[Fix] Return a File instead of raw text for new export endpoint (#442)

Co-authored-by: Tim Lorsbach <tim@lorsba.ch>
Reviewed-on: enviPath/enviPy#442

bridge/contracts.py

@@ -1,4 +1,5 @@
 import enum
+from typing import Any, Dict
 from abc import ABC, abstractmethod
 
 from envipy_additional_information import EnviPyModel
@@ -69,6 +70,12 @@ class Plugin(ABC):
 
 
 class Property(Plugin):
+    def parameters(self) -> Dict[str, Any]:
+        """
+        Returns the parameters of the PropertyPlugin.
+        """
+        return {}
+
     @classmethod
     @abstractmethod
     def requires_rule_packages(cls) -> bool:
@@ -300,6 +307,12 @@ class Classifier(Plugin):
         """
         pass
 
+    def parameters(self) -> Dict[str, Any]:
+        """
+        Returns the parameters of the ClassifierPlugin.
+        """
+        return {}
+
     @abstractmethod
     def build(self, eP: EnviPyDTO, *args, **kwargs) -> BuildResult | None:
         """

epapi/v1/interfaces/iuclid/projections.py

@@ -56,7 +56,7 @@ def get_pathway_for_iuclid_export(user, pathway_uuid: UUID) -> PathwayExportDTO:
 
         ai_for_node = []
         scenario_entries: list[PathwayScenarioDTO] = []
-        for scenario in sorted(node.scenarios.all(), key=lambda item: item.pk):
+        for scenario in sorted(node.get_scenarios(), key=lambda item: item.pk):
             ai_for_scenario = list(scenario.get_additional_information(direct_only=True))
             ai_for_node.extend(ai_for_scenario)
             scenario_entries.append(

epdb/legacy_api.py

@@ -46,6 +46,10 @@ from .models import (
 Package = s.GET_PACKAGE_MODEL()
 
 
+def get_package_for_read(user, package_uuid):
+    return PackageManager.get_package_by_id(user, package_uuid)
+
+
 def get_package_for_write(user, package_uuid):
     p = PackageManager.get_package_by_id(user, package_uuid)
     if not PackageManager.writable(user, p):
@@ -2291,3 +2295,42 @@ def predict(request, np: Form[NonPersistent]):
         return 403, {
             "message": f"Getting Setting with id {np.setting_url} failed due to insufficient rights!"
         }
+
+
+##########
+# Export #
+##########
+class PackageExportInSchema(Schema):
+    package_uuid: str
+    additional_information_types: List[str] | None = None
+
+
+@router.get("/export", response={200: Any, 403: Error})
+def export(request, q: Query[PackageExportInSchema]):
+    try:
+        p = get_package_for_read(request.user, q.package_uuid)
+
+        from envipy_additional_information import registry
+        from utilities.misc import PathwayExporter
+
+        ai_types = []
+        if q.additional_information_types is not None:
+            for ai_type in q.additional_information_types:
+                if registry.get_model(ai_type) is None:
+                    return 400, {
+                        "message": f"Exporting Package with id {q.package_uuid} failed as {ai_type} is not a valid additional information type!"
+                    }
+                ai_types.append(ai_type)
+
+        exporter = PathwayExporter(p, add_infs_to_export=ai_types)
+        res = exporter.do_export()
+
+        filename = f"{p.get_name().replace(' ', '_')}_{p.uuid}.tsv"
+        response = HttpResponse(res, content_type="text/csv")
+        response["Content-Disposition"] = f'attachment; filename="{filename}"'
+
+        return response
+    except ValueError:
+        return 403, {
+            "message": f"Exporting Package with id {q.package_uuid} failed due to insufficient rights!"
+        }

epdb/models.py

@@ -2162,7 +2162,7 @@ class Pathway(EnviPathModel, AliasMixin, ScenarioMixin, AdditionalInformationMix
 
             row += [cs.smiles, cs.get_name(), n.depth]
 
-            edges = self.edges.filter(end_nodes__in=[n])
+            edges = self.edges.filter(end_nodes=n)
             if len(edges):
                 for e in edges:
                     _row = row.copy()
@@ -2847,6 +2847,58 @@ class PackageBasedModel(EPModel):
 
         return res
 
+    def parameters(self):
+        params = {
+            "Model Evaluation Threshold": f"{self.threshold:.2f}",
+            "Multi Gen Evaluation": "Yes" if self.multigen_eval else "No",
+        }
+
+        if self.app_domain:
+            params["Applicability Domain Num Neighbors"] = f"{self.app_domain.num_neighbours:.2f}"
+            params["Applicability Domain Reliability Threshold"] = (
+                f"{self.app_domain.reliability_threshold:.2f}"
+            )
+            params["Applicability Domain Local Compatibility Threshold"] = (
+                f"{self.app_domain.local_compatibilty_threshold:.2f}"
+            )
+
+        return params
+
+    def statistics(self):
+        from sklearn.metrics import auc
+
+        recall = list(self.eval_results["average_recall_per_threshold"].values())
+        precision = list(self.eval_results["average_precision_per_threshold"].values())
+        mg_recall = list(
+            self.eval_results.get("multigen_average_recall_per_threshold", {}).values()
+        )
+        mg_precision = list(
+            self.eval_results.get("multigen_average_precision_per_threshold", {}).values()
+        )
+
+        return {
+            "accuracy": [
+                self.eval_results["average_accuracy"],
+                self.eval_results.get("multigen_average_accuracy"),
+            ],
+            "precision": [
+                self.eval_results["average_precision_per_threshold"][f"{self.threshold:.2f}"],
+                self.eval_results.get("multigen_average_precision_per_threshold", {}).get(
+                    f"{self.threshold:.2f}"
+                ),
+            ],
+            "recall": [
+                self.eval_results["average_recall_per_threshold"][f"{self.threshold:.2f}"],
+                self.eval_results.get("multigen_average_recall_per_threshold", {}).get(
+                    f"{self.threshold:.2f}"
+                ),
+            ],
+            "Area under PR Curve": [
+                auc(recall, precision),
+                auc(mg_recall, mg_precision) if self.multigen_eval else None,
+            ],
+        }
+
     @cached_property
     def applicable_rules(self) -> List["Rule"]:
         """
@@ -3027,8 +3079,6 @@ class PackageBasedModel(EPModel):
                 thresholds.append(np.float64(threshold))
                 thresholds.sort()
 
-            logger.info(f"Thresholds: {thresholds}")
-
             precision = {f"{t:.2f}": [] for t in thresholds}
             recall = {f"{t:.2f}": [] for t in thresholds}
 
@@ -3077,14 +3127,17 @@ class PackageBasedModel(EPModel):
             for t in thresholds:
                 for true, pred in zip(pathways, pred_pathways):
                     acc, pre, rec = multigen_eval(true, pred, t)
-                    if abs(t - threshold) < 0.01:
+
-                        mg_acc = acc
+                    if f"{t:.2f}" == f"{threshold:.2f}":
+                        mg_acc += acc
+
                     precision[f"{t:.2f}"].append(pre)
                     recall[f"{t:.2f}"].append(rec)
 
+            avg_mg_acc = mg_acc / len(root_compounds)
             precision = {k: sum(v) / len(v) if len(v) > 0 else 0 for k, v in precision.items()}
             recall = {k: sum(v) / len(v) if len(v) > 0 else 0 for k, v in recall.items()}
-            return mg_acc, precision, recall
+            return avg_mg_acc, precision, recall
 
         # If there are eval packages perform single generation evaluation on them instead of random splits
         if self.eval_packages.count() > 0:
@@ -4216,6 +4269,9 @@ class ClassifierPluginModel(PackageBasedModel):
         instance = impl(conf)
         return instance
 
+    def parameters(self):
+        return self.instance().parameters()
+
     def build_dataset(self):
         """
         Required by general model contract but actual implementation resides in plugin.
@@ -4436,6 +4492,9 @@ class PropertyPluginModel(PackageBasedModel):
         instance = impl()
         return instance
 
+    def parameters(self):
+        return self.instance().parameters()
+
     def build_dataset(self):
         """
         Required by general model contract but actual implementation resides in plugin.

epdb/tasks.py

@@ -477,8 +477,7 @@ def batch_predict(
             limit=None,
             setting_overrides={
                 "max_nodes": num_tps,
-                "max_depth": num_tps,
+                "model_threshold": 0.0,
-                "model_threshold": 0.001,
             },
         )
 

epiuclid/serializers/pathway_mapper.py

@@ -351,7 +351,8 @@ class PathwayMapper:
 
         props = SoilPropertiesData()
 
-        for ai in ai_list:
+        for ai_obj in ai_list:
+            ai = ai_obj.get()
             if isinstance(ai, SoilTexture1) and props.soil_type is None:
                 props.soil_type = ai.type.value
             elif isinstance(ai, SoilTexture2):

templates/objects/model/_model_base.html

@@ -117,6 +117,39 @@
           </div>
         </div>
       {% endif %}
+
+      {% if model.parameters %}
+        <!-- Model Parameters Panel -->
+        <div class="collapse-arrow bg-base-200 collapse">
+          <input type="checkbox" checked />
+          <div class="collapse-title text-xl font-medium">Model Parameters</div>
+          <div class="collapse-content">
+            <div class="flex justify-center">
+              <div
+                id="model-stats"
+                class="overflow-x-auto rounded-box shadow-md bg-base-100"
+              >
+                <table class="table table-fixed w-full">
+                  <thead class="text-base">
+                    <tr>
+                      <th class="w-3/5">Parameter</th>
+                      <th>Value</th>
+                    </tr>
+                  </thead>
+                  <tbody>
+                    {% for param, value in model.parameters.items %}
+                      <tr>
+                        <td>{{ param }}</td>
+                        <td>{{ value }}</td>
+                      </tr>
+                    {% endfor %}
+                  </tbody>
+                </table>
+              </div>
+            </div>
+          </div>
+        </div>
+      {% endif %}
     {% endblock %}
 
     {% block usemodel %}

templates/objects/model/classification_model.html

@@ -313,6 +313,44 @@
                 </div>
             </div>
         {% endif %}
+
+      <!-- Model Statistics Panel -->
+      <div class="collapse-arrow bg-base-200 collapse">
+        <input type="checkbox" checked />
+        <div class="collapse-title text-xl font-medium">Model Statistics for threshold {{ model.threshold }}</div>
+        <div class="collapse-content">
+          <div class="flex justify-center">
+            <div
+              id="model-stats"
+              class="overflow-x-auto rounded-box shadow-md bg-base-100"
+            >
+              <table class="table table-fixed w-full">
+                <thead class="text-base">
+                  <tr>
+                    <th class="w-1/5">Metric</th>
+                    <th>Single Gen Value</th>
+                    {% if model.multigen_eval %}
+                      <th>Multi Gen Value</th>
+                    {% endif %}
+                  </tr>
+                </thead>
+                <tbody>
+                  {% for metric, value in model.statistics.items %}
+                    <tr>
+                      <td>{{ metric|upper }}</td>
+                      <td>{{ value.0|floatformat:3 }}</td>
+                      {% if model.multigen_eval %}
+                        <td>{{ value.1|floatformat:3 }}</td>
+                      {% endif %}
+                    </tr>
+                  {% endfor %}
+                </tbody>
+              </table>
+            </div>
+          </div>
+        </div>
+      </div>
+
     {% endif %}
     <script>
         function makeChart(selector, data) {

templates/objects/node.html

@@ -139,7 +139,7 @@
                           </tr>
                           <tr>
                             <td>Model</td>
-                            <td>{{ half_lifes.0.model }}</td>
+                            <td>{{ half_lifes.0.model.value }}</td>
                           </tr>
                         </tbody>
                       </table>

templates/objects/pathway.html

@@ -535,7 +535,7 @@
         <input type="checkbox" />
         <div class="collapse-title text-xl font-medium">Setting</div>
         <div class="collapse-content">
-          {% with setting_to_render=pathway.setting can_be_default=False %}
+          {% with setting_to_render=pathway.setting_with_overrides can_be_default=False %}
             {% include "objects/setting_template.html" %}
           {% endwith %}
         </div>

templates/objects/user.html

@@ -95,21 +95,23 @@
     </div>
 
     <!-- Other Prediction Settings -->
-    <div class="collapse-arrow bg-base-200 collapse">
+    {% if meta.available_settings|length > 1 %}
-      <input type="checkbox" />
+      <div class="collapse-arrow bg-base-200 collapse">
-      <div class="collapse-title text-xl font-medium">
+        <input type="checkbox" />
-        Other Prediction Settings
+        <div class="collapse-title text-xl font-medium">
+          Other Prediction Settings
+        </div>
+        <div class="collapse-content space-y-3">
+          {% for setting in meta.available_settings %}
+            {% if setting != user.default_setting %}
+              {% with setting_to_render=setting can_be_default=True %}
+                {% include "objects/setting_template.html" %}
+              {% endwith %}
+            {% endif %}
+          {% endfor %}
+        </div>
       </div>
-      <div class="collapse-content space-y-3">
+    {% endif %}
-        {% for setting in meta.available_settings %}
-          {% if setting != user.default_setting %}
-            {% with setting_to_render=setting can_be_default=True %}
-              {% include "objects/setting_template.html" %}
-            {% endwith %}
-          {% endif %}
-        {% endfor %}
-      </div>
-    </div>
   </div>
 
   <script>

tests/test_jobs.py

@@ -58,7 +58,7 @@ class MultiGenTest(TestCase):
 
         self.assertEqual(
             pw.setting_with_overrides.max_depth,
-            f"{num_tps} (this is an override for this particular pathway)",
+            5,
         )
         self.assertEqual(
             pw.setting_with_overrides.max_nodes,

utilities/misc.py

@@ -4,11 +4,13 @@ import hmac
 import json
 import logging
 import uuid
+from abc import ABC, abstractmethod
 from collections import defaultdict
 from datetime import datetime
-from typing import Any, Dict, List, Optional, TYPE_CHECKING
+from typing import Any, Dict, List, Optional, TYPE_CHECKING, Type
 
 from django.conf import settings as s
+from envipy_additional_information import EnviPyModel, UIConfig
 from ninja import Schema
 from pydantic import HttpUrl, ValidationError
 
@@ -132,7 +134,7 @@ class EnzymeExportSchema(RefEnzymeExportSchema):
 
 
 class EnzymeRuleExportSchema(RefRuleExportSchema):
-    enzymes: List[EnzymeExportSchema] | None = None
+    enzymes: List[EnzymeExportSchema] = []
 
     @staticmethod
     def resolve_enzymes(obj):
@@ -277,15 +279,20 @@ class PackageExportSchema(Schema):
         return AdditionalInformation.objects.filter(package=obj)
 
 
-class PackageExporter:
+class Exporter(ABC):
     def __init__(self, package: Package):
         self._raw_package = package
 
     def do_export(self):
-        return PackageExporter._export_package_as_json(self._raw_package)
+        return self._export()
 
-    @staticmethod
+    @abstractmethod
-    def _export_package_as_json(package: Package) -> Dict[str, Any]:
+    def _export(self):
+        pass
+
+
+class PackageExporter(Exporter):
+    def _export(self) -> Dict[str, Any]:
         """
         Dumps a Package and all its related objects as JSON.
 
@@ -296,11 +303,126 @@ class PackageExporter:
             Dict containing the complete package data as JSON-serializable structure
         """
 
-        data = PackageExportSchema.from_orm(package)
+        data = PackageExportSchema.from_orm(self._raw_package)
 
         return data.model_dump(mode="json")
 
 
+class PathwayExporter(Exporter):
+    def __init__(self, package: Package, add_infs_to_export: List[str] = []):
+        super().__init__(package)
+        self._add_infs_to_export = add_infs_to_export
+
+    def _flatten_additional_information(self, ai: AdditionalInformation) -> dict[str, Any]:
+        model_cls: Type[EnviPyModel] = type(ai.get())
+
+        def _flatten(d: dict, parent_key: str = "") -> dict[str, Any]:
+            items: dict[str, Any] = {}
+            for key, value in d.items():
+                new_key = f"{parent_key}__{key.lower()}" if parent_key else key.lower()
+                if isinstance(value, dict):
+                    items.update(_flatten(value, new_key))
+                else:
+                    items[new_key] = value
+            return items
+
+        flat = _flatten(ai.data, ai.type)
+
+        ui_class = getattr(model_cls, "UI", None)
+        if ui_class is None:
+            return flat
+
+        for f in model_cls.model_fields:
+            ui_info = getattr(ui_class, f, None)
+
+            if not isinstance(ui_info, UIConfig) or ui_info.unit is None:
+                continue
+
+            flat[f"{model_cls.__name__}__{f}__unit"] = ui_info.unit
+
+        return flat
+
+    def _export(self):
+        from io import StringIO
+        from csv import DictWriter
+
+        rows = []
+
+        for pw in self._raw_package.pathways.all():
+            for n in pw.nodes:
+                for scen in pw.scenarios.all():
+                    row = {
+                        "pathway_name": pw.name,
+                        "pathway_id": str(pw.url),
+                        "node_depth": n.depth,
+                        "compound_id": str(n.default_node_label.compound.url),
+                        "pubchem_ID": n.default_node_label.pubchem_compound_id,
+                        "compound_name": n.default_node_label.compound.name,
+                        "compound_smiles": n.default_node_label.smiles,
+                        "scenario_id": str(scen.url),
+                        "scenario_name": scen.name,
+                        "scenario_type": scen.scenario_type,
+                        "scenario_description": scen.description,
+                    }
+
+                    if self._add_infs_to_export:
+                        ai_qs = AdditionalInformation.objects.filter(
+                            scenario=scen, type__in=self._add_infs_to_export
+                        )
+                    else:
+                        ai_qs = AdditionalInformation.objects.filter(scenario=scen)
+
+                    for ai in ai_qs:
+                        if ai.type == "ProposedIntermediate" and ai.content_object == n:
+                            row.update({"proposed_intermediate": True})
+                        elif ai.type == "SpikeCompound":
+                            spike = {"SpikeCompound__url": ai.get().url}
+
+                            try:
+                                struc = CompoundStructure.objects.get(
+                                    compound__package=self._raw_package, url=ai.get().url
+                                )
+                                spike["SpikeCompound__smiles"] = struc.smiles
+                            except Exception:
+                                spike["SpikeCompound__smiles"] = None
+
+                            row.update(**spike)
+                        else:
+                            row.update(self._flatten_additional_information(ai))
+                    rows.append(row)
+
+        # Get all header fields
+        all_header_fields = set()
+        for row in rows:
+            all_header_fields.update(row.keys())
+
+        # Per request the CSV should start with these fields
+        header = [
+            "pathway_name",
+            "pathway_id",
+            "node_depth",
+            "compound_id",
+            "pubchem_ID",
+            "compound_name",
+            "compound_smiles",
+            "scenario_id",
+            "scenario_name",
+            "scenario_type",
+            "scenario_description",
+        ]
+
+        # User remaining fields and place them after the predefined values in a sorted manner
+        remainder = sorted(list(all_header_fields.difference(set(header))))
+        header.extend(remainder)
+
+        buffer = StringIO()
+        writer = DictWriter(buffer, fieldnames=header, delimiter="\t")
+        writer.writeheader()
+        writer.writerows(rows)
+        buffer.seek(0)
+        return buffer.getvalue()
+
+
 class PackageImporter:
     def __init__(self, package: Dict[str, Any], preserve_uuids: bool = False):
         self.preserve_uuids = preserve_uuids