feat(chat): open cited documents at the cited passage (#3700)
* feat(chat): open cited documents at the cited passage
Clicking a citation used to show only the parsed chunk. It now opens the
original file in the references panel, jumps to the cited place and
highlights it, narrowed to the sentence the citation supports.
Parsing records where each chunk sits in the original file
(chunks.source_locators): PDF page and paragraph boxes from the docreader,
MinerU and PaddleOCR-VL layouts; Word paragraphs, slides, sheet rows and
EPUB chapters aligned in Go against any parser engine's Markdown; text
offsets for Markdown/TXT; segment times for audio. Pipeline text rewrites
are followed with a line-diff remapper. Scanned PDF pages get their OCR
text as quotes at search time so a sentence can pick its page.
The frontend renders PDFs with pdf.js and every other previewable type
with the existing preview, then highlights with the CSS Custom Highlight
API in a highlighter yellow that stays visible against the brand green.
* fix(docreader): serialize source blocks outside main so tests import it
test_docx_tables stubs docreader.parser in sys.modules, so importing
docreader.main from a later test failed in the full unittest discover run.
The wire conversion now lives in docreader.source_wire, which needs only
the proto.
* fix(chat): tighten source locators from review
- Remap rewritten-line-for-line stretches by line so scanned PDF pages
no longer drift onto the wrong page.
- Keep the source preview mounted for citations in the file already open.
- Wait for in-flight PDF page renders before looking for the text layer.
- Count image rune offsets in one pass; stream xlsx rows up to maxUnits.
- Start text locators after a UTF-8 BOM, as viewers decode without it.
- Only re-query image children for scanned-page results.
- Seek audio citations without autoplay and drop stale seek listeners.
- Fix the SQLite migration's cross-reference to versioned/000113.
* fix(migrations): renumber chunk source locators after main's chunk_images
main took versioned 000113 and sqlite 000032 for chunk_images, so the
source_locators column moves to versioned 000114 and sqlite 000033.
2026-09-25 05:26:16 +00:00
|
|
|
import json
|
|
|
|
|
import unittest
|
|
|
|
|
|
|
|
|
|
from docreader.models.document import Document
|
|
|
|
|
from docreader.parser.source_locator import (
|
|
|
|
|
PageGeometry,
|
|
|
|
|
locate_lines,
|
|
|
|
|
page_blocks,
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def _ln(text, x0, y0, x1, y1):
|
|
|
|
|
return {"text": text, "bbox": (x0, y0, x1, y1)}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class LocateLinesTest(unittest.TestCase):
|
|
|
|
|
def test_finds_lines_in_order_and_pulls_in_leading_markup(self):
|
|
|
|
|
text = "## 标题\n【版权声明】\n正文 第一行\n正文 第一行"
|
|
|
|
|
hits = locate_lines(text, ["标题", "版权声明】", "正文第一行", "正文 第一行", "无此行"])
|
|
|
|
|
self.assertEqual(hits[0], 0)
|
|
|
|
|
self.assertEqual(text[hits[1]:hits[1] + 2], "【版")
|
|
|
|
|
self.assertEqual(hits[2], text.index("正文"))
|
|
|
|
|
self.assertEqual(hits[3], text.rindex("正文"))
|
|
|
|
|
self.assertIsNone(hits[4])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class PageGeometryTest(unittest.TestCase):
|
|
|
|
|
def test_flips_to_top_left_origin(self):
|
|
|
|
|
g = PageGeometry((0, 0, 200, 100))
|
|
|
|
|
self.assertEqual(g.bbox((20, 80, 100, 90)), [0.1, 0.1, 0.5, 0.2])
|
|
|
|
|
|
|
|
|
|
def test_applies_cropbox_offset_and_rotation(self):
|
|
|
|
|
g = PageGeometry((100, 100, 300, 200), rotation=90)
|
|
|
|
|
# Unrotated [0.1, 0.1, 0.5, 0.2] turns clockwise.
|
|
|
|
|
self.assertEqual(g.bbox((120, 180, 200, 190)), [0.8, 0.1, 0.9, 0.5])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class PageBlocksTest(unittest.TestCase):
|
|
|
|
|
def test_groups_lines_into_boxed_paragraphs(self):
|
|
|
|
|
page = (
|
|
|
|
|
"8.14.11 索夹与吊索施工应符合下列规定:\n"
|
|
|
|
|
"1 在满足施工需要的前提下,应减小猫道面层开孔面积,并应在开孔位置四周绑\n"
|
|
|
|
|
"扎防滑木条,设立警示标志。\n"
|
|
|
|
|
"2 索夹在主缆上定位后,应紧固螺栓。"
|
|
|
|
|
)
|
|
|
|
|
lines = [
|
|
|
|
|
_ln("8.14.11 索夹与吊索施工应符合下列规定:", 50, 700, 300, 712),
|
|
|
|
|
_ln("1 在满足施工需要的前提下,应减小猫道面层开孔面积,并应在开孔位置四周绑", 60, 684, 540, 696),
|
|
|
|
|
_ln("扎防滑木条,设立警示标志。", 50, 668, 200, 680),
|
|
|
|
|
_ln("2 索夹在主缆上定位后,应紧固螺栓。", 60, 652, 400, 664),
|
|
|
|
|
]
|
|
|
|
|
blocks = page_blocks(page, 10, 57, lines, PageGeometry((0, 0, 600, 800)))
|
|
|
|
|
self.assertEqual(len(blocks), 3)
|
|
|
|
|
self.assertEqual(blocks[0]["start"], 10)
|
|
|
|
|
item1 = blocks[1]
|
|
|
|
|
self.assertEqual(page[item1["start"] - 10:item1["end"] - 10].split("\n")[0][:1], "1")
|
|
|
|
|
self.assertIn("设立警示标志。", page[item1["start"] - 10:item1["end"] - 10])
|
|
|
|
|
self.assertEqual(item1["locator"]["page"], 57)
|
|
|
|
|
x0, y0, x1, y1 = item1["locator"]["bbox"]
|
|
|
|
|
self.assertAlmostEqual(y0, 1 - 696 / 800, places=3)
|
|
|
|
|
self.assertAlmostEqual(y1, 1 - 668 / 800, places=3)
|
|
|
|
|
self.assertEqual(blocks[-1]["end"], 10 + len(page))
|
|
|
|
|
|
|
|
|
|
def test_page_without_lines_gets_page_block(self):
|
|
|
|
|
blocks = page_blocks("文字", 5, 2, [], None)
|
2026-09-27 12:32:26 +00:00
|
|
|
self.assertEqual(blocks, [{"start": 5, "end": 7, "locator": {"type": "pdf", "page": 2, "mapping": "exact"}}])
|
feat(chat): open cited documents at the cited passage (#3700)
* feat(chat): open cited documents at the cited passage
Clicking a citation used to show only the parsed chunk. It now opens the
original file in the references panel, jumps to the cited place and
highlights it, narrowed to the sentence the citation supports.
Parsing records where each chunk sits in the original file
(chunks.source_locators): PDF page and paragraph boxes from the docreader,
MinerU and PaddleOCR-VL layouts; Word paragraphs, slides, sheet rows and
EPUB chapters aligned in Go against any parser engine's Markdown; text
offsets for Markdown/TXT; segment times for audio. Pipeline text rewrites
are followed with a line-diff remapper. Scanned PDF pages get their OCR
text as quotes at search time so a sentence can pick its page.
The frontend renders PDFs with pdf.js and every other previewable type
with the existing preview, then highlights with the CSS Custom Highlight
API in a highlighter yellow that stays visible against the brand green.
* fix(docreader): serialize source blocks outside main so tests import it
test_docx_tables stubs docreader.parser in sys.modules, so importing
docreader.main from a later test failed in the full unittest discover run.
The wire conversion now lives in docreader.source_wire, which needs only
the proto.
* fix(chat): tighten source locators from review
- Remap rewritten-line-for-line stretches by line so scanned PDF pages
no longer drift onto the wrong page.
- Keep the source preview mounted for citations in the file already open.
- Wait for in-flight PDF page renders before looking for the text layer.
- Count image rune offsets in one pass; stream xlsx rows up to maxUnits.
- Start text locators after a UTF-8 BOM, as viewers decode without it.
- Only re-query image children for scanned-page results.
- Seek audio citations without autoplay and drop stale seek listeners.
- Fix the SQLite migration's cross-reference to versioned/000113.
* fix(migrations): renumber chunk source locators after main's chunk_images
main took versioned 000113 and sqlite 000032 for chunk_images, so the
source_locators column moves to versioned 000114 and sqlite 000033.
2026-09-25 05:26:16 +00:00
|
|
|
|
|
|
|
|
|
|
|
|
|
class WireTest(unittest.TestCase):
|
|
|
|
|
def test_source_blocks_serialize_and_skip_malformed(self):
|
|
|
|
|
from docreader.source_wire import source_blocks_to_proto
|
|
|
|
|
|
|
|
|
|
doc = Document(
|
|
|
|
|
content="abc",
|
|
|
|
|
source_blocks=[
|
|
|
|
|
{"start": 0, "end": 3, "locator": {"type": "pdf", "page": 1}},
|
|
|
|
|
{"start": 2, "end": 2, "locator": {"type": "pdf", "page": 1}},
|
|
|
|
|
{"start": 0, "end": 1},
|
|
|
|
|
],
|
|
|
|
|
)
|
|
|
|
|
wire = source_blocks_to_proto(doc)
|
|
|
|
|
self.assertEqual(len(wire), 1)
|
|
|
|
|
self.assertEqual(json.loads(wire[0].locator_json), {"type": "pdf", "page": 1})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class LocateColumnsTest(unittest.TestCase):
|
|
|
|
|
def test_repeated_line_follows_its_own_column(self):
|
|
|
|
|
text = (
|
|
|
|
|
"alpha left first line\nsame tail phrase here\n"
|
|
|
|
|
"bravo left third line\nsame tail phrase here\n"
|
|
|
|
|
"omega right first line\nsigma right second line"
|
|
|
|
|
)
|
|
|
|
|
left, right = (0, 100), (120, 220)
|
|
|
|
|
lines = [
|
|
|
|
|
("alpha left first line", left, 700),
|
|
|
|
|
("omega right first line", right, 700),
|
|
|
|
|
("same tail phrase here", left, 680),
|
|
|
|
|
("sigma right second line", right, 680),
|
|
|
|
|
("bravo left third line", left, 660),
|
|
|
|
|
("same tail phrase here", left, 640),
|
|
|
|
|
]
|
|
|
|
|
hits = locate_lines(
|
|
|
|
|
text,
|
|
|
|
|
[t for t, _, _ in lines],
|
|
|
|
|
[(x[0], y, x[1], y + 10) for _, x, y in lines],
|
|
|
|
|
)
|
|
|
|
|
self.assertEqual(hits[2], text.index("same tail"))
|
|
|
|
|
self.assertEqual(hits[5], text.rindex("same tail"))
|
|
|
|
|
|
|
|
|
|
def test_short_repeated_label_is_not_placed(self):
|
|
|
|
|
text = "Figure shows the LLM pipeline\nThe LLM is prompted twice"
|
|
|
|
|
hits = locate_lines(text, ["LLM", "The LLM is prompted twice"])
|
|
|
|
|
self.assertIsNone(hits[0])
|
|
|
|
|
self.assertEqual(hits[1], text.index("The LLM"))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class SplitLineAtGapsTest(unittest.TestCase):
|
|
|
|
|
@staticmethod
|
|
|
|
|
def _line(chars):
|
|
|
|
|
glyphs = [
|
|
|
|
|
{"ch": ch, "x0": x, "x1": x + 5, "y0": 0.0, "y1": 10.0, "i": i}
|
|
|
|
|
for ch, x, i in chars
|
|
|
|
|
]
|
|
|
|
|
return {"text": "".join(c for c, _, _ in chars), "bbox": (0, 0, 0, 0), "chars": glyphs}
|
|
|
|
|
|
|
|
|
|
def test_cuts_fused_columns_at_a_stream_jump(self):
|
|
|
|
|
from docreader.parser.pdf_parser import _split_line_at_gaps
|
|
|
|
|
|
|
|
|
|
left = [(ch, 10 + 5 * k, 100 + k) for k, ch in enumerate("left side")]
|
|
|
|
|
right = [(ch, 80 + 5 * k, 900 + k) for k, ch in enumerate("right side")]
|
|
|
|
|
parts = _split_line_at_gaps(self._line(left + right))
|
|
|
|
|
self.assertEqual([p["bbox"][0] for p in parts], [10, 80])
|
|
|
|
|
|
|
|
|
|
def test_keeps_a_wide_word_space_within_one_column(self):
|
|
|
|
|
from docreader.parser.pdf_parser import _split_line_at_gaps
|
|
|
|
|
|
|
|
|
|
chars = [(ch, 10 + 5 * k, k) for k, ch in enumerate("wide")]
|
|
|
|
|
chars += [(ch, 60 + 5 * k, 5 + k) for k, ch in enumerate("space")]
|
|
|
|
|
self.assertEqual(len(_split_line_at_gaps(self._line(chars))), 1)
|
|
|
|
|
|
|
|
|
|
|
2026-09-27 12:32:26 +00:00
|
|
|
|
|
|
|
|
class PrecisionRegressionTest(unittest.TestCase):
|
|
|
|
|
def test_common_prefix_cannot_place_a_different_conclusion(self):
|
|
|
|
|
prefix = "Common opening repeated before the important conclusion "
|
|
|
|
|
self.assertEqual(locate_lines(prefix + "failed", [prefix + "passed"]), [None])
|
|
|
|
|
|
|
|
|
|
def test_numeric_meaning_is_preserved(self):
|
|
|
|
|
for source, quote in [("limit 15 mm", "limit 1.5 mm"), ("temperature 5 C", "temperature -5 C"), ("rate 5", "rate 5%")]:
|
|
|
|
|
self.assertEqual(locate_lines(source, [quote]), [None])
|
|
|
|
|
|
|
|
|
|
def test_unaccounted_duplicate_does_not_take_first_region(self):
|
|
|
|
|
quote = "A repeated paragraph with enough text"
|
|
|
|
|
self.assertEqual(locate_lines(quote + "\n" + quote, [quote]), [None])
|
|
|
|
|
|
|
|
|
|
def test_unmatched_text_never_inherits_previous_geometry(self):
|
|
|
|
|
page = "First matched line.\nUnmatched evidence in between.\nLast matched line."
|
|
|
|
|
lines = [_ln("First matched line.", 50, 700, 300, 712), _ln("Last matched line.", 50, 668, 300, 680)]
|
|
|
|
|
blocks = page_blocks(page, 0, 2, lines, PageGeometry((0, 0, 600, 800)))
|
|
|
|
|
at = page.index("Unmatched")
|
|
|
|
|
covering = [b for b in blocks if b["start"] <= at < b["end"]]
|
|
|
|
|
self.assertEqual(len(covering), 1)
|
|
|
|
|
self.assertNotIn("bbox", covering[0]["locator"])
|
|
|
|
|
self.assertEqual(covering[0]["locator"]["page"], 2)
|
|
|
|
|
|
|
|
|
|
def test_trailing_unmatched_text_has_only_page_provenance(self):
|
|
|
|
|
page = "Matched paragraph.\nUnrecognized trailing evidence."
|
|
|
|
|
blocks = page_blocks(page, 0, 7, [_ln("Matched paragraph.", 50, 700, 300, 712)], PageGeometry((0, 0, 600, 800)))
|
|
|
|
|
self.assertEqual(blocks[0]["end"], page.index("Unrecognized"))
|
|
|
|
|
self.assertNotIn("bbox", blocks[-1]["locator"])
|
|
|
|
|
|
|
|
|
|
def test_wire_preserves_verified_geometry_and_identity(self):
|
|
|
|
|
from docreader.source_wire import source_blocks_to_proto
|
|
|
|
|
locator = {"type": "pdf", "page": 501, "mapping": "exact", "source_id": "pdf:501:0:3", "bbox": [.1, .2, .8, .4]}
|
|
|
|
|
wire = source_blocks_to_proto(Document(content="abc", source_blocks=[{"start": 0, "end": 3, "locator": locator}]))
|
|
|
|
|
self.assertEqual(json.loads(wire[0].locator_json), locator)
|
|
|
|
|
|
feat(chat): open cited documents at the cited passage (#3700)
* feat(chat): open cited documents at the cited passage
Clicking a citation used to show only the parsed chunk. It now opens the
original file in the references panel, jumps to the cited place and
highlights it, narrowed to the sentence the citation supports.
Parsing records where each chunk sits in the original file
(chunks.source_locators): PDF page and paragraph boxes from the docreader,
MinerU and PaddleOCR-VL layouts; Word paragraphs, slides, sheet rows and
EPUB chapters aligned in Go against any parser engine's Markdown; text
offsets for Markdown/TXT; segment times for audio. Pipeline text rewrites
are followed with a line-diff remapper. Scanned PDF pages get their OCR
text as quotes at search time so a sentence can pick its page.
The frontend renders PDFs with pdf.js and every other previewable type
with the existing preview, then highlights with the CSS Custom Highlight
API in a highlighter yellow that stays visible against the brand green.
* fix(docreader): serialize source blocks outside main so tests import it
test_docx_tables stubs docreader.parser in sys.modules, so importing
docreader.main from a later test failed in the full unittest discover run.
The wire conversion now lives in docreader.source_wire, which needs only
the proto.
* fix(chat): tighten source locators from review
- Remap rewritten-line-for-line stretches by line so scanned PDF pages
no longer drift onto the wrong page.
- Keep the source preview mounted for citations in the file already open.
- Wait for in-flight PDF page renders before looking for the text layer.
- Count image rune offsets in one pass; stream xlsx rows up to maxUnits.
- Start text locators after a UTF-8 BOM, as viewers decode without it.
- Only re-query image children for scanned-page results.
- Seek audio citations without autoplay and drop stale seek listeners.
- Fix the SQLite migration's cross-reference to versioned/000113.
* fix(migrations): renumber chunk source locators after main's chunk_images
main took versioned 000113 and sqlite 000032 for chunk_images, so the
source_locators column moves to versioned 000114 and sqlite 000033.
2026-09-25 05:26:16 +00:00
|
|
|
if __name__ == "__main__":
|
|
|
|
|
unittest.main()
|