Skip to content

Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.

test(webui): align page selection with persistent previews

myhloli4692886e7919
diff --git a/tests/unittest/test_pdf_page_id.py b/tests/unittest/test_pdf_page_id.py
index ebccb755..07fde7f5 100644
--- a/tests/unittest/test_pdf_page_id.py
+++ b/tests/unittest/test_pdf_page_id.py
@@ -28,6 +28,7 @@ from mineru.errors import InvalidRequestError, MineruError
from mineru.kit.gradio.page_range import effective_page_range as _effective_page_range
from mineru.kit.gradio.artifacts import persist_parse_result
from mineru.kit.gradio.client import V1ArtifactClient
+from mineru.kit.gradio.sources import prepare_source
from mineru.kit.main import app as kit_app
from mineru.parser import MinerUParser, api_server
from mineru.parser.api_client import MinerUApiParser
@@ -40,6 +41,7 @@ from mineru.parser.page_range import (
parse_page_range,
parse_page_range_set,
)
+from mineru.types import MiddleJson
VALID_RANGES = [
("5", [5], "5"),
@@ -237,7 +239,7 @@ def _numbered_pdf(path: Path, page_count: int = 5) -> None:
def test_real_pdf_selection_preserves_original_indices_and_artifact_order(tmp_path: Path) -> None:
- """实测非连续选页与末页组合,同时核对解析页号和原文预览 PDF。"""
+ """实测非连续选页与末页组合,核对产物原始页号并保留完整原文预览。"""
source = tmp_path / "numbered.pdf"
_numbered_pdf(source)
parser = MinerUParser(tier="flash")
@@ -247,11 +249,18 @@ def test_real_pdf_selection_preserves_original_indices_and_artifact_order(tmp_pa
assert "Source page 3" in result.markdown()
assert "Source page 2" not in result.markdown()
artifacts = persist_parse_result(result, source, output_root=tmp_path / "review", page_range="r1,3,1,3")
- assert artifacts.page_indices == (0, 2, 4)
- assert artifacts.origin_pdf_path is not None
- reader = PdfReader(str(artifacts.origin_pdf_path))
- assert [page.extract_text().strip() for page in reader.pages] == ["Source page 1", "Source page 3", "Source page 5"]
- assert not (artifacts.root / "layout.pdf").exists()
+ stored = MiddleJson.model_validate_json(artifacts.middle_json_path.read_text(encoding="utf-8"))
+ assert [page.page_idx for page in stored.pages] == [0, 2, 4]
+ assert artifacts.initial_page_idx == 0
+ markdown = artifacts.markdown_path.read_text(encoding="utf-8")
+ assert markdown.index("Source page 1") < markdown.index("Source page 3") < markdown.index("Source page 5")
+ assert "Source page 2" not in markdown and "Source page 4" not in markdown
+ preview = prepare_source(source, source_id="a" * 32, output_root=tmp_path / "review")
+ assert preview.preview_pdf_path is not None
+ assert preview.preview_pdf_path.read_bytes() == source.read_bytes()
+ reader = PdfReader(str(preview.preview_pdf_path))
+ assert [page.extract_text().strip() for page in reader.pages] == [f"Source page {index}" for index in range(1, 6)]
+ assert not list(artifacts.root.glob("*.pdf"))
prepared = parser._prepare_input(source, "r1")
assert prepared.retained_page_indices == [4]
assert PdfReader(io.BytesIO(prepared.file_bytes)).pages[0].extract_text().strip() == "Source page 5"