Files
docling-core/tests/test_groundtruth_diff.py
2026-11-03 16:13:54 +01:00

164 lines
6.4 KiB
Python

# SPDX-FileCopyrightText: The Docling Contributors
# SPDX-License-Identifier: MIT
from __future__ import annotations
import copy
import importlib.util
import json
import sys
from pathlib import Path
from typing import Any
SCRIPTS_DIR = Path(__file__).resolve().parents[1] / ".github/scripts"
sys.path.insert(0, str(SCRIPTS_DIR))
SPEC = importlib.util.spec_from_file_location("groundtruth_diff", SCRIPTS_DIR / "groundtruth_diff.py")
assert SPEC is not None and SPEC.loader is not None
gt = importlib.util.module_from_spec(SPEC)
sys.modules[SPEC.name] = gt
SPEC.loader.exec_module(gt)
DOC_PATH = "tests/data/doc/sample.json"
def make_doc() -> dict[str, Any]:
prov = [{"page_no": 1, "bbox": {"l": 1, "t": 2, "r": 3, "b": 4}}]
return {
"body": {"children": [{"$ref": "#/texts/0"}, {"$ref": "#/tables/0"}]},
"texts": [
{"label": "section_header", "text": "Intro", "prov": prov, "children": []},
{"label": "text", "text": "Hello world", "prov": prov, "children": []},
],
"tables": [
{
"label": "table",
"prov": prov,
"children": [{"$ref": "#/texts/1"}],
"data": {
"num_rows": 1,
"num_cols": 2,
"table_cells": [{"text": "a"}, {"text": "b"}],
},
}
],
"pages": {"1": {"size": {"width": 10, "height": 10}}},
}
def summarize(old: dict[str, Any], new: dict[str, Any]) -> Any:
return gt.summarize_file(
DOC_PATH,
json.dumps(old).encode(),
json.dumps(new, indent=2).encode(),
status="modified",
)
def test_test_data_counts_except_source_code() -> None:
assert gt.is_groundtruth_path("tests/data/doc/2408.09869v3_enriched.gt.md")
assert gt.is_groundtruth_path("tests/data/doc/virtual_text_list/input.json")
assert gt.is_groundtruth_path("tests/data/chunker/0_out_chunks.json")
assert gt.is_groundtruth_path("tests/data/viz/2408.09869v3_enriched_viz_p1.png")
# Source code stays in the model's diff: pytest collects tests/ recursively.
assert not gt.is_groundtruth_path("tests/data/chunker_repo/repos/docling/export.py")
assert not gt.is_groundtruth_path("tests/data/repo_chunking/sample.ts")
assert not gt.is_groundtruth_path("tests/test_utils.py")
assert not gt.is_groundtruth_path("docs/data/a.json")
def test_images_are_compared_by_bytes() -> None:
path = "tests/data/viz/a.png"
png = b"\x89PNG\r\n\x1a\n\x00\xff"
assert gt.summarize_file(path, png, png, status="modified").kind == gt.ChangeKind.FORMATTING
change = gt.summarize_file(path, png, png + b"\x00", status="modified")
assert change.kind == gt.ChangeKind.BINARY
assert change.details == ["10 → 11 bytes"]
def test_reformatted_json_and_moved_boxes_are_not_content_changes() -> None:
old = make_doc()
assert summarize(old, copy.deepcopy(old)).kind == gt.ChangeKind.FORMATTING
moved = copy.deepcopy(old)
moved["texts"][0]["prov"][0]["bbox"]["l"] = 5
moved["pages"]["1"]["size"]["width"] = 11
assert summarize(old, moved).kind == gt.ChangeKind.COORDINATES
def test_text_table_and_reading_order_changes_are_classified() -> None:
old = make_doc()
text = copy.deepcopy(old)
text["texts"][1]["text"] = "Hello there"
change = summarize(old, text)
assert change.kind == gt.ChangeKind.TEXT
assert change.details == ["1 text item(s) changed"]
cells = copy.deepcopy(old)
cells["tables"][0]["data"]["table_cells"][1]["text"] = "c"
assert summarize(old, cells).details == ["table 0: 1 cell(s) changed"]
order = copy.deepcopy(old)
order["body"]["children"].reverse()
change = summarize(old, order)
assert change.kind == gt.ChangeKind.STRUCTURE
assert "reading order changed" in change.details
relabel = copy.deepcopy(old)
relabel["texts"][0]["label"] = "title"
change = summarize(old, relabel)
assert change.kind == gt.ChangeKind.STRUCTURE
assert "section_header: 1 → 0" in change.details
def test_markdown_whitespace_and_table_rows() -> None:
path = "tests/data/doc/a.gt.md"
old = b"# T\n\n| a | b |\n|---|---|\n| 1 | 2 |\n"
formatted = gt.summarize_file(path, old, old + b"\n\n", status="modified")
assert formatted.kind == gt.ChangeKind.FORMATTING
joined = gt.summarize_file(path, b"one\n\ntwo\n", b"one two\n", status="modified")
assert joined.kind == gt.ChangeKind.TEXT
assert joined.details == ["line or paragraph breaks changed"]
change = gt.summarize_file(path, old, old.replace(b"| 1 |", b"| 9 |"), status="modified")
assert change.kind == gt.ChangeKind.TABLES
assert change.details == ["+1 / -1 lines", "Markdown table rows changed"]
def test_a_failed_read_of_a_modified_file_is_not_reported_as_added() -> None:
# read_blob gives None for any git failure, for example a blob that the
# partial clone cannot fetch.
doc = json.dumps(make_doc()).encode()
for old, new, side in ((None, doc, "base"), (doc, None, "head")):
change = gt.summarize_file(DOC_PATH, old, new, status="modified")
assert change.kind == gt.ChangeKind.UNPARSABLE
assert change.details == [f"cannot read the {side} version"]
added = gt.summarize_file(DOC_PATH, None, doc, status="added")
assert added.kind == gt.ChangeKind.ADDED
renamed = gt.summarize_file(DOC_PATH, None, doc, status="renamed")
assert renamed.kind == gt.ChangeKind.UNPARSABLE
def test_render_lists_relevant_kinds_first() -> None:
changes = [
gt.FileChange("f1.json", gt.ChangeKind.FORMATTING),
gt.FileChange("f2.json", gt.ChangeKind.STRUCTURE, ["reading order changed"]),
]
lines = gt.render_markdown(changes).splitlines()
assert lines[2] == "| structure | 1 |"
assert lines[5] == "- **structure** `f2.json`: reading order changed"
def test_render_markdown_escapes_data_from_the_pr() -> None:
doc = make_doc()
changed = copy.deepcopy(doc)
changed["texts"][0]["label"] = "x\n\n**safe** @org/team ![i](http://evil/p.png)"
change = gt.summarize_file(DOC_PATH, json.dumps(doc).encode(), json.dumps(changed).encode(), status="modified")
path = "tests/data/a` @org/team `.md"
rendered = gt.render_markdown([change, gt.FileChange(path, gt.ChangeKind.ADDED)]).splitlines()
assert len(rendered) == 7
detail, added = rendered[5], rendered[6]
assert "**safe**" not in detail
assert "![i](" not in detail
assert "@org" not in detail
assert added == f"- **added** ``{path}``"