Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
Python ai4sciencedocument-analysisdocxextract-datalayout-analysisocrparserpdfpdf-converterpdf-extractor-llmpdf-extractor-pretrainpdf-extractor-ragpdf-parserpptxpythonxlsx
from __future__ import annotations
from pathlib import Path
import pytest
from pydantic import ValidationError
from mineru.config import Config, _collect_env_overrides, _load_config
def test_render_config_defaults() -> None:
"""验证 config.yaml 缺省时使用美元公式定界符。"""
delimiters = Config().render.latex_delimiters
assert delimiters.display.left == "$$"
assert delimiters.display.right == "$$"
assert delimiters.inline.left == "$"
assert delimiters.inline.right == "$"
def test_render_config_reads_yaml_shape(tmp_path: Path) -> None:
"""验证 render.latex_delimiters YAML 结构能进入 typed Config。"""
config_file = tmp_path / "config.yaml"
config_file.write_text(
"""
render:
latex_delimiters:
display:
left: '\\['
right: '\\]'
inline:
left: '\\('
right: '\\)'
""".strip(),
encoding="utf-8",
)
config = Config(**_load_config(str(config_file)))
assert config.render.latex_delimiters.display.left == r"\["
assert config.render.latex_delimiters.inline.right == r"\)"
def test_render_config_supports_environment_overrides(monkeypatch: pytest.MonkeyPatch) -> None:
"""验证现有环境变量递归覆盖机制可定位 render delimiter。"""
prefix = "TEST_MINERU_"
monkeypatch.setenv(f"{prefix}RENDER_LATEX_DELIMITERS_INLINE_LEFT", r"\(")
monkeypatch.setenv(f"{prefix}RENDER_LATEX_DELIMITERS_INLINE_RIGHT", r"\)")
overrides, _paths = _collect_env_overrides(prefix=prefix)
config = Config(**overrides)
assert config.render.latex_delimiters.inline.left == r"\("
assert config.render.latex_delimiters.inline.right == r"\)"
@pytest.mark.parametrize("side", ["left", "right"])
def test_render_config_rejects_empty_delimiters(side: str) -> None:
"""验证任一空定界符都会在 typed config 边界报错。"""
pair = {"left": "$", "right": "$"}
pair[side] = ""
with pytest.raises(ValidationError):
Config(render={"latex_delimiters": {"inline": pair}})
def test_render_config_does_not_read_legacy_root_key() -> None:
"""验证旧根级 latex-delimiter-config 不影响新 render 配置。"""
config = Config.model_validate(
{
"latex-delimiter-config": {
"inline": {"left": "OLD", "right": "OLD"},
}
}
)
assert config.render.latex_delimiters.inline.left == "$"