From 4a618e8dcb6f941e6239e2494be4ba714c0b2ea5 Mon Sep 17 00:00:00 2001 From: Aman Mehtar Date: Tue, 20 Oct 2026 22:28:08 +0530 Subject: [PATCH] Edit text: do not group or join by an empty font name (#3661) Review follow-up. `Metrics::from_dict` leaves `base_font` empty when a font has no `/BaseFont`, and a bare subset tag such as `/ABCDEF+` left an empty name after stripping. Two consequences: - `same_style` split one paragraph into separate reading blocks for the exporters when both names were empty, though they grouped before subset joining; - `/ABCDEF+` and `/GHIJKL+` compared equal after stripping, so unrelated malformed fonts could join into one line. `same_base_font` now strips both tags first and requires a non-empty, equal name, so a missing name or a bare tag matches nothing. `same_style` keeps grouping lines whose fonts both lack a name, since the resource or fallback face `same_font` already matched is their whole identity. Tests: a page in a font without `/BaseFont` still reads as one block (it split into two before this change), and a bare tag matches no other name. --- crates/edit/src/tests.rs | 20 +++++++++++++++++++- crates/edit/src/text.rs | 16 +++++++++++++++- crates/edit/src/text_regions.rs | 5 ++++- 3 files changed, 38 insertions(+), 3 deletions(-) diff --git a/crates/edit/src/tests.rs b/crates/edit/src/tests.rs index efde39a7f..65292daa3 100644 --- a/crates/edit/src/tests.rs +++ b/crates/edit/src/tests.rs @@ -883,7 +883,7 @@ fn text_page(content: &str) -> Document { let objs: Vec = vec![ "<< /Type /Catalog /Pages 2 0 R >>".into(), "<< /Type /Pages /Kids [3 0 R] /Count 1 >>".into(), - "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 600 800] /Contents 4 0 R /Resources << /Font << /F1 5 0 R /F2 6 0 R /F3 7 0 R /F4 8 0 R >> >> >>".into(), + "<< /Type /Page /Parent 2 0 R /MediaBox [0 0 600 800] /Contents 4 0 R /Resources << /Font << /F1 5 0 R /F2 6 0 R /F3 7 0 R /F4 8 0 R /F5 9 0 R >> >> >>".into(), format!("<< /Length {} >>\nstream\n{content}\nendstream", content.len()), "<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica /Encoding /WinAnsiEncoding >>".into(), // Subset: glyphs for a (97) and b (98) only. @@ -894,6 +894,8 @@ fn text_page(content: &str) -> Document { .into(), // Helvetica again, in a second resource. "<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica /Encoding /WinAnsiEncoding >>".into(), + // No `/BaseFont`: `Metrics::from_dict` leaves its base-font name empty. + "<< /Type /Font /Subtype /Type1 /Encoding /WinAnsiEncoding >>".into(), ]; let mut out = b"%PDF-1.7\n".to_vec(); let mut offs = Vec::new(); @@ -1760,6 +1762,22 @@ fn different_fonts_start_a_new_line() { assert_eq!(text::text_lines(&doc, 0).unwrap().len(), 2); } +/// A font without `/BaseFont` has no name to compare. Its lines still group by resource, as they +/// did before subset joining: an empty name must not split one paragraph into two blocks. +#[test] +fn lines_in_a_font_without_a_base_font_stay_one_block() { + let doc = text_page("BT /F5 12 Tf 72 700 Td (First) Tj 0 -15 Td (second) Tj ET"); + let lines = text::text_lines(&doc, 0).unwrap(); + assert_eq!(lines.len(), 2); + assert!(lines.iter().all(|l| l.base_font.is_empty()), "{:?}", lines.iter().map(|l| &l.base_font).collect::>()); + let blocks = text::text_blocks(&doc, 0).unwrap(); + assert_eq!(blocks.len(), 1); + assert_eq!(blocks[0].text, "First second"); + let reading = text::reading_blocks(&doc, 0).unwrap(); + assert_eq!(reading.len(), 1, "{:?}", reading.iter().map(|b| &b.text).collect::>()); + assert_eq!(reading[0].text, "First second"); +} + #[test] fn text_lines_are_found_and_replaced_in_place() { let mut doc = diff --git a/crates/edit/src/text.rs b/crates/edit/src/text.rs index 1aa3e5840..c7adfdb36 100644 --- a/crates/edit/src/text.rs +++ b/crates/edit/src/text.rs @@ -1466,8 +1466,10 @@ fn same_font(line: &TextLine, font: &[u8], base_font: &str, face: Option<&str>) /// Whether two `/BaseFont` names name the same font: a subset's six-letter tag and `+` prefix do /// not change the font (ISO 32000-2 §9.6.4), so `ABCDEF+Arial` and `GHIJKL+Arial` are one font. +/// A name that is missing, or only a tag (`ABCDEF+`), names no font and matches nothing. fn same_base_font(a: &str, b: &str) -> bool { - !a.is_empty() && (a == b || strip_subset_tag(a) == strip_subset_tag(b)) + let (a, b) = (strip_subset_tag(a), strip_subset_tag(b)); + !a.is_empty() && a == b } /// `ABCDEF+Arial` → `Arial`. A name without a subset tag is returned unchanged. @@ -2804,6 +2806,18 @@ pub(crate) fn translate_blocks(doc: &mut Document, page: usize, indexes: &[usize mod tests { use super::*; + #[test] + fn a_subset_tag_is_the_only_difference_between_subsets_of_one_font() { + assert!(same_base_font("ABCDEF+Arial", "GHIJKL+Arial")); + assert!(same_base_font("Arial", "ABCDEF+Arial")); + assert!(same_base_font("Arial", "Arial")); + assert!(!same_base_font("ABCDEF+Arial", "GHIJKL+Helvetica")); + // A tag with no name after it, or no name at all, names no font. + assert!(!same_base_font("ABCDEF+", "GHIJKL+")); + assert!(!same_base_font("", "")); + assert!(!same_base_font("Arial", "")); + } + #[test] fn paragraph_wrapping_keeps_ideographic_spaces() { let width = |s: &str| s.chars().count() as f64; diff --git a/crates/edit/src/text_regions.rs b/crates/edit/src/text_regions.rs index cd3a25887..a00e42f15 100644 --- a/crates/edit/src/text_regions.rs +++ b/crates/edit/src/text_regions.rs @@ -52,7 +52,10 @@ fn invisible(line: &TextLine) -> bool { fn same_style(a: &TextLine, b: &TextLine) -> bool { same_font(a, b.font.as_bytes(), &b.base_font, b.face.as_deref()) - && same_base_font(&a.base_font, &b.base_font) + // With no `/BaseFont` (or only a subset tag) there is no name to compare: the resource or + // fallback face `same_font` matched is the whole identity, so keep the lines together. + && (same_base_font(&a.base_font, &b.base_font) + || (strip_subset_tag(&a.base_font).is_empty() && strip_subset_tag(&b.base_font).is_empty())) && a.bold == b.bold && a.italic == b.italic && a.color == b.color