fix: import error for TestsetGeneration and small fixes (#1516)

src/ragas/dataset_schema.py

@@ -2,6 +2,7 @@
 
 import json
 import typing as t
+from abc import ABC, abstractmethod
 from dataclasses import dataclass, field
 
 from datasets import Dataset as HFDataset
@@ -12,6 +13,8 @@
 from ragas.utils import safe_nanmean
 
 if t.TYPE_CHECKING:
+    from pathlib import Path
+
     from datasets import Dataset as HFDataset
     from pandas import DataFrame as PandasDataframe
 
@@ -136,9 +139,20 @@ def pretty_repr(self):
 Sample = t.TypeVar("Sample", bound=BaseSample)
 
 
-class RagasDataset(BaseModel, t.Generic[Sample]):
+class RagasDataset(ABC, BaseModel, t.Generic[Sample]):
     samples: t.List[Sample]
 
+    @abstractmethod
+    def to_list(self) -> t.List[t.Dict]:
+        """Converts the samples to a list of dictionaries."""
+        pass
+
+    @classmethod
+    @abstractmethod
+    def from_list(cls, data: t.List[t.Dict]) -> RagasDataset[Sample]:
+        """Creates an EvaluationDataset from a list of dictionaries."""
+        pass
+
     @field_validator("samples")
     def validate_samples(cls, samples: t.List[BaseSample]) -> t.List[BaseSample]:
         """Validates that all samples are of the same type."""
@@ -155,20 +169,6 @@ def get_sample_type(self) -> t.Type[Sample]:
         """Returns the type of the samples in the dataset."""
         return type(self.samples[0])
 
-    def _to_list(self) -> t.List[t.Dict]:
-        """Converts the samples to a list of dictionaries."""
-        rows = [sample.to_dict() for sample in self.samples]
-
-        if self.get_sample_type() == MultiTurnSample:
-            for sample in rows:
-                for item in sample["user_input"]:
-                    if not isinstance(item["content"], str):
-                        item["content"] = json.dumps(
-                            item["content"], ensure_ascii=False
-                        )
-
-        return rows
-
     def to_hf_dataset(self) -> HFDataset:
         """Converts the dataset to a Hugging Face Dataset."""
         try:
@@ -178,7 +178,7 @@ def to_hf_dataset(self) -> HFDataset:
                 "datasets is not installed. Please install it to use this function."
             )
 
-        return HFDataset.from_list(self._to_list())
+        return HFDataset.from_list(self.to_list())
 
     @classmethod
     def from_hf_dataset(cls, dataset: HFDataset):
@@ -194,26 +194,13 @@ def to_pandas(self) -> PandasDataframe:
                 "pandas is not installed. Please install it to use this function."
             )
 
-        data = self._to_list()
+        data = self.to_list()
         return pd.DataFrame(data)
 
     def features(self):
         """Returns the features of the samples."""
         return self.samples[0].get_features()
 
-    @classmethod
-    def from_list(cls, mapping: t.List[t.Dict]):
-        """Creates an EvaluationDataset from a list of dictionaries."""
-        samples = []
-        if all(
-            "user_input" in item and isinstance(mapping[0]["user_input"], list)
-            for item in mapping
-        ):
-            samples.extend(MultiTurnSample(**sample) for sample in mapping)
-        else:
-            samples.extend(SingleTurnSample(**sample) for sample in mapping)
-        return cls(samples=samples)
-
     @classmethod
     def from_dict(cls, mapping: t.Dict):
         """Creates an EvaluationDataset from a dictionary."""
@@ -227,40 +214,30 @@ def from_dict(cls, mapping: t.Dict):
             samples.extend(SingleTurnSample(**sample) for sample in mapping)
         return cls(samples=samples)
 
-    @classmethod
-    def from_csv(cls, path: str):
-        """Creates an EvaluationDataset from a CSV file."""
-        import csv
-
-        with open(path, "r", newline="") as csvfile:
-            reader = csv.DictReader(csvfile)
-            data = [row for row in reader]
-        return cls.from_list(data)
-
-    def to_csv(self, path: str):
+    def to_csv(self, path: t.Union[str, Path]):
         """Converts the dataset to a CSV file."""
         import csv
 
-        data = self._to_list()
+        data = self.to_list()
         if not data:
             return
 
-        fieldnames = self.features()
+        fieldnames = data[0].keys()
 
         with open(path, "w", newline="") as csvfile:
             writer = csv.DictWriter(csvfile, fieldnames=fieldnames)
             writer.writeheader()
             for row in data:
                 writer.writerow(row)
 
-    def to_jsonl(self, path: str):
+    def to_jsonl(self, path: t.Union[str, Path]):
         """Converts the dataset to a JSONL file."""
         with open(path, "w") as jsonlfile:
             for sample in self.samples:
                 jsonlfile.write(json.dumps(sample.to_dict(), ensure_ascii=False) + "\n")
 
     @classmethod
-    def from_jsonl(cls, path: str):
+    def from_jsonl(cls, path: t.Union[str, Path]):
         """Creates an EvaluationDataset from a JSONL file."""
         with open(path, "r") as jsonlfile:
             data = [json.loads(line) for line in jsonlfile]
@@ -307,8 +284,6 @@ class EvaluationDataset(RagasDataset[SingleTurnSampleOrMultiTurnSample]):
         Creates an EvaluationDataset from a list of dictionaries.
     from_dict(mapping)
         Creates an EvaluationDataset from a dictionary.
-    from_csv(path)
-        Creates an EvaluationDataset from a CSV file.
     to_csv(path)
         Converts the dataset to a CSV file.
     to_jsonl(path)
@@ -333,6 +308,37 @@ def __getitem__(
         else:
             raise TypeError("Index must be int or slice")
 
+    def to_list(self) -> t.List[t.Dict]:
+        rows = [sample.to_dict() for sample in self.samples]
+
+        if self.get_sample_type() == MultiTurnSample:
+            for sample in rows:
+                for item in sample["user_input"]:
+                    if not isinstance(item["content"], str):
+                        item["content"] = json.dumps(
+                            item["content"], ensure_ascii=False
+                        )
+
+        return rows
+
+    @classmethod
+    def from_list(cls, data: t.List[t.Dict]) -> EvaluationDataset:
+        samples = []
+        if all(
+            "user_input" in item and isinstance(data[0]["user_input"], list)
+            for item in data
+        ):
+            samples.extend(MultiTurnSample(**sample) for sample in data)
+        else:
+            samples.extend(SingleTurnSample(**sample) for sample in data)
+        return cls(samples=samples)
+
+
+class EvaluationResultRow(BaseModel):
+    dataset_row: t.Dict
+    scores: t.Dict[str, t.Any]
+    trace: t.Dict[str, t.Any] = field(default_factory=dict)  # none for now
+
 
 @dataclass
 class EvaluationResult:
@@ -352,7 +358,7 @@ class EvaluationResult:
     """
 
     scores: t.List[t.Dict[str, t.Any]]
-    dataset: t.Optional[EvaluationDataset] = None
+    dataset: EvaluationDataset
     binary_columns: t.List[str] = field(default_factory=list)
     cost_cb: t.Optional[CostCallbackHandler] = None
 
@@ -407,6 +413,18 @@ def to_pandas(self, batch_size: int | None = None, batched: bool = False):
         dataset_df = self.dataset.to_pandas()
         return pd.concat([dataset_df, scores_df], axis=1)
 
+    def serialized(self) -> t.List[EvaluationResultRow]:
+        """
+        Convert the result to a list of EvaluationResultRow.
+        """
+        return [
+            EvaluationResultRow(
+                dataset_row=self.dataset[i].to_dict(),
+                scores=self.scores[i],
+            )
+            for i in range(len(self.scores))
+        ]
+
     def total_tokens(self) -> t.Union[t.List[TokenUsage], TokenUsage]:
         """
         Compute the total tokens used in the evaluation.

src/ragas/testset/synthesizers/testset_schema.py

@@ -2,14 +2,16 @@
 
 import typing as t
 
-from ragas.dataset_schema import BaseSample, RagasDataset
+from ragas.dataset_schema import (
+    BaseSample,
+    EvaluationDataset,
+    MultiTurnSample,
+    RagasDataset,
+    SingleTurnSample,
+)
 
 if t.TYPE_CHECKING:
-    from ragas.dataset_schema import (
-        EvaluationDataset,
-        MultiTurnSample,
-        SingleTurnSample,
-    )
+    from ragas.dataset_schema import MultiTurnSample, SingleTurnSample
 
 
 class TestsetSample(BaseSample):
@@ -48,13 +50,34 @@ def to_evaluation_dataset(self) -> EvaluationDataset:
             samples=[sample.eval_sample for sample in self.samples]
         )
 
-    def _to_list(self) -> t.List[t.Dict]:
-        eval_list = self.to_evaluation_dataset()._to_list()
-        testset_list_without_eval_sample = [
-            sample.model_dump(exclude={"eval_sample"}) for sample in self.samples
-        ]
-        testset_list = [
-            {**eval_sample, **sample}
-            for eval_sample, sample in zip(eval_list, testset_list_without_eval_sample)
-        ]
-        return testset_list
+    def to_list(self) -> t.List[t.Dict]:
+        """
+        Converts the Testset to a list of dictionaries.
+        """
+        return [sample.model_dump() for sample in self.samples]
+
+    @classmethod
+    def from_list(cls, data: t.List[t.Dict]) -> Testset:
+        """
+        Converts a list of dictionaries to a Testset.
+        """
+        # first create the samples
+        samples = []
+        for sample in data:
+            eval_sample = sample["eval_sample"]
+
+            # if user_input is a list it is MultiTurnSample
+            if "user_input" in eval_sample and not isinstance(
+                eval_sample.get("user_input"), list
+            ):
+                eval_sample = SingleTurnSample(**sample["eval_sample"])
+            else:
+                eval_sample = MultiTurnSample(**sample["eval_sample"])
+
+            samples.append(
+                TestsetSample(
+                    eval_sample=eval_sample, synthesizer_name=sample["synthesizer_name"]
+                )
+            )
+        # then create the testset
+        return Testset(samples=samples)

tests/unit/test_dataset_schema.py

@@ -3,46 +3,66 @@
 import pytest
 from pydantic import ValidationError
 
-from ragas.dataset_schema import EvaluationDataset, MultiTurnSample, SingleTurnSample
-
-
-def test_evaluation_dataset():
-    single_turn_sample = SingleTurnSample(user_input="What is X", response="Y")
-
-    dataset = EvaluationDataset(samples=[single_turn_sample, single_turn_sample])
+from ragas.dataset_schema import (
+    EvaluationDataset,
+    HumanMessage,
+    MultiTurnSample,
+    SingleTurnSample,
+)
+
+samples = [
+    SingleTurnSample(user_input="What is X", response="Y"),
+    MultiTurnSample(
+        user_input=[HumanMessage(content="What is X")],
+        reference="Y",
+    ),
+]
+
+
+@pytest.mark.parametrize("eval_sample", samples)
+def test_evaluation_dataset(eval_sample):
+    dataset = EvaluationDataset(samples=[eval_sample, eval_sample])
 
     hf_dataset = dataset.to_hf_dataset()
 
-    assert dataset.get_sample_type() == SingleTurnSample
+    assert dataset.get_sample_type() is type(eval_sample)
     assert len(hf_dataset) == 2
-    assert dataset.features() == ["user_input", "response"]
     assert len(dataset) == 2
-    assert dataset[0] == single_turn_sample
+    assert dataset[0] == eval_sample
 
 
-def test_evaluation_dataset_save_load(tmpdir):
-    single_turn_sample = SingleTurnSample(user_input="What is X", response="Y")
+@pytest.mark.parametrize("eval_sample", samples)
+def test_evaluation_dataset_save_load_csv(tmpdir, eval_sample):
+    dataset = EvaluationDataset(samples=[eval_sample, eval_sample])
 
-    dataset = EvaluationDataset(samples=[single_turn_sample, single_turn_sample])
+    # save and load to csv
+    csv_path = tmpdir / "csvfile.csv"
+    dataset.to_csv(csv_path)
 
-    hf_dataset = dataset.to_hf_dataset()
 
-    # save and load to csv
-    dataset.to_csv(tmpdir / "csvfile.csv")
-    loaded_dataset = EvaluationDataset.from_csv(tmpdir / "csvfile.csv")
-    assert loaded_dataset == dataset
+@pytest.mark.parametrize("eval_sample", samples)
+def test_evaluation_dataset_save_load_jsonl(tmpdir, eval_sample):
+    dataset = EvaluationDataset(samples=[eval_sample, eval_sample])
 
     # save and load to jsonl
-    dataset.to_jsonl(tmpdir / "jsonlfile.jsonl")
-    loaded_dataset = EvaluationDataset.from_jsonl(tmpdir / "jsonlfile.jsonl")
+    jsonl_path = tmpdir / "jsonlfile.jsonl"
+    dataset.to_jsonl(jsonl_path)
+    loaded_dataset = EvaluationDataset.from_jsonl(jsonl_path)
     assert loaded_dataset == dataset
 
-    # load from hf dataset
+
+@pytest.mark.parametrize("eval_sample", samples)
+def test_evaluation_dataset_load_from_hf(eval_sample):
+    dataset = EvaluationDataset(samples=[eval_sample, eval_sample])
+
+    # convert to and load from hf dataset
+    hf_dataset = dataset.to_hf_dataset()
     loaded_dataset = EvaluationDataset.from_hf_dataset(hf_dataset)
     assert loaded_dataset == dataset
 
 
-def test_single_type_evaluation_dataset():
+@pytest.mark.parametrize("eval_sample", samples)
+def test_single_type_evaluation_dataset(eval_sample):
     single_turn_sample = SingleTurnSample(user_input="What is X", response="Y")
     multi_turn_sample = MultiTurnSample(
         user_input=[{"content": "What is X"}],