Merge pull request #99 from emiliom/transect-select-bug

Fix transect subset selection bug and increment version for new release

Author: emiliom

EchoPro/__init__.py

@@ -2,5 +2,5 @@
 
 __all__ = ["Survey"]
 
-__version__ = "0.3.0"
+__version__ = "0.3.1"
 

EchoPro/computation/transect_results.py

@@ -1233,34 +1233,48 @@ def set_adult_NASC(self) -> None:
             columns=["val"], index=len_age_dist_all_norm.stratum_num, dtype=np.float64
         )
 
-        for i in len_age_dist_all_norm.stratum_num.values:
+        # for stratum in self.all_strata:
+        for stratum in len_age_dist_all_norm.stratum_num.values:
             sig_bs_aged_ave = np.sum(
                 self.survey.params["sig_b_coef"]
                 * np.matmul(
                     (self.survey.params["bio_hake_len_bin"] ** 2),
-                    len_age_dist_all_norm.sel(stratum_num=i).values,
+                    len_age_dist_all_norm.sel(stratum_num=stratum).values,
                 )
             )
 
             temp = self.survey.params["sig_b_coef"] * np.matmul(
                 (self.survey.params["bio_hake_len_bin"] ** 2),
-                len_age_dist_all_norm.sel(stratum_num=i).isel(age_bin=0).values,
+                len_age_dist_all_norm.sel(stratum_num=stratum).isel(age_bin=0).values,
             )
 
             age1_nasc_proportion = temp / sig_bs_aged_ave
 
-            nasc_fraction_adult_df.loc[i] = abs(1.0 - age1_nasc_proportion)
+            nasc_fraction_adult_df.loc[stratum] = abs(1.0 - age1_nasc_proportion)
+
+        # Identify and populate strata missing in nasc_fraction_adult_df
+        # (and therefore in self.bin_ds.len_age_dist_all)
+        # based on surrounding strata as pre-allocated in stratum_choices
+        # TODO: This scheme actually reflects the simpler implementation from v0.1.0-alpha.
+        #   Revisit the Matlab code and related missing-stratum handling in this module
+        #   https://github.com/uw-echospace/EchoPro_matlab/blob/26b939c9c7c0ccbf8828d402e249e1fdf6ed2b5a/general/load_files_parameters/get_historical_strata_data.m#L731-L795 # noqa
+        missing_stratum_num = (
+            set(self.nasc_df.stratum_num.values) - set(len_age_dist_all_norm.stratum_num.values)
+        )
+        for stratum in missing_stratum_num:
+            nasc_fraction_adult_df.loc[stratum] = (
+                nasc_fraction_adult_df.loc[self.stratum_choices[stratum]]['val'].mean()
+            )
+        # Storing nasc_fraction_adult_df is not required but will be very useful
+        # in the interim for ongoing development and testing
+        self.nasc_fraction_adult_df = nasc_fraction_adult_df
 
-        # obtain the adult NASC proportion coefficient for each stratum value
         fraction_adult_stratum_df = nasc_fraction_adult_df.loc[
             self.nasc_df.stratum_num
         ].values.flatten()
 
-        # obtain the NASC for adults
-        NASC_adult = self.nasc_df["NASC"] * fraction_adult_stratum_df
-
-        # assign values to results gdf
-        self.transect_results_gdf["NASC_adult"] = NASC_adult
+        # Calculate NASC for adults and assign values to results gdf
+        self.transect_results_gdf["NASC_adult"] = self.nasc_df["NASC"] * fraction_adult_stratum_df
 
     def _construct_results_gdf(self) -> None:
         """

EchoPro/tests/computation/test_transect_selection.py

@@ -1,10 +1,55 @@
 import pytest
 
+import math
+
 import numpy as np
 import EchoPro
 from EchoPro.computation import SemiVariogram as SV
 
 
+@pytest.mark.parametrize(["removal_percentage"], [
+    [38.0], [50.0], [60.0]
+])
+def test_transects_selection_success(removal_percentage, config_base_path):
+    # initialize Survey object
+    survey_2019 = EchoPro.Survey(
+        init_file_path=config_base_path / 'initialization_config.yml',
+        survey_year_file_path=config_base_path / 'survey_year_2019_config.yml',
+        source=3,
+        exclude_age1=True
+    )
+
+    # load all data
+    survey_2019.load_survey_data()
+
+    # obtain all unique transects in nasc_df
+    unique_transects = survey_2019.nasc_df.index.unique().values
+
+    # determine the number of transects that should be selected
+    num_sel_transects = math.floor(len(unique_transects) * (1.0 - removal_percentage / 100.0))
+
+    # initialize the random number generator object and fix the seed
+    rng = np.random.default_rng(seed=1234)
+
+    num_iterations = 5
+    for iteration in range(num_iterations):
+        # randomly select transects without replacement
+        selected_transects = list(rng.choice(unique_transects, num_sel_transects, replace=False))
+
+        print("Iteration #", iteration, len(selected_transects))
+
+        # compute all transect variables
+        survey_2019.compute_transect_results(selected_transects=selected_transects)
+
+        nasc_fraction_adult_df_set = set(survey_2019.bio_calc.nasc_fraction_adult_df.index)
+        nasc_stratum_num_set = set(survey_2019.bio_calc.nasc_df.stratum_num.unique())
+
+        # This test is intended mainly to verify that survey_2019.compute_transect_results
+        # runs successfully. Not sure what other assertion test would be better
+        assert nasc_stratum_num_set <= nasc_fraction_adult_df_set
+        # assert nasc_stratum_num_set <= nasc_fraction_adult_df_set | {0}
+
+
 def test_transect_selection_output(config_base_path):
     # initialize Survey object
     survey_2019 = EchoPro.Survey(