feature: samples to/from csv

climate_health/datatypes.py

@@ -260,6 +260,20 @@ class SummaryStatistics(TimeSeriesData):
 class Samples(TimeSeriesData):
     samples: float
 
+    def topandas(self):
+        n_samples = self.samples.shape[-1]
+        df = pd.DataFrame({'time_period': self.time_period.topandas()} | {f'sample_{i}': self.samples[:, i] for i in
+                                                                          range(n_samples)})
+        return df
+
+    @classmethod
+    def from_pandas(cls, data: pd.DataFrame, fill_missing=False) -> 'TimeSeriesData':
+        ptime = PeriodRange.from_strings(data.time_period.astype(str), fill_missing=fill_missing)
+        n_samples = sum(1 for col in data.columns if col.startswith('sample_'))
+        samples = np.array([data[f'sample_{i}'].values for i in range(n_samples)]).T
+        return cls(ptime, samples)
+
+    to_pandas = topandas
 
 @dataclasses.dataclass
 class Quantile:

tests/test_datatypes.py

@@ -3,7 +3,7 @@
 import bionumpy as bnp
 import pytest
 from bionumpy.util.testing import assert_bnpdataclass_equal
-from climate_health.datatypes import ClimateHealthTimeSeries, HealthData
+from climate_health.datatypes import ClimateHealthTimeSeries, HealthData, Samples
 from climate_health.spatio_temporal_data.temporal_dataclass import DataSet
 from climate_health.time_period import PeriodRange
 from climate_health.time_period.dataclasses import Year
@@ -59,3 +59,17 @@ def test_dataset_with_missing(dataset_with_missing):
     for location, data in health_data.items():
         # assert data.start_timestamp == start
         assert data.end_timestamp == end
+
+@pytest.fixture()
+def samples():
+    time_period = PeriodRange.from_strings(['2010', '2011', '2012'])
+    return Samples(
+        time_period=time_period,
+        samples=np.random.rand(3, 10))
+
+
+def test_samples(samples, tmp_path):
+    path = tmp_path/'samples.csv'
+    samples.to_csv(path)
+    samples2 = Samples.from_csv(path)
+    assert_bnpdataclass_equal(samples, samples2)