Haipeng Liu

"""
Integration tests for the base Extractor class.
"""
import pytest
import hashlib
import math
from unittest.mock import Mock, patch, MagicMock
from collections import Counter

pytestmark = [pytest.mark.integration]


class TestExtractorStaticMethods:
    """Tests for static methods in Extractor class."""

    def test_process_binary_string_printable(self):
        """Test processing a printable binary string."""
        from redb.extractors.extractor import Extractor

        # Test with printable ASCII characters
        input_bytes = b"Hello World\x00\x00\x00"
        result = Extractor.process_binary_string(input_bytes)
        assert result == "Hello World"

    def test_process_binary_string_non_printable(self):
        """Test processing a string with non-printable characters."""
        from redb.extractors.extractor import Extractor

        # Test with non-printable characters
        input_bytes = b"\x01\x02Hello\x00"
        result = Extractor.process_binary_string(input_bytes)
        assert "\\x01" in result
        assert "\\x02" in result
        assert "Hello" in result

    def test_process_binary_string_empty(self):
        """Test processing an empty binary string."""
        from redb.extractors.extractor import Extractor

        input_bytes = b"\x00\x00\x00"
        result = Extractor.process_binary_string(input_bytes)
        assert result == ""

    def test_process_binary_string_all_printable(self):
        """Test processing a fully printable string."""
        from redb.extractors.extractor import Extractor

        input_bytes = b"TestString123"
        result = Extractor.process_binary_string(input_bytes)
        assert result == "TestString123"

    def test_remove_non_utf8_valid(self):
        """Test removing non-UTF8 characters from valid UTF8 string."""
        from redb.extractors.extractor import Extractor

        input_bytes = b"Hello World"
        result = Extractor.remove_non_utf8(input_bytes)
        assert result == b"Hello World"

    def test_remove_non_utf8_with_invalid(self):
        """Test removing non-UTF8 characters from mixed string."""
        from redb.extractors.extractor import Extractor

        # Mix of valid UTF8 and invalid bytes
        input_bytes = b"Hello\xff\xfeWorld"
        result = Extractor.remove_non_utf8(input_bytes)
        # Should remove the invalid bytes
        assert b"\xff" not in result
        assert b"\xfe" not in result
        assert b"Hello" in result
        assert b"World" in result

    def test_remove_non_utf8_empty(self):
        """Test removing non-UTF8 from empty string."""
        from redb.extractors.extractor import Extractor

        result = Extractor.remove_non_utf8(b"")
        assert result == b""


class TestExtractorEntropyCalculation:
    """Tests for entropy calculation in Extractor class."""

    def test_calculate_entropy_empty_data(self, elf_binary_path, mock_logger):
        """Test entropy calculation with empty data."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)
            result = extractor.calculate_entropy(b"")
            assert result == 0.0

    def test_calculate_entropy_uniform_data(self, elf_binary_path, mock_logger):
        """Test entropy calculation with uniform (low entropy) data."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)
            # All same bytes = 0 entropy
            uniform_data = b"\x00" * 256
            result = extractor.calculate_entropy(uniform_data)
            assert result == 0.0

    def test_calculate_entropy_random_data(self, elf_binary_path, mock_logger, high_entropy_data):
        """Test entropy calculation with random (high entropy) data."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)
            result = extractor.calculate_entropy(high_entropy_data)
            # Random data should have entropy close to 8 (max for bytes)
            assert result > 7.0

    def test_calculate_entropy_binary_data(self, elf_binary_path, mock_logger):
        """Test entropy calculation with binary data."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)
            # Two equally distributed values = 1 bit entropy
            binary_data = b"\x00\x01" * 128
            result = extractor.calculate_entropy(binary_data)
            assert abs(result - 1.0) < 0.01

    def test_calculate_entropy_string_data(self, elf_binary_path, mock_logger):
        """Test entropy calculation with string data."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)
            # String with two equally distributed characters
            string_data = "ab" * 100
            result = extractor.calculate_entropy(string_data)
            assert abs(result - 1.0) < 0.01


class TestExtractorInitialization:
    """Tests for Extractor initialization."""

    def test_extractor_hashes_computed_on_init(self, elf_binary_path, mock_logger, known_hashes):
        """Test that hashes are computed during initialization."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)

            assert extractor.md5 == known_hashes['md5']
            assert extractor.sha1 == known_hashes['sha1']
            assert extractor.sha256 == known_hashes['sha256']

    def test_extractor_hash_dataclass_created(self, elf_binary_path, mock_logger, known_hashes):
        """Test that Hash dataclass is created during initialization."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)

            assert extractor.hash.md5 == known_hashes['md5']
            assert extractor.hash.sha1 == known_hashes['sha1']
            assert extractor.hash.sha256 == known_hashes['sha256']

    def test_extractor_binary_cached_property(self, elf_binary_path, mock_logger, elf_binary_content):
        """Test that binary content is loaded as cached property."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)

            # Access binary property
            binary = extractor.binary
            assert binary == elf_binary_content

            # Second access should return cached value (same id)
            assert extractor.binary is binary

    def test_extractor_with_exporters(self, elf_binary_path, mock_logger, mock_print_exporter):
        """Test extractor initialization with exporters."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(
                elf_binary_path,
                mock_logger,
                exporters=[mock_print_exporter]
            )

            assert len(extractor.exporters) == 1
            assert extractor.exporters[0] is mock_print_exporter

    def test_extractor_with_index_prefix(self, elf_binary_path, mock_logger):
        """Test extractor initialization with custom index prefix."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(
                elf_binary_path,
                mock_logger,
                index_prefix="custom_prefix"
            )

            assert "custom_prefix" in extractor.elastic_index

    def test_extractor_known_benign_flag(self, elf_binary_path, mock_logger):
        """Test extractor with known_benign flag."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(
                elf_binary_path,
                mock_logger,
                known_benign=True
            )

            assert extractor.known_benign is True
            assert extractor.known_malicious is False

    def test_extractor_known_malicious_flag(self, elf_binary_path, mock_logger):
        """Test extractor with known_malicious flag."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(
                elf_binary_path,
                mock_logger,
                known_malicious=True
            )

            assert extractor.known_benign is False
            assert extractor.known_malicious is True


class TestExtractorExportData:
    """Tests for export_data method."""

    def test_export_data_with_print_exporter(self, elf_binary_path, mock_logger, mock_print_exporter, capsys):
        """Test export_data with PrintExporter."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(
                elf_binary_path,
                mock_logger,
                exporters=[mock_print_exporter]
            )

            result = extractor.export_data()
            # Should succeed with PrintExporter
            assert result is True

    def test_export_data_no_exporters(self, elf_binary_path, mock_logger):
        """Test export_data with no exporters configured."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(
                elf_binary_path,
                mock_logger,
                exporters=[]
            )

            result = extractor.export_data()
            # Should succeed with no exporters (nothing to export)
            assert result is True


class TestExtractorEdgeCases:
    """Tests for edge cases in Extractor class."""

    def test_extractor_with_nonexistent_file(self, mock_logger):
        """Test extractor with non-existent file."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            with pytest.raises(FileNotFoundError):
                BasicPropertiesExtractor("/nonexistent/file.bin", mock_logger)

    def test_entropy_none_data(self, elf_binary_path, mock_logger):
        """Test entropy calculation with None returns 0."""
        from redb.extractors.basicproperties import BasicPropertiesExtractor

        with patch('redb.settings.ELASTIC_BINARIES_COLLECTION', 'test'):
            extractor = BasicPropertiesExtractor(elf_binary_path, mock_logger)
            result = extractor.calculate_entropy(None)
            assert result == 0.0

    def test_process_binary_string_high_bytes(self):
        """Test processing binary string with high byte values."""
        from redb.extractors.extractor import Extractor

        # Characters above 126 are non-printable
        input_bytes = b"\x7f\x80\x81\xff"
        result = Extractor.process_binary_string(input_bytes)
        assert "\\x7f" in result
        assert "\\x80" in result
        assert "\\xff" in result