fix(waterdata): materialize numpy/Series numeric params instead of str()-ing them (#308)

A numeric (_NO_NORMALIZE_PARAMS) param — water_year, year, month, day,
thresholds, … — passed as a numpy array or pandas Series fell into the
`args[k] = v` passthrough in _get_args without being materialized to a list.
Downstream, the GET comma-join and the chunker both test `list`/`tuple`, so an
ndarray/Series was neither comma-joined nor chunked: e.g.
get_peaks(water_year=np.array([2020, 2021])) produced
`water_year=%5B2020+2021%5D` (the array's repr) instead of
`water_year=2020,2021`, which the API rejects with HTTP 400. Plain lists
already worked.

Split the branch so _NO_NORMALIZE_PARAMS values keep their element types (no
string-normalization) but a non-string iterable is still materialized to a
list of native Python scalars — `.tolist()` for numpy/pandas, `list()` for
generators and other iterables — so the values comma-join in the URL, chunk,
and stay JSON-serializable (no numpy reprs in args).

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: thodson-usgs

dataretrieval/waterdata/utils.py

@@ -2024,9 +2024,18 @@ def _get_args(
             args[k] = _as_str_list(v, k)
         elif (
             k in _NO_NORMALIZE_PARAMS
-            or isinstance(v, str)
-            or not isinstance(v, Iterable)
+            and isinstance(v, Iterable)
+            and not isinstance(v, str)
         ):
+            # Numeric params (water_year, bbox, thresholds, …) keep their
+            # element types — no string-normalization — but a non-string
+            # iterable (numpy array, pandas Series, generator) is materialized
+            # to a list so the GET comma-join and the chunker, which test
+            # ``list``/``tuple``, handle it instead of str()-ing the whole
+            # array. ``.tolist()`` yields native int/float; ``list()`` covers
+            # generators and other iterables. Scalars/strings fall through.
+            args[k] = v.tolist() if hasattr(v, "tolist") else list(v)
+        elif isinstance(v, str) or not isinstance(v, Iterable):
             args[k] = v
         else:
             args[k] = _normalize_str_iterable(v, k)

tests/waterdata_test.py

@@ -35,6 +35,7 @@
     _check_profiles,
     _construct_api_requests,
     _construct_cql_request,
+    _get_args,
     _normalize_str_iterable,
 )
 
@@ -294,6 +295,27 @@ def test_construct_api_requests_numeric_list_joins_with_str():
     assert "water_year=2020%2C2021" in str(req.url)
 
 
+def test_get_args_materializes_numpy_and_series_numeric_params():
+    """Regression: numeric (_NO_NORMALIZE_PARAMS) params given as a numpy array
+    or pandas Series must be materialized to a list of native Python scalars so
+    they comma-join in the URL (and stay JSON-serializable) — previously the
+    array/Series repr leaked into the query string."""
+    for value in (np.array([2020, 2021]), pd.Series([2020, 2021])):
+        args = _get_args({"water_year": value})
+        assert args["water_year"] == [2020, 2021]
+        # native Python ints, not numpy scalars (JSON-serializable, no np reprs)
+        assert [type(x) for x in args["water_year"]] == [int, int]
+        req = _construct_api_requests("peaks", **args)
+        assert "water_year=2020%2C2021" in str(req.url)
+
+    # float coordinate arrays (e.g. bbox) likewise materialize to native floats
+    args = _get_args({"bbox": np.array([-92.8, 44.2, -88.9, 46.0])})
+    assert args["bbox"] == [-92.8, 44.2, -88.9, 46.0]
+    assert all(type(x) is float for x in args["bbox"])
+    req = _construct_api_requests("daily", **args)
+    assert "bbox=-92.8%2C44.2%2C-88.9%2C46.0" in str(req.url)
+
+
 def test_construct_api_requests_two_element_date_list_becomes_interval():
     """A two-element date list is interpreted as start/end of an OGC datetime
     interval (joined with '/'), NOT as two discrete dates. The OGC `datetime`