import pytest
from src.ingestion.transform.image_captioner import ImageCaptioner
from unittest.mock import MagicMock
class TestImageCaptionerLanguageDetection:
@pytest.fixture
def captioner(self):
settings = MagicMock()
settings.ingestion.transform.image_captioner.enabled = True
settings.ingestion.transform.image_captioner.prompt_path = None
return ImageCaptioner(settings)
def test_empty_text_defaults_to_chinese(self, captioner):
assert captioner._detect_language("") == "zh"
assert captioner._detect_language(None) == "zh"
def test_only_image_markdown_defaults_to_chinese(self, captioner):
text = ""
assert captioner._detect_language(text) == "zh"
def test_chinese_content_returns_zh(self, captioner):
text = "这是一个测试。"
assert captioner._detect_language(text) == "zh"
text = "Mixed content with 中文 characters."
assert captioner._detect_language(text) == "zh"
def test_short_english_defaults_to_chinese(self, captioner):
# "Table 1" has 6 chars -> 5 letters. Threshold is > 10.
text = "Table 1"
assert captioner._detect_language(text) == "zh"
text = "12345"
assert captioner._detect_language(text) == "zh"
def test_long_english_returns_en(self, captioner):
text = "This is a sufficiently long English sentence to be detected."
assert captioner._detect_language(text) == "en"
# "Table 1: Sales Data for 2023" -> ~20 english chars
text = "Table 1: Sales Data for 2023"
assert captioner._detect_language(text) == "en"
def test_image_markdown_ignored_in_detection(self, captioner):
# The image path has many english letters, but should be ignored.
# ""
# If not ignored, this might trigger 'en'.
# But since it's stripped, the remaining text is empty -> 'zh'.
text = ""
assert captioner._detect_language(text) == "zh"
