move columns of dependencies into config

odc/stats/plugins/lc_level34.py

@@ -2,20 +2,22 @@
 Plugin of Module A3 in LandCover PipeLine
 """
 
-from typing import Optional, Dict
+from typing import Optional, Dict, List
 
 import xarray as xr
 import s3fs
 import os
 import pandas as pd
 import dask.array as da
+import logging
 
 from ._registry import StatsPluginInterface, register
 from ._utils import rasterize_vector_mask, generate_numexpr_expressions
 from odc.stats._algebra import expr_eval
 from osgeo import gdal
 
 NODATA = 255
+_log = logging.getLogger(__name__)
 
 
 class StatsLccsLevel4(StatsPluginInterface):
@@ -27,6 +29,7 @@ class StatsLccsLevel4(StatsPluginInterface):
     def __init__(
         self,
         class_def_path: str = None,
+        class_condition: Dict[str, List] = None,
         urban_mask: str = None,
         filter_expression: str = None,
         mask_threshold: Optional[float] = None,
@@ -43,6 +46,9 @@ def __init__(
         elif not os.path.exists(class_def_path):
             raise FileNotFoundError(f"{class_def_path} not found")
 
+        if class_condition is None:
+            raise ValueError("Missing input to generate classification conditions")
+
         if urban_mask is None:
             raise ValueError("Missing urban mask shapefile")
 
@@ -54,8 +60,12 @@ def __init__(
             raise ValueError("Missing urban mask filter")
 
         self.class_def = pd.read_csv(class_def_path)
-        cols = list(self.class_def.columns[:6]) + list(self.class_def.columns[9:-6])
-        self.class_def = self.class_def[cols].astype(str).fillna("nan")
+        self.class_condition = class_condition
+        cols = set()
+        for k, v in self.class_condition.items():
+            cols |= {k} | set(v)
+
+        self.class_def = self.class_def[list(cols)].astype(str).fillna("nan")
 
         self.urban_mask = urban_mask
         self.filter_expression = filter_expression
@@ -77,6 +87,7 @@ def classification(self, xx, class_def, con_cols, class_col):
         res = da.full(xx.level_3_4.shape, 0, dtype="uint8")
 
         for expression in expressions:
+            _log.info(expression)
             local_dict.update({"res": res})
             res = expr_eval(
                 expression,
@@ -98,9 +109,10 @@ def classification(self, xx, class_def, con_cols, class_col):
         return res
 
     def reduce(self, xx: xr.Dataset) -> xr.Dataset:
-        con_cols = ["level1", "artificial_surface", "cultivated"]
         class_col = "level3"
-        level3 = self.classification(xx, self.class_def, con_cols, class_col)
+        level3 = self.classification(
+            xx, self.class_def, self.class_condition[class_col], class_col
+        )
 
         # apply urban mask
         # 215 -> 216 if urban_mask == 0
@@ -119,6 +131,8 @@ def reduce(self, xx: xr.Dataset) -> xr.Dataset:
             dtype="uint8",
         )
 
+        # append level3 to the input dataset so it can be used
+        # to classify level4
         attrs = xx.attrs.copy()
         attrs["nodata"] = NODATA
         dims = xx.level_3_4.dims[1:]
@@ -127,18 +141,10 @@ def reduce(self, xx: xr.Dataset) -> xr.Dataset:
             level3.squeeze(), dims=dims, attrs=attrs, coords=coords
         )
 
-        con_cols = [
-            "level1",
-            "level3",
-            "woody",
-            "water_season",
-            "water_frequency",
-            "pv_pc_50",
-            "bs_pc_50",
-        ]
         class_col = "level4"
-
-        level4 = self.classification(xx, self.class_def, con_cols, class_col)
+        level4 = self.classification(
+            xx, self.class_def, self.class_condition[class_col], class_col
+        )
 
         data_vars = {
             k: xr.DataArray(v, dims=dims, attrs=attrs)