Merge pull request #71 from GeoOcean/67-swan-generation-bathy

[JTH] merge fixed parameters structre implementation in swan wrapper and more

Author: tausiaj

bluemath_tk/core/decorators.py

@@ -120,6 +120,7 @@ def wrapper(
         directional_variables: List[str] = [],
         custom_scale_factor: dict = {},
         min_number_of_points: int = None,
+        max_number_of_iterations: int = 10,
         normalize_data: bool = True,
     ):
         if data is None:
@@ -133,6 +134,11 @@ def wrapper(
         if min_number_of_points is not None:
             if not isinstance(min_number_of_points, int) or min_number_of_points <= 0:
                 raise ValueError("Minimum number of points must be integer and > 0")
+        if (
+            not isinstance(max_number_of_iterations, int)
+            or max_number_of_iterations <= 0
+        ):
+            raise ValueError("Maximum number of iterations must be integer and > 0")
         if not isinstance(normalize_data, bool):
             raise TypeError("Normalize data must be a boolean")
         return func(

bluemath_tk/datamining/kma.py

@@ -182,6 +182,7 @@ def fit(
         directional_variables: List[str] = [],
         custom_scale_factor: dict = {},
         min_number_of_points: int = None,
+        max_number_of_iterations: int = 10,
         normalize_data: bool = True,
     ) -> None:
         """
@@ -206,6 +207,10 @@ def fit(
         min_number_of_points : int, optional
             The minimum number of points to consider a cluster.
             Default is None.
+        max_number_of_iterations : int, optional
+            The maximum number of iterations for the K-Means algorithm.
+            This is used when min_number_of_points is not None.
+            Default is 10.
         normalize_data : bool, optional
             A flag to normalize the data. Default is True.
         """
@@ -248,9 +253,10 @@ def fit(
                 if np.all(counts >= min_number_of_points):
                     stable_kma_child = True
                 number_of_tries += 1
-                if number_of_tries > 10:
+                if number_of_tries > max_number_of_iterations:
                     raise ValueError(
-                        "Failed to find a stable K-Means configuration after 10 attempts."
+                        f"Failed to find a stable K-Means configuration after {max_number_of_iterations} attempts."
+                        "Change max_number_of_iterations or min_number_of_points."
                     )
             self.logger.info(
                 f"Found a stable K-Means configuration after {number_of_tries} attempts."
@@ -318,6 +324,7 @@ def fit_predict(
         directional_variables: List[str] = [],
         custom_scale_factor: dict = {},
         min_number_of_points: int = None,
+        max_number_of_iterations: int = 10,
         normalize_data: bool = True,
     ) -> Tuple[pd.DataFrame, pd.DataFrame]:
         """
@@ -337,6 +344,10 @@ def fit_predict(
         min_number_of_points : int, optional
             The minimum number of points to consider a cluster.
             Default is None.
+        max_number_of_iterations : int, optional
+            The maximum number of iterations for the K-Means algorithm.
+            This is used when min_number_of_points is not None.
+            Default is 10.
         normalize_data : bool, optional
             A flag to normalize the data. Default is True.
 
@@ -352,6 +363,7 @@ def fit_predict(
             directional_variables=directional_variables,
             custom_scale_factor=custom_scale_factor,
             min_number_of_points=min_number_of_points,
+            max_number_of_iterations=max_number_of_iterations,
             normalize_data=normalize_data,
         )
 

bluemath_tk/predictor/xwt.py

@@ -286,6 +286,8 @@ def fit(
 
         kma: KMA = self.steps.get("kma")
         self.num_clusters = kma.num_clusters
+        # TODO: standarize PCs by first PC variance
+        # pca.pcs_df / pca.pcs.stds.isel(n_component=0).values
         kma_bmus, _kma_bmus_df = kma.fit_predict(
             data=pca.pcs_df,
             **fit_params.get("kma", {}),

bluemath_tk/topo_bathy/swan_grid.py

@@ -23,17 +23,26 @@ def generate_grid_parameters(bathy_data: xr.DataArray) -> dict:
     """
 
     return {
-        "xpc": np.nanmin(bathy_data.lon),  # x origin
-        "ypc": np.nanmin(bathy_data.lat),  # y origin
+        "xpc": int(np.nanmin(bathy_data.lon)),  # x origin
+        "ypc": int(np.nanmin(bathy_data.lat)),  # y origin
         "alpc": 0,  # x-axis direction
-        "xlenc": np.nanmax(bathy_data.lon)
-        - np.nanmin(bathy_data.lon),  # grid length in x
-        "ylenc": np.nanmax(bathy_data.lat)
-        - np.nanmin(bathy_data.lat),  # grid length in y
+        "xlenc": int(
+            np.nanmax(bathy_data.lon) - np.nanmin(bathy_data.lon)
+        ),  # grid length in x
+        "ylenc": int(
+            np.nanmax(bathy_data.lat) - np.nanmin(bathy_data.lat)
+        ),  # grid length in y
         "mxc": len(bathy_data.lon) - 1,  # number mesh x, una menos pq si no SWAN peta
         "myc": len(bathy_data.lat) - 1,  # number mesh y, una menos pq si no SWAN peta
-        "dxinp": bathy_data.lon[1].values
-        - bathy_data.lon[0].values,  # size mesh x (resolution in x)
-        "dyinp": bathy_data.lat[1].values
-        - bathy_data.lat[0].values,  # size mesh y (resolution in y)
+        "xpinp": np.nanmin(bathy_data.lon),  # x origin
+        "ypinp": np.nanmin(bathy_data.lat),  # y origin
+        "alpinp": 0,  # x-axis direction
+        "mxinp": len(bathy_data.lon) - 1,  # number mesh x
+        "myinp": len(bathy_data.lat) - 1,  # number mesh y
+        "dxinp": abs(
+            bathy_data.lon[1].values - bathy_data.lon[0].values
+        ),  # size mesh x (resolution in x)
+        "dyinp": abs(
+            bathy_data.lat[1].values - bathy_data.lat[0].values
+        ),  # size mesh y (resolution in y)
     }

bluemath_tk/waves/binwaves.py

@@ -110,12 +110,11 @@ def process_kp_coefficients(
             print(f"Error processing {input_spec_file} and {output_spec_file}")
             print(e)
 
-    return (
-        xr.concat(output_kp_list, dim="case_num")
-        .fillna(0.0)
-        .sortby("freq")
-        .sortby("dir")
-    )
+    # Concat files one by one
+    concatened_kp = output_kp_list[0]
+    for file in output_kp_list[1:]:
+        concatened_kp = xr.concat([concatened_kp, file], dim="case_num")
+    return concatened_kp.fillna(0.0).sortby("freq").sortby("dir")
 
 
 def reconstruc_spectra(
@@ -149,15 +148,15 @@ def reconstruc_spectra(
 
     # Setup Dask client
     if num_workers is None:
-        num_workers = os.environ.get("BLUEMATH_NUM_WORKERS", 2)
+        num_workers = os.environ.get("BLUEMATH_NUM_WORKERS", 4)
     client = setup_dask_client(n_workers=num_workers, memory_limit=memory_limit)
 
     try:
         # Process with controlled chunks
         offshore_spectra_chunked = offshore_spectra.chunk(
-            {"time": chunk_sizes.get("time", 24)}
+            {"time": chunk_sizes.get("time", 24 * 7)}
         )
-        kp_coeffs_chunked = kp_coeffs.chunk({"site": 1})
+        kp_coeffs_chunked = kp_coeffs.chunk({"site": 10})
         with ProgressBar():
             onshore_spectra = (
                 (offshore_spectra_chunked * kp_coeffs_chunked).sum(dim="case_num")
@@ -250,7 +249,7 @@ def plot_selected_cases_grid(
     ax = fig.add_subplot(1, 1, 1, projection="polar")
 
     # prepare data
-    x = np.append(np.deg2rad(directions - 7.5), np.deg2rad(directions - 7.5)[0])
+    x = np.append(np.deg2rad(directions), np.deg2rad(directions)[0])
     y = np.append(0, frequencies)
     z = (
         np.array(range(len(frequencies) * len(directions)))