sc-340182 - fix the report with new versions of Flask, fix the plugin…

CHANGELOG.md

@@ -1,5 +1,9 @@
 # Changelog
 
+## Version 1.3.6 (2026-09-14)
+- Add support for newer versions of Flask (enabling the model error analysis of custom plugin's algorithms)
+- Fix bug now that sparse matrices can be given as a preprocessed dataframe
+
 ## Version 1.3.5 (2026-01-29)
 - Add python 3.12, 3.13 and 3.14 official support
 

Jenkinsfile

@@ -2,7 +2,7 @@ pipeline {
    options { disableConcurrentBuilds() }
    agent { label 'dss-plugin-tests'}
    environment {
-        PLUGIN_INTEGRATION_TEST_INSTANCE="$HOME/instance_config.json"
+        PLUGIN_INTEGRATION_TEST_INSTANCE="/home/jenkins-agent/instance_config.json"
         UNIT_TEST_FILES_STATUS_CODE = sh(script: 'ls ./tests/*/unit/test*', returnStatus: true)
         INTEGRATION_TEST_FILES_STATUS_CODE = sh(script: 'ls ./tests/*/integration/test*', returnStatus: true)
     }

plugin.json

@@ -1,6 +1,6 @@
 {
   "id" : "model-error-analysis",
-  "version" : "1.3.5",
+  "version" : "1.3.6",
   "meta" : {
     "label" : "Model Error Analysis",
     "description" : "Debug model performance with error analysis. A code env is only required to use the Jupyter Notebook.",

python-lib/dku_error_analysis_tree_parsing/tree_parser.py

@@ -293,7 +293,10 @@ def parse_nodes(self, tree, preprocessed_x):
             preprocessed_feature = self.feature_list[feature_idx]
             split_parameters = self._get_split_parameters(preprocessed_feature)
             if split_parameters.feature not in tree.df:
-                tree.df[split_parameters.feature] = split_parameters.add_preprocessed_feature(preprocessed_x, feature_idx)
+                feature_values = split_parameters.add_preprocessed_feature(preprocessed_x, feature_idx)
+                if hasattr(feature_values, "toarray"): # for sparse matrices
+                    feature_values = feature_values.toarray().reshape(-1)
+                tree.df[split_parameters.feature] = feature_values
             value = split_parameters.value
             if value is None:
                 value = split_parameters.value_func(threshold)

resource/py/test_dku_visualizer.py

@@ -43,7 +43,7 @@ def _get_stats(node_id, feature_name, nr_bins, bins=None):
         return _get_stats
     return f
 
-def test_plot_feature_distributions_no_show_global(mocker, mocked_nodes, mocked_get_stats, caplog):
+def test_plot_feature_distributions_no_show_global(mocker, mocked_nodes, mocked_get_stats, caplog, dss_target):
     analyzer = mocker.Mock(spec=DkuErrorAnalyzer)
     mocker.patch.object(analyzer, "_get_ranked_leaf_ids", return_value=[1])
     analyzer.tree.ranked_features = [
@@ -120,7 +120,7 @@ def test_plot_feature_distributions_no_show_global(mocker, mocked_nodes, mocked_
     assert patched_get_node.call_args_list[2][0] == (1,) and patched_get_node.call_count == 3
     patched_get_stats.assert_called_once_with(1, "num", 10)
 
-def test_plot_feature_distributions_show_global(mocker, mocked_nodes, mocked_get_stats, caplog):
+def test_plot_feature_distributions_show_global(mocker, mocked_nodes, mocked_get_stats, caplog, dss_target):
     analyzer = mocker.Mock(spec=DkuErrorAnalyzer)
     mocker.patch.object(analyzer, "_get_ranked_leaf_ids", return_value=[1])
     analyzer.tree.ranked_features = [
@@ -226,6 +226,6 @@ def test_plot_feature_distributions_show_global(mocker, mocked_nodes, mocked_get
     assert patched_get_stats.call_args_list[0][0] == (1, "num", 10)
     assert patched_get_stats.call_args_list[1][0] == (0, "num", 10, [])
 
-def test_failed_init(mocker):
+def test_failed_init(mocker, dss_target):
     with pytest.raises(TypeError, match="You need to input a DkuErrorAnalyzer object."):
         DkuErrorVisualizer(mocker.Mock(spec=ErrorAnalyzer))

resource/py/test_tree_parser.py

@@ -8,6 +8,7 @@
 import numpy as np
 import pytest
 import logging
+from scipy import sparse
 
 @pytest.fixture
 def create_parser(mocker):
@@ -250,6 +251,30 @@ def test_build_tree(mocker, df, create_parser, dss_target):
 
     pd.testing.assert_frame_equal(tree.df, pd.concat([dataframe, pd.Series(["toast"]*12, name="super_cat_1"), pd.Series(["hellow"]*12, name="super_cat_2")], axis=1))
 
+
+@pytest.mark.parsing
+def test_build_tree_with_sparse_preprocessed_features(mocker, create_parser, dss_target):
+    mocker.patch("dku_error_analysis_tree_parsing.tree_parser.descale_numerical_thresholds",
+                 return_value=[.5, -2, -2])
+
+    error_model = mocker.Mock(classes_=np.array([
+        ErrorAnalyzerConstants.WRONG_PREDICTION,
+        ErrorAnalyzerConstants.CORRECT_PREDICTION
+    ]))
+    error_model.tree_.children_left = np.array([1, -2, -2])
+    error_model.tree_.children_right = np.array([2, -2, -2])
+    error_model.tree_.feature = np.array([0, -2, -2])
+    error_model.tree_.value = np.array([[[1, 2]], [[1, 0]], [[0, 2]]])
+
+    tree = mocker.Mock(df=pd.DataFrame(index=range(3)))
+    parser = create_parser(error_model=error_model, feature_names=["feature"])
+
+    parser.parse_nodes(tree, sparse.csr_matrix([[0], [1], [2]]))
+
+    pd.testing.assert_series_equal(
+        tree.df["feature"], pd.Series([0, 1, 2], name="feature")
+    )
+
 # CATEGORICAL HANDLINGS
 def check_dummy(split, name, value=None, others=False):
     assert split.node_type == Node.TYPES.CAT

webapps/error-analysis/backend.py

@@ -10,7 +10,18 @@
 
 from dku_error_analysis_decision_tree.tree_handler import TreeHandler
 
-app.json_encoder = DKUJSONEncoder
+try:
+    from flask.json.provider import DefaultJSONProvider
+    class DSSJSONProvider(DefaultJSONProvider):
+        def default(self, obj):
+            return DKUJSONEncoder().default(obj)
+
+    # Since Flask 2.2, jsonify delegates to app.json; app.json_encoder was
+    # deprecated then and removed in Flask 2.3.
+    app.json = DSSJSONProvider(app)
+except ImportError:
+    # Flask < 2.2 still uses the legacy JSON encoder API.
+    app.json_encoder = DKUJSONEncoder
 
 LOGGER = logging.getLogger(__name__)
 logging.basicConfig(level=logging.INFO, format="Error Analysis Plugin %(levelname)s - %(message)s")