zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-08-25 01:20:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix(pdf): align TSR crop margin with Python (fixed 30px, not 3%) (#18387)

Browse Source
This commit is contained in:
Jack
2026-08-18 13:11:41 +08:00
committed by GitHub
parent 344bd4d57a
commit 5ec11f7b07
5 changed files with 124 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
internal/deepdoc/parser/pdf/table/table_construct_test.go
View File

@@ -143,13 +143,14 @@ func TestExtractTableAndReplace_CellTextFilled(t *testing.T) {
// Simulate 公司差旅费 page 0 table coordinates.
// DLA region: X0=217, X1=1584, Y0=985, Y1=1599 at 216 DPI → PDF: 72-528 x 328-533
// Scale = 216/72 = 3.0
// cropOff ≈ region.X - region.W*0.03
// cropOff = region.X - 10pt*ZM = region.X - 30px (fixed margin, matches
// Python's MARGIN=10; NOT the old Go 3% proportional margin).
const scale = 3.0
const cropOffX = 176.0
const cropOffY = 967.0
const cropOffX = 187.0
const cropOffY = 955.0
// Post-merge boxes in PDF point space (inside the table region).
// PDF Y=470 → crop Top = 470*3-967 = 443 → overlaps crop cell at Y0=441.
// PDF Y=470 → crop Top = 470*3-955 = 455 → overlaps crop cell at Y0=441.
// Boxes must have R (row) and C (col) annotations matching cells,
// matching Python's construct_table which assigns boxes to cells by R/C.
boxes := []pdf.TextBox{