use frequency validation from utilsforecast (#717)

dev/environment.yml

@@ -23,4 +23,4 @@ dependencies:
     - polars
     - supersmoother
     - tqdm
-    - utilsforecast>=0.0.17
+    - utilsforecast>=0.0.19

dev/local_environment.yml

@@ -21,4 +21,4 @@ dependencies:
     - polars
     - supersmoother
     - tqdm
-    - utilsforecast>=0.0.17
+    - utilsforecast>=0.0.19

nbs/src/core/core.ipynb

@@ -101,7 +101,7 @@
     "from triad import conditional_dispatcher\n",
     "from utilsforecast.compat import DataFrame, pl_DataFrame, pl_Series\n",
     "from utilsforecast.grouped_array import GroupedArray as BaseGroupedArray\n",
-    "from utilsforecast.validation import ensure_time_dtype\n",
+    "from utilsforecast.validation import ensure_time_dtype, validate_freq\n",
     "\n",
     "import statsforecast.config as sf_config\n",
     "from statsforecast.utils import ConformalIntervals"
@@ -945,30 +945,8 @@
     "            _warn_df_constructor()\n",
     "            return\n",
     "        df = ensure_time_dtype(df, 'ds')\n",
-    "        time_dtype = df['ds'].head(1).to_numpy().dtype\n",
-    "        time_is_int = np.issubdtype(time_dtype, np.integer)\n",
-    "        if time_is_int and not isinstance(self.freq, int):\n",
-    "            raise ValueError(\n",
-    "                'Time column contains integer but the specified frequency is not an integer. '\n",
-    "                'Please provide a valid integer, like `freq=1`'\n",
-    "            )\n",
-    "        elif not time_is_int and not isinstance(self.freq, str):\n",
-    "            # the ensure_time_dtype function makes sure that ds is either int or timestamp\n",
-    "            raise ValueError(\n",
-    "                'Time column contains timestamps but the specified frequency is an integer. '\n",
-    "                'Please provide a valid pandas or polars offset.'\n",
-    "            )\n",
-    "        # try to catch pandas frequency in polars dataframe\n",
-    "        if isinstance(df, pl_DataFrame) and isinstance(self.freq, str):\n",
-    "            missing_n = re.search(r'\\d+', self.freq) is None\n",
-    "            uppercase = re.sub('\\d+', '', self.freq).isupper()\n",
-    "            if missing_n or uppercase:\n",
-    "                raise ValueError(\n",
-    "                    'You must specify a valid polars offset when using polars dataframes. '\n",
-    "                    'You can find the available offsets in '\n",
-    "                    'https://pola-rs.github.io/polars/py-polars/html/reference/expressions/api/polars.Expr.dt.offset_by.html'\n",
-    "                )\n",
-    "        elif isinstance(df, pd.DataFrame) and df.index.name == 'unique_id':\n",
+    "        validate_freq(df['ds'], self.freq)\n",
+    "        if isinstance(df, pd.DataFrame) and df.index.name == 'unique_id':\n",
     "            warnings.warn(\n",
     "                \"Passing unique_id as the index is deprecated. \"\n",
     "                \"Please provide it as a column instead.\",\n",

settings.ini

@@ -15,7 +15,7 @@ language = English
 custom_sidebar = True
 license = apache2
 status = 2
-requirements = numba>=0.55.0 numpy>=1.21.6 pandas>=1.3.5 scipy>=1.7.3 statsmodels>=0.13.2 tqdm fugue>=0.8.1 utilsforecast>=0.0.17
+requirements = numba>=0.55.0 numpy>=1.21.6 pandas>=1.3.5 scipy>=1.7.3 statsmodels>=0.13.2 tqdm fugue>=0.8.1 utilsforecast>=0.0.19
 polars_requirements = polars
 ray_requirements = fugue[ray]>=0.8.1 protobuf>=3.15.3,<4.0.0
 dask_requirements = fugue[dask]>=0.8.1

statsforecast/core.py

@@ -27,7 +27,7 @@
 from triad import conditional_dispatcher
 from utilsforecast.compat import DataFrame, pl_DataFrame, pl_Series
 from utilsforecast.grouped_array import GroupedArray as BaseGroupedArray
-from utilsforecast.validation import ensure_time_dtype
+from utilsforecast.validation import ensure_time_dtype, validate_freq
 
 import statsforecast.config as sf_config
 from .utils import ConformalIntervals
@@ -516,30 +516,8 @@ def _prepare_fit(
             _warn_df_constructor()
             return
         df = ensure_time_dtype(df, "ds")
-        time_dtype = df["ds"].head(1).to_numpy().dtype
-        time_is_int = np.issubdtype(time_dtype, np.integer)
-        if time_is_int and not isinstance(self.freq, int):
-            raise ValueError(
-                "Time column contains integer but the specified frequency is not an integer. "
-                "Please provide a valid integer, like `freq=1`"
-            )
-        elif not time_is_int and not isinstance(self.freq, str):
-            # the ensure_time_dtype function makes sure that ds is either int or timestamp
-            raise ValueError(
-                "Time column contains timestamps but the specified frequency is an integer. "
-                "Please provide a valid pandas or polars offset."
-            )
-        # try to catch pandas frequency in polars dataframe
-        if isinstance(df, pl_DataFrame) and isinstance(self.freq, str):
-            missing_n = re.search(r"\d+", self.freq) is None
-            uppercase = re.sub("\d+", "", self.freq).isupper()
-            if missing_n or uppercase:
-                raise ValueError(
-                    "You must specify a valid polars offset when using polars dataframes. "
-                    "You can find the available offsets in "
-                    "https://pola-rs.github.io/polars/py-polars/html/reference/expressions/api/polars.Expr.dt.offset_by.html"
-                )
-        elif isinstance(df, pd.DataFrame) and df.index.name == "unique_id":
+        validate_freq(df["ds"], self.freq)
+        if isinstance(df, pd.DataFrame) and df.index.name == "unique_id":
             warnings.warn(
                 "Passing unique_id as the index is deprecated. "
                 "Please provide it as a column instead.",