fix: warn when a task resolves to zero datasets

src/workrb/tasks/abstract/base.py

@@ -70,11 +70,28 @@ def __init__(
 
         # Select dataset identifiers that match the requested languages
         self.dataset_ids = self.languages_to_dataset_ids(self.languages)
+        if not self.dataset_ids:
+            logger.warning(
+                "Task '%s' has no datasets for the requested languages %s. "
+                "The task will produce no results. For cross-lingual tasks, the "
+                "requested languages must cover both the query and the target "
+                "language of a dataset (e.g. languages=['de', 'en'] for a "
+                "German-to-English dataset).",
+                self.name,
+                [lang.value for lang in self.languages],
+            )
 
         # Load datasets for the selected dataset identifiers
         self.datasets = self._load_datasets(dataset_ids=self.dataset_ids, split=self.split)
 
         # Skipped datasets (DatasetConfigNotSupported) are excluded from self.dataset_ids
+        if self.dataset_ids and not self.datasets:
+            logger.warning(
+                "Task '%s': all %d selected datasets were skipped during loading. "
+                "The task will produce no results.",
+                self.name,
+                len(self.dataset_ids),
+            )
         self.dataset_ids = list(self.datasets.keys())
 
     def _parse_languages(

tests/test_task_loading.py

@@ -1,5 +1,6 @@
 """Test loading of all concrete task implementations."""
 
+import logging
 from typing import Literal
 
 import pytest
@@ -17,13 +18,19 @@
     JobTitleSimilarityRanking,
     MELORanking,
     MELSRanking,
+    SearchQueryCandidateRanking,
     SkillMatch1kSkillSimilarityRanking,
     SkillSkapeExtractRanking,
     SkillSkapeGradedSkillExtractRanking,
     TechGradedSkillExtractRanking,
     TechSkillExtractRanking,
 )
-from workrb.tasks.abstract.base import DatasetSplit, Language, TaskType
+from workrb.tasks.abstract.base import (
+    DatasetConfigNotSupported,
+    DatasetSplit,
+    Language,
+    TaskType,
+)
 from workrb.types import DatasetLanguages
 
 
@@ -233,6 +240,68 @@ def test_init_task_with_skip_on_unsupported_languages(
         )
 
 
+class TestEmptyDatasetWarnings:
+    """Tasks that resolve to zero datasets must warn instead of staying silent."""
+
+    @pytest.fixture(autouse=True)
+    def propagate_workrb_logs(self):
+        """Enable propagation on the 'workrb' logger for the duration of each test.
+
+        The package logger disables propagation, which hides its records
+        from caplog's root handler.
+        """
+        workrb_logger = logging.getLogger("workrb")
+        original = workrb_logger.propagate
+        workrb_logger.propagate = True
+        yield
+        workrb_logger.propagate = original
+
+    def test_warns_when_no_dataset_matches_requested_languages(self, caplog):
+        """Warn when the language selection matches no dataset.
+
+        A cross-lingual task requested with a single language selects no
+        datasets: each dataset needs both its query and target languages in
+        the requested set.
+        """
+        with caplog.at_level(logging.WARNING, logger="workrb.tasks.abstract.base"):
+            task = SearchQueryCandidateRanking(split="test", languages=[Language.DE.value])
+
+        assert task.dataset_ids == []
+        assert any(
+            "no datasets for the requested languages" in record.message for record in caplog.records
+        )
+
+    def test_warns_when_all_selected_datasets_are_skipped(self, caplog):
+        """Warn when every selected dataset raises DatasetConfigNotSupported.
+
+        The task ends up with no datasets after loading and must say so.
+        """
+
+        class AllSkippedTask(StubSupportedLanguagesRankingTask):
+            def load_dataset(self, dataset_id, split):
+                raise DatasetConfigNotSupported(f"unsupported: {dataset_id}")
+
+        with caplog.at_level(logging.WARNING, logger="workrb.tasks.abstract.base"):
+            task = AllSkippedTask(languages=[Language.EN.value], split="test")
+
+        assert task.dataset_ids == []
+        assert any(
+            "selected datasets were skipped during loading" in record.message
+            for record in caplog.records
+        )
+
+    def test_no_warning_when_datasets_load(self, caplog):
+        """A task with at least one loaded dataset must not emit either warning."""
+        with caplog.at_level(logging.WARNING, logger="workrb.tasks.abstract.base"):
+            StubSupportedLanguagesRankingTask(languages=[Language.EN.value], split="test")
+
+        assert not any(
+            "no datasets for the requested languages" in record.message
+            or "selected datasets were skipped during loading" in record.message
+            for record in caplog.records
+        )
+
+
 class TestMELORankingDatasetIds:
     """Test MELORanking.languages_to_dataset_ids filtering logic."""