kb-skills/tests/test_processors/test_download.py

"""Tests for DownloadProcessor and DownloadInput/DownloadOutput schemas.

Test coverage:
- DownloadInput schema validation
- DownloadOutput schema validation
- DownloadProcessor metadata fetching
- DownloadProcessor content downloading
- DownloadProcessor state recording
"""

import pytest
from datetime import datetime
from unittest.mock import AsyncMock, MagicMock, patch
from pathlib import Path

from kb_workflow.pipeline.schemas import DownloadInput, DownloadOutput


class TestDownloadInputSchema:
    """Tests for DownloadInput schema validation."""

    def test_schema_download_input_validates_required_fields(self):
        """Test 1: DownloadInput validates required fields"""
        # Test with all required fields
        input_data = DownloadInput(
            content_id="BV1xx",
            platform="bilibili",
            content_dir="/tmp/content",
            content_type="video"
        )
        assert input_data.content_id == "BV1xx"
        assert input_data.platform == "bilibili"
        assert input_data.content_dir == "/tmp/content"
        assert input_data.content_type == "video"

    def test_schema_download_input_requires_content_dir(self):
        """DownloadInput should require content_dir"""
        with pytest.raises(Exception):  # ValidationError from pydantic
            DownloadInput(
                content_id="BV1xx",
                platform="bilibili",
                content_type="video"
                # Missing content_dir
            )

    def test_schema_download_input_requires_content_type(self):
        """DownloadInput should require content_type"""
        with pytest.raises(Exception):  # ValidationError from pydantic
            DownloadInput(
                content_id="BV1xx",
                platform="bilibili",
                content_dir="/tmp/content"
                # Missing content_type
            )


class TestDownloadOutputSchema:
    """Tests for DownloadOutput schema validation."""

    def test_schema_download_output_validates_required_fields(self):
        """Test 2: DownloadOutput validates required fields"""
        output = DownloadOutput(
            success=True,
            metadata_file="/tmp/content/bilibili_BV1xx_metadata.json",
            content_file="/tmp/content/bilibili_BV1xx_content.md",
            title="Test Video",
            author="Test Author"
        )
        assert output.success is True
        assert output.metadata_file == "/tmp/content/bilibili_BV1xx_metadata.json"
        assert output.content_file == "/tmp/content/bilibili_BV1xx_content.md"
        assert output.title == "Test Video"
        assert output.author == "Test Author"
        assert output.duration is None  # Optional field

    def test_schema_download_output_handles_video_content_type(self):
        """Test 3: DownloadOutput handles video content type with duration"""
        output = DownloadOutput(
            success=True,
            metadata_file="/tmp/content/bilibili_BV1xx_metadata.json",
            content_file="",  # Video content handled by ASR
            title="Test Video",
            author="Test Author",
            duration=300  # 5 minutes
        )
        assert output.success is True
        assert output.duration == 300
        assert output.content_file == ""

    def test_schema_download_output_handles_article_content_type(self):
        """Test 3: DownloadOutput handles article content type without duration"""
        output = DownloadOutput(
            success=True,
            metadata_file="/tmp/content/bilibili_CV123_metadata.json",
            content_file="/tmp/content/bilibili_CV123_content.md",
            title="Test Article",
            author="Test Author",
            duration=None  # Articles don't have duration
        )
        assert output.success is True
        assert output.duration is None
        assert output.content_file == "/tmp/content/bilibili_CV123_content.md"

    def test_schema_download_output_handles_zhihu_answer(self):
        """Test 3: DownloadOutput handles Zhihu answer content type"""
        output = DownloadOutput(
            success=True,
            metadata_file="/tmp/content/zhihu_123456_metadata.json",
            content_file="/tmp/content/zhihu_123456_content.md",
            title="Test Answer",
            author="Test Author"
        )
        assert output.success is True
        assert "zhihu" in output.metadata_file

    def test_schema_download_output_handles_failure(self):
        """DownloadOutput should handle failure case"""
        output = DownloadOutput(
            success=False,
            error_message="Network error",
            metadata_file="",
            content_file="",
            title="",
            author=""
        )
        assert output.success is False
        assert output.error_message == "Network error"


class TestDownloadProcessorMetadata:
    """Tests for DownloadProcessor metadata fetching."""

    @pytest.fixture
    def mock_bilibili_adapter(self):
        """Create a mock BilibiliAdapter."""
        adapter = MagicMock()
        adapter.fetch_metadata = AsyncMock(return_value={
            "metadata": {
                "title": "Test Video Title",
                "author": "Test UP Author",
                "mid": 12345,
                "bvid": "BV1xx",
            },
            "stats": {
                "view": 1000,
                "like": 100,
            },
            "description": "Test description"
        })
        adapter.get_output_filename = MagicMock(
            return_value="bilibili_BV1xx_metadata.json"
        )
        return adapter

    @pytest.fixture
    def mock_zhihu_adapter(self):
        """Create a mock ZhihuAdapter."""
        adapter = MagicMock()
        adapter.fetch_metadata = AsyncMock(return_value={
            "title": "Test Answer Title",
            "author": "Test Author",
            "url": "https://www.zhihu.com/question/123/answer/456",
            "content_markdown": "Test content"
        })
        adapter.fetch_content = AsyncMock(return_value="---\ntitle: Test\n---\nTest content")
        adapter.get_output_filename = MagicMock(
            return_value="zhihu_456_metadata.json"
        )
        return adapter

    @pytest.fixture
    def mock_state_manager(self):
        """Create a mock WorkflowStateManager."""
        manager = MagicMock()
        manager.load = MagicMock(return_value=None)
        manager.save = MagicMock()
        return manager

    @pytest.mark.asyncio
    async def test_metadata_fetches_bilibili_video_metadata_correctly(
        self, mock_bilibili_adapter, mock_state_manager, tmp_path
    ):
        """Test 1: DownloadProcessor fetches Bilibili video metadata correctly"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput

        input_data = DownloadInput(
            content_id="BV1xx",
            platform="bilibili",
            content_dir=str(tmp_path),
            content_type="video"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_bilibili_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is True
        assert result.title == "Test Video Title"
        assert "bilibili_BV1xx_metadata.json" in result.metadata_file

        # Verify adapter was called correctly
        mock_bilibili_adapter.fetch_metadata.assert_called_once_with(
            "BV1xx", "video"
        )

    @pytest.mark.asyncio
    async def test_metadata_fetches_zhihu_answer_metadata_correctly(
        self, mock_zhihu_adapter, mock_state_manager, tmp_path
    ):
        """Test 2: DownloadProcessor fetches Zhihu answer metadata correctly"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput

        input_data = DownloadInput(
            content_id="456",
            platform="zhihu",
            content_dir=str(tmp_path),
            content_type="answer"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_zhihu_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is True
        assert result.title == "Test Answer Title"

        # Verify adapter was called correctly
        mock_zhihu_adapter.fetch_metadata.assert_called_once_with(
            "456", "answer"
        )

    @pytest.mark.asyncio
    async def test_metadata_records_state_on_success(
        self, mock_bilibili_adapter, mock_state_manager, tmp_path
    ):
        """Test 3: DownloadProcessor records state on success"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput, WorkflowState

        # Create a mock state that will be updated
        mock_state = MagicMock(spec=WorkflowState)
        mock_state.is_step_completed = MagicMock(return_value=False)
        mock_state_manager.load.return_value = mock_state

        input_data = DownloadInput(
            content_id="BV1xx",
            platform="bilibili",
            content_dir=str(tmp_path),
            content_type="video"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_bilibili_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is True

        # Verify state was recorded
        mock_state.mark_step_completed.assert_called_once()
        call_args = mock_state.mark_step_completed.call_args
        assert call_args[0][0] == "download"  # step_name
        mock_state_manager.save.assert_called_once_with(mock_state)

    @pytest.mark.asyncio
    async def test_metadata_records_state_on_failure(
        self, mock_bilibili_adapter, mock_state_manager, tmp_path
    ):
        """Test 4: DownloadProcessor records state on failure"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput, WorkflowState
        from kb_workflow.pipeline.errors import NetworkError

        # Make adapter raise an error
        mock_bilibili_adapter.fetch_metadata = AsyncMock(
            side_effect=NetworkError(
                step_name="download",
                input_summary="BV1xx",
                status_code=503,
                cause=Exception("Network timeout")
            )
        )

        # Create a mock state that will be updated
        mock_state = MagicMock(spec=WorkflowState)
        mock_state.is_step_completed = MagicMock(return_value=False)
        mock_state_manager.load.return_value = mock_state

        input_data = DownloadInput(
            content_id="BV1xx",
            platform="bilibili",
            content_dir=str(tmp_path),
            content_type="video"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_bilibili_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is False
        assert "503" in result.error_message  # NetworkError message includes status code

        # Verify failure was recorded
        mock_state.mark_step_failed.assert_called_once()
        call_args = mock_state.mark_step_failed.call_args
        assert call_args[0][0] == "download"  # step_name
        mock_state_manager.save.assert_called_once_with(mock_state)


class TestDownloadProcessorContent:
    """Tests for DownloadProcessor content downloading."""

    @pytest.fixture
    def mock_bilibili_article_adapter(self):
        """Create a mock BilibiliAdapter for articles."""
        adapter = MagicMock()
        adapter.fetch_metadata = AsyncMock(return_value={
            "title": "Test Article Title",
            "author": "Test Author",
            "cvid": "CV123",
            "url": "https://www.bilibili.com/read/cv123",
        })
        adapter.fetch_content = AsyncMock(return_value="---\ntitle: Test Article\n---\nArticle content here")
        adapter.get_output_filename = MagicMock(side_effect=lambda cid, ctype, suffix: f"bilibili_{cid}_{suffix}")
        return adapter

    @pytest.fixture
    def mock_bilibili_video_adapter(self):
        """Create a mock BilibiliAdapter for videos."""
        adapter = MagicMock()
        adapter.fetch_metadata = AsyncMock(return_value={
            "metadata": {
                "title": "Test Video Title",
                "author": "Test Video Author",
                "mid": 12345,
                "bvid": "BV1xx",
                "duration": 300,
            },
            "stats": {"view": 1000},
        })
        adapter.get_output_filename = MagicMock(side_effect=lambda cid, ctype, suffix: f"bilibili_{cid}_{suffix}")
        return adapter

    @pytest.fixture
    def mock_state_manager(self):
        """Create a mock WorkflowStateManager."""
        manager = MagicMock()
        manager.load = MagicMock(return_value=None)
        manager.save = MagicMock()
        return manager

    @pytest.mark.asyncio
    async def test_content_downloads_article_content(
        self, mock_bilibili_article_adapter, mock_state_manager, tmp_path
    ):
        """Test 1: DownloadProcessor downloads Bilibili article content"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput

        input_data = DownloadInput(
            content_id="CV123",
            platform="bilibili",
            content_dir=str(tmp_path),
            content_type="article"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_bilibili_article_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is True
        assert result.content_file != ""  # Content should be downloaded
        assert "bilibili_CV123_content.md" in result.content_file

        # Verify content file was created
        content_path = Path(result.content_file)
        assert content_path.exists()

        # Verify fetch_content was called
        mock_bilibili_article_adapter.fetch_content.assert_called_once_with("CV123", "article")

    @pytest.mark.asyncio
    async def test_content_skips_video_download(
        self, mock_bilibili_video_adapter, mock_state_manager, tmp_path
    ):
        """Test 2: DownloadProcessor skips video download (handled by ASR)"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput

        input_data = DownloadInput(
            content_id="BV1xx",
            platform="bilibili",
            content_dir=str(tmp_path),
            content_type="video"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_bilibili_video_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is True
        assert result.content_file == ""  # Video content not downloaded (handled by ASR)
        assert result.duration == 300  # Duration should be extracted

        # Verify fetch_content was NOT called for video
        mock_bilibili_video_adapter.fetch_content.assert_not_called()

    @pytest.mark.asyncio
    async def test_content_creates_correct_output_file_paths(
        self, mock_bilibili_article_adapter, mock_state_manager, tmp_path
    ):
        """Test 3: DownloadProcessor creates correct output file paths"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor
        from kb_workflow.pipeline.schemas import DownloadInput

        input_data = DownloadInput(
            content_id="CV123",
            platform="bilibili",
            content_dir=str(tmp_path),
            content_type="article"
        )

        with patch("kb_workflow.pipeline.processors.base.get_settings") as mock_settings:
            mock_settings.return_value = MagicMock()
            with patch.object(
                DownloadProcessor, '_get_adapter', return_value=mock_bilibili_article_adapter
            ):
                processor = DownloadProcessor(state_manager=mock_state_manager)
                result = await processor.process(input_data)

        assert result.success is True

        # Check file paths follow naming convention
        assert "bilibili_CV123_metadata.json" in result.metadata_file
        assert "bilibili_CV123_content.md" in result.content_file

        # Check files exist
        assert Path(result.metadata_file).exists()
        assert Path(result.content_file).exists()


class TestDownloadProcessorModuleExport:
    """Tests for DownloadProcessor module export."""

    def test_download_processor_importable_from_processors(self):
        """Test 4: DownloadProcessor can be imported from kb_workflow.pipeline.processors"""
        from kb_workflow.pipeline.processors import DownloadProcessor

        assert DownloadProcessor is not None
        assert hasattr(DownloadProcessor, 'STEP_NAME')
        assert DownloadProcessor.STEP_NAME == "download"

    def test_all_processors_exported(self):
        """All processors should be exported from kb_workflow.pipeline.processors"""
        from kb_workflow.pipeline.processors import BaseProcessor, ParseProcessor, DownloadProcessor, ASRProcessor

        assert BaseProcessor is not None
        assert ParseProcessor is not None
        assert DownloadProcessor is not None
        assert ASRProcessor is not None


class TestDownloadProcessorAuthorExtraction:
    """Tests for _extract_author method - ensure it returns author name, not mid."""

    def test_extract_author_returns_author_name_not_mid(self):
        """_extract_author should return author name, not mid (user ID)"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor

        processor = DownloadProcessor()

        # Bilibili metadata with both author and mid
        metadata = {
            "metadata": {
                "title": "Test Video",
                "author": "HOLA小测佬",
                "mid": 406636263,
                "bvid": "BV1QXcCzfECp",
            }
        }

        author = processor._extract_author(metadata, "bilibili")

        # Should return author name, not mid
        assert author == "HOLA小测佬"
        assert author != "406636263"
        assert author != str(metadata["metadata"]["mid"])

    def test_extract_author_handles_nested_metadata(self):
        """_extract_author should handle nested metadata structure"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor

        processor = DownloadProcessor()

        # Nested metadata structure
        metadata = {
            "metadata": {
                "author": "Test Author Name",
                "mid": 12345,
            }
        }

        author = processor._extract_author(metadata, "bilibili")
        assert author == "Test Author Name"

    def test_extract_author_handles_flat_metadata(self):
        """_extract_author should handle flat metadata structure"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor

        processor = DownloadProcessor()

        # Flat metadata structure (no nested 'metadata' key)
        metadata = {
            "author": "Flat Author",
            "mid": 67890,
        }

        author = processor._extract_author(metadata, "bilibili")
        assert author == "Flat Author"

    def test_extract_author_handles_zhihu_platform(self):
        """_extract_author should handle zhihu platform"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor

        processor = DownloadProcessor()

        # Zhihu metadata structure
        metadata = {
            "author": "Zhihu User",
        }

        author = processor._extract_author(metadata, "zhihu")
        assert author == "Zhihu User"

    def test_extract_author_returns_empty_string_for_missing_author(self):
        """_extract_author should return empty string when author is missing"""
        from kb_workflow.pipeline.processors.download import DownloadProcessor

        processor = DownloadProcessor()

        # Metadata without author
        metadata = {
            "metadata": {
                "title": "Test Video",
                "mid": 12345,
            }
        }

        author = processor._extract_author(metadata, "bilibili")
        assert author == ""