured = changes.collect()
        captured.run(realtime=False, cycles=50)
        indices = captured.peek_value()[-1]["indices"]
        self.assertTrue(indices, "expected a changepoint after the shift")
        self.assertTrue(any(i >= 10 for i in indices))

    def test_quiet_when_steady(self):
        import math

        series = ticker(1.0).count().map(lambda n: 10.0 + 0.1 * math.sin(n * 0.5))
        changes = series.augurs_changepoint(40)
        captured = changes.collect()
        captured.run(realtime=False, cycles=40)
        self.assertEqual(captured.peek_value()[-1]["indices"], [])

    def test_hazard_and_min_points_marshal(self):
        # Exercise the hazard/min_points optionals end to end: a level shift
        # is still detected with a non-default hazard and min_points.
        series = ticker(1.0).count().map(lambda n: 50.0 if n > 20 else 0.0)
        changes = series.augurs_changepoint(60, min_points=12, hazard=100.0)
        captured = changes.collect()
        captured.run(realtime=False, cycles=50)
        self.assertTrue(captured.peek_value()[-1]["indices"])


class TestSeasons(unittest.TestCase):
    def test_detects_period(self):
        seasons = _sine(12).augurs_seasons(96)
        captured = seasons.collect()
        captured.run(realtime=False, cycles=96)
        periods = captured.peek_value()[-1]["periods"]
        self.assertTrue(any(10 <= p <= 14 for p in periods), periods)

    def test_period_range_and_min_points_marshal(self):
        # Exercise the min_period/max_period/min_points optionals: a bracket
        # around 12 must still surface the period, and min_points must gate.
        seasons = _sine(12).augurs_seasons(
            96, min_points=40, min_period=8, max_period=16
        )
        captured = seasons.collect()
        captured.run(realtime=False, cycles=96)
        periods = captured.peek_value()[-1]["periods"]
        self.assertTrue(any(8 <= p <= 16 for p in periods), periods)

    def test_min_points_gates(self):
        seasons = _sine(12).augurs_seasons(96, min_points=50)
        captured = seasons.collect()
        captured.run(realtime=False, cycles=20)
        # 20 cycles < 50 min_points → never ticked.
        self.assertFalse(captured.peek_value())


class TestDtw(unittest.TestCase):
    def test_distance_matrix_shape_and_order(self):
        import math

        readings = ticker(1.0).count().map(
            lambda n: [
                math.sin(n * 0.3),
                math.sin(n * 0.3) + 0.02,
                5.0 * math.sin(n * 0.3) + 10.0,
            ]
        )
        dists = readings.augurs_dtw(30)
        captured = dists.collect()
        captured.run(realtime=False, cycles=30)
        rows = captured.peek_value()[-1]["rows"]
        self.assertEqual(len(rows), 3)
        self.assertEqual(len(rows[0]), 3)
        self.assertLess(rows[0][0], 1e-9)
        self.assertGreater(rows[0][2], rows[0][1])


class TestCluster(unittest.TestCase):
    def test_groups_series(self):
        import math

        def reading(n):
            low = math.sin(n * 0.3)
            high = math.sin(n * 0.3) + 20.0
            return [low, low + 0.02, high, high + 0.02, 50.0 * math.cos(n * 0.9) + 100.0]

        clusters = ticker(1.0).count().map(reading).augurs_cluster(30, 1.0, 2)
        captured = clusters.collect()
        captured.run(realtime=False, cycles=30)
        labels = captured.peek_value()[-1]["labels"]
        self.assertEqual(len(labels), 5)
        self.assertEqual(labels[0], labels[1])
        self.assertEqual(labels[2], labels[3])
        self.assertNotEqual(labels[0], labels[2])
        self.assertEqual(labels[4], -1)


class TestConstruction(unittest.TestCase):
    """Construction-only: exercise argument parsing without depending on output."""

    def test_forecast_constructs(self):
        stream = _ramp().augurs_forecast(32, 2)
        self.assertIsNotNone(stream)

    def test_forecast_mstl_constructs(self):
        stream = _ramp().augurs_forecast(48, 4, periods=[12])
        self.assertIsNotNone(stream)

    def test_outlier_constructs(self):
        stream = _series().augurs_outlier(16, 0.25)
        self.assertIsNotNone(stream)

    def test_new_ops_construct(self):
        self.assertIsNotNone(_ramp().augurs_changepoint(32))
        self.assertIsNotNone(_ramp().augurs_seasons(48))
        self.assertIsNotNone(_series().augurs_dtw(16))
        self.assertIsNotNone(_series().augurs_cluster(16, 1.0, 2))


class TestValidation(unittest.TestCase):
    """Invalid arguments are rejected loudly rather than silently coerced."""

    def test_unknown_detector_raises(self):
        with self.assertRaises(ValueError):
            _series().augurs_outlier(30, 0.5, detector="dbscann")

    def test_unknown_metric_raises(self):
        with self.assertRaises(ValueError):
            _series().augurs_dtw(30, metric="manhatten")

    def test_unknown_cluster_metric_raises(self):
        with self.assertRaises(ValueError):
            _series().augurs_cluster(30, 1.0, 2, metric="chebyshev")

    def test_out_of_range_sensitivity_raises(self):
        # A sensitivity outside (0, 1) raises ValueError at construction rather
        # than panicking across the FFI boundary.
        for bad in (0.0, 1.0, 1.5, -0.1):
            with self.assertRaises(ValueError):
                _series().augurs_outlier(30, bad)

    def test_out_of_range_level_raises(self):
        for bad in (0.0, 1.0, 1.5):
            with self.assertRaises(ValueError):
                _ramp().augurs_forecast(48, 2, level=bad)

    def test_invalid_mstl_period_raises(self):
        # STL cannot decompose a period below 2; caught at construction.
        for bad in ([1], [0], [12, 1]):
            with self.assertRaises(ValueError):
                _ramp().augurs_forecast(64, 4, periods=bad)

    def test_non_float_input_fails_fast(self):
        # A non-float value is a hard error (like .average()/.sum()), not a
        # silently substituted NaN that would poison the model window.
        bad = ticker(1.0).count().map(lambda n: "not a float")
        captured = bad.augurs_forecast(16, 1).collect()
        with self.assertRaises(Exception):
            captured.run(realtime=False, cycles=20)


if __name__ == "__main__":
    unittest.main()