[back][front] feat: support patching score and score max of criteria (#…

…1947)


Co-authored-by: Adrien Matissart <amatissart@users.noreply.github.com>

backend/ml/inputs.py

@@ -70,6 +70,7 @@ def get_comparisons(self, criteria=None, user_id=None) -> pd.DataFrame:
 
         values = scores_queryset.values(
             "score",
+            "score_max",
             "criteria",
             "weight",
             entity_a=F("comparison__entity_1_id"),
@@ -78,7 +79,9 @@ def get_comparisons(self, criteria=None, user_id=None) -> pd.DataFrame:
         )
         if len(values) > 0:
             dtf = pd.DataFrame(values)
-            return dtf[["user_id", "entity_a", "entity_b", "criteria", "score", "weight"]]
+            return dtf[
+                ["user_id", "entity_a", "entity_b", "criteria", "score", "score_max", "weight"]
+            ]
 
         return pd.DataFrame(
             columns=[
@@ -87,6 +90,7 @@ def get_comparisons(self, criteria=None, user_id=None) -> pd.DataFrame:
                 "entity_b",
                 "criteria",
                 "score",
+                "score_max",
                 "weight",
             ]
         )

backend/ml/mehestan/parameters.py

@@ -1,8 +1,7 @@
 from solidago.pipeline.legacy2023.parameters import PipelineParameters
 
-from tournesol.utils.constants import COMPARISON_MAX, MEHESTAN_MAX_SCALED_SCORE
+from tournesol.utils.constants import MEHESTAN_MAX_SCALED_SCORE
 
 
 class MehestanParameters(PipelineParameters):
-    r_max = COMPARISON_MAX
     max_squashed_score = MEHESTAN_MAX_SCALED_SCORE

backend/tournesol/lib/public_dataset.py

@@ -54,6 +54,7 @@ def get_comparisons_data(poll_name: str, until_: datetime) -> QuerySet:
             comparisoncriteriascore.criteria,
             comparisoncriteriascore.weight,
             comparisoncriteriascore.score,
+            comparisoncriteriascore.score_max,
             DATE(DATE_TRUNC('week', datetime_add)) AS week_date
 
         FROM tournesol_comparison
@@ -283,7 +284,15 @@ def write_comparisons_file(
 
     # If we want this function to be generic, the specific video_a and video_b
     # columns should be renamed entity_a and entity_b.
-    fieldnames = ["public_username", "video_a", "video_b", "criteria", "score", "week_date"]
+    fieldnames = [
+        "public_username",
+        "video_a",
+        "video_b",
+        "criteria",
+        "score",
+        "score_max",
+        "week_date"
+    ]
     writer = csv.DictWriter(write_target, fieldnames=fieldnames)
     writer.writeheader()
     writer.writerows(
@@ -293,6 +302,7 @@ def write_comparisons_file(
             "video_b": comparison.uid_b.split(UID_DELIMITER)[1],
             "criteria": comparison.criteria,
             "score": int(round(comparison.score)),
+            "score_max": comparison.score_max,
             "week_date": comparison.week_date,
         }
         for comparison in get_comparisons_data(poll_name, until_).iterator()

backend/tournesol/management/commands/load_public_dataset.py

@@ -110,6 +110,7 @@ def handle(self, *args, **options):
                         comparison=comparison,
                         criteria=values["criteria"],
                         score=values["score"],
+                        score_max=values["score_max"],
                     )
                 nb_comparisons += 1
             print(f"Created {nb_comparisons} comparisons")

backend/tournesol/migrations/0061_comparisoncriteriascore_score_max_and_more.py

@@ -0,0 +1,29 @@
+# Generated by Django 4.2.11 on 2024-05-02 09:05
+
+import django.core.validators
+from django.db import migrations, models
+
+
+class Migration(migrations.Migration):
+
+    dependencies = [
+        ("tournesol", "0060_remove_entity_tournesol_score"),
+    ]
+
+    operations = [
+        migrations.AddField(
+            model_name="comparisoncriteriascore",
+            name="score_max",
+            field=models.IntegerField(
+                default=10,
+                help_text="The absolute value of the maximum score.",
+                validators=[django.core.validators.MinValueValidator(1)],
+            ),
+            preserve_default=False,
+        ),
+        migrations.AlterField(
+            model_name="comparisoncriteriascore",
+            name="score",
+            field=models.FloatField(help_text="Score for the given comparison"),
+        ),
+    ]

backend/tournesol/models/comparisons.py

@@ -5,12 +5,11 @@
 import uuid
 
 import computed_property
-from django.core.validators import MaxValueValidator, MinValueValidator
+from django.core.validators import MinValueValidator
 from django.db import models
 from django.db.models import F, ObjectDoesNotExist, Q
 
 from core.models import User
-from tournesol.utils.constants import COMPARISON_MAX
 
 from .poll import Poll
 
@@ -129,7 +128,10 @@ class ComparisonCriteriaScore(models.Model):
     )
     score = models.FloatField(
         help_text="Score for the given comparison",
-        validators=[MinValueValidator(-COMPARISON_MAX), MaxValueValidator(COMPARISON_MAX)],
+    )
+    score_max = models.IntegerField(
+        help_text="The absolute value of the maximum score.",
+        validators=[MinValueValidator(1)],
     )
     # TODO: ask Lê if weights should be in a certain range (maybe always > 0)
     # and add validation if required
@@ -144,3 +146,17 @@ class Meta:
 
     def __str__(self):
         return f"{self.comparison}/{self.criteria}/{self.score}"
+
+    def save(self, *args, **kwargs):
+        if self.score_max is None:
+            raise TypeError("The value of score_max cannot be None.")
+
+        if self.score_max <= 0:
+            raise ValueError("The value of score_max must be greater than or equal to 1.")
+
+        if abs(self.score) > self.score_max:
+            raise ValueError(
+                f"The absolute value of the score {self.score} given to the criterion "
+                f"{self.criteria} can't be greater than the value of score_max {self.score_max}."
+            )
+        return super().save(*args, **kwargs)

backend/tournesol/resources/export_readme.txt

@@ -70,10 +70,15 @@ List of columns:
 
 - score:
 
-  The score is an integer between -10 and +10. Negative values indicate that
-  the user considers the video_a better, and positive values indicate that
-  they prefer the video_b. A score of 0 or close to 0 means that they find the
-  two videos similar.
+  The score is an integer in the range [-score_max, +score_max]. Negative
+  values indicate that the user considers the video_a better, and positive
+  values indicate that they prefer the video_b. A score of 0 or close to 0
+  means that they find the two videos similar.
+
+- score_max:
+
+  The absolute value of the minimum and maximum score. 10 means the user could
+  have given a score between -10 and +10 for the criterion.
 
 - week_date:
 

backend/tournesol/serializers/comparison.py

@@ -11,7 +11,7 @@
 class ComparisonCriteriaScoreSerializer(ModelSerializer):
     class Meta:
         model = ComparisonCriteriaScore
-        fields = ["criteria", "score", "weight"]
+        fields = ["criteria", "score", "score_max", "weight"]
 
     def validate_criteria(self, value):
         current_poll = self.context["poll"]
@@ -37,11 +37,15 @@ def reverse_criteria_scores(self, criteria_scores):
 
     def validate_criteria_scores(self, value):
         current_poll = self.context["poll"]
-        missing_criterias = set(current_poll.required_criterias_list) - set(
-            score["criteria"] for score in value
-        )
-        if missing_criterias:
-            raise ValidationError(f"Missing required criteria: {','.join(missing_criterias)}")
+        partial_update = self.context.get("partial_update")
+
+        if not partial_update:
+            missing_criterias = set(current_poll.required_criterias_list) - set(
+                score["criteria"] for score in value
+            )
+            if missing_criterias:
+                raise ValidationError(f"Missing required criteria: {','.join(missing_criterias)}")
+
         return value
 
 
@@ -193,9 +197,19 @@ def update(self, instance, validated_data):
             instance.duration_ms = validated_data.get("duration_ms")
 
         instance.save()
-        instance.criteria_scores.all().delete()
 
-        for criteria_score in validated_data.pop("criteria_scores"):
-            instance.criteria_scores.create(**criteria_score)
+        partial_update = self.context.get("partial_update")
+
+        if partial_update:
+            for criteria_score in validated_data.pop("criteria_scores"):
+                instance.criteria_scores.update_or_create(
+                    criteria=criteria_score["criteria"],
+                    defaults={**criteria_score}
+                )
+        else:
+            instance.criteria_scores.all().delete()
+
+            for criteria_score in validated_data.pop("criteria_scores"):
+                instance.criteria_scores.create(**criteria_score)
 
         return instance

backend/tournesol/serializers/inconsistencies.py

@@ -56,6 +56,7 @@ class ScoreInconsistencySerializer(Serializer):
     entity_1_rating = serializers.FloatField()
     entity_2_rating = serializers.FloatField()
     comparison_score = serializers.FloatField()
+    comparison_score_max = serializers.IntegerField()
     expected_comparison_score = serializers.FloatField()
 
 

backend/tournesol/tests/factories/comparison.py

@@ -31,3 +31,4 @@ class Meta:
     comparison = factory.SubFactory(ComparisonFactory)
     criteria = "better_habits"
     score = fuzzy.FuzzyDecimal(-10, 10)
+    score_max = 10