|
| 1 | +use anyhow::Result; |
| 2 | +use serde_json::{Value, json}; |
| 3 | +use std::collections::HashMap; |
| 4 | + |
| 5 | +use super::assets; |
| 6 | + |
| 7 | +const SCHEMA: &str = "paperdown.layout.v1"; |
| 8 | + |
| 9 | +pub(crate) fn render_layout_json( |
| 10 | + layout_details: &[Value], |
| 11 | + figure_replacements: &HashMap<String, String>, |
| 12 | + tables_raw_written: usize, |
| 13 | + source_rel: &str, |
| 14 | +) -> Result<String> { |
| 15 | + let table_regions = count_table_regions(layout_details); |
| 16 | + let link_tables = table_regions > 0 && table_regions == tables_raw_written; |
| 17 | + let table_artifact_match = if link_tables { "order" } else { "none" }; |
| 18 | + let mut table_index = 0usize; |
| 19 | + |
| 20 | + let pages: Vec<Value> = layout_details |
| 21 | + .iter() |
| 22 | + .enumerate() |
| 23 | + .map(|(page_index, page_blocks)| { |
| 24 | + let regions = page_blocks |
| 25 | + .as_array() |
| 26 | + .map(|blocks| { |
| 27 | + blocks |
| 28 | + .iter() |
| 29 | + .enumerate() |
| 30 | + .map(|(block_index, block)| { |
| 31 | + render_region( |
| 32 | + block, |
| 33 | + block_index, |
| 34 | + figure_replacements, |
| 35 | + link_tables, |
| 36 | + &mut table_index, |
| 37 | + ) |
| 38 | + }) |
| 39 | + .collect::<Vec<_>>() |
| 40 | + }) |
| 41 | + .unwrap_or_default(); |
| 42 | + |
| 43 | + json!({ |
| 44 | + "page": page_index + 1, |
| 45 | + "regions": regions, |
| 46 | + }) |
| 47 | + }) |
| 48 | + .collect(); |
| 49 | + |
| 50 | + let rendered = json!({ |
| 51 | + "schema": SCHEMA, |
| 52 | + "source_pdf": source_rel, |
| 53 | + "ocr_model": "glm-ocr", |
| 54 | + "bbox_format": { |
| 55 | + "order": "x1,y1,x2,y2", |
| 56 | + "origin": "top-left", |
| 57 | + }, |
| 58 | + "table_artifact_match": table_artifact_match, |
| 59 | + "pages": pages, |
| 60 | + }); |
| 61 | + |
| 62 | + let mut text = serde_json::to_string_pretty(&rendered)?; |
| 63 | + text.push('\n'); |
| 64 | + Ok(text) |
| 65 | +} |
| 66 | + |
| 67 | +fn count_table_regions(layout_details: &[Value]) -> usize { |
| 68 | + layout_details |
| 69 | + .iter() |
| 70 | + .filter_map(Value::as_array) |
| 71 | + .flatten() |
| 72 | + .filter(|block| label(block) == Some("table")) |
| 73 | + .count() |
| 74 | +} |
| 75 | + |
| 76 | +fn render_region( |
| 77 | + block: &Value, |
| 78 | + index: usize, |
| 79 | + figure_replacements: &HashMap<String, String>, |
| 80 | + link_tables: bool, |
| 81 | + table_index: &mut usize, |
| 82 | +) -> Value { |
| 83 | + let label = label(block); |
| 84 | + let artifact = match label { |
| 85 | + Some("image") => assets::extract_image_url(block) |
| 86 | + .and_then(|url| figure_replacements.get(&url).cloned()) |
| 87 | + .map(Value::String) |
| 88 | + .unwrap_or(Value::Null), |
| 89 | + Some("table") if link_tables => { |
| 90 | + *table_index += 1; |
| 91 | + Value::String(format!("tables/table_{:03}.html", *table_index)) |
| 92 | + } |
| 93 | + Some("table") => { |
| 94 | + *table_index += 1; |
| 95 | + Value::Null |
| 96 | + } |
| 97 | + _ => Value::Null, |
| 98 | + }; |
| 99 | + |
| 100 | + json!({ |
| 101 | + "index": index, |
| 102 | + "label": label.map(str::to_owned).map(Value::String).unwrap_or(Value::Null), |
| 103 | + "bbox": bbox(block), |
| 104 | + "content": block.get("content").cloned().unwrap_or(Value::Null), |
| 105 | + "artifact": artifact, |
| 106 | + }) |
| 107 | +} |
| 108 | + |
| 109 | +fn label(block: &Value) -> Option<&str> { |
| 110 | + block.get("label").and_then(Value::as_str) |
| 111 | +} |
| 112 | + |
| 113 | +fn bbox(block: &Value) -> Value { |
| 114 | + block |
| 115 | + .get("bbox_2d") |
| 116 | + .or_else(|| block.get("bbox")) |
| 117 | + .cloned() |
| 118 | + .unwrap_or(Value::Null) |
| 119 | +} |
| 120 | + |
| 121 | +#[cfg(test)] |
| 122 | +mod tests { |
| 123 | + use super::render_layout_json; |
| 124 | + use serde_json::{Value, json}; |
| 125 | + use std::collections::HashMap; |
| 126 | + |
| 127 | + // Render `layout_details` through the real renderer and parse the output back |
| 128 | + // into a `Value` so assertions are structural, not brittle pretty-text checks. |
| 129 | + fn render( |
| 130 | + layout: &[Value], |
| 131 | + replacements: &HashMap<String, String>, |
| 132 | + tables_written: usize, |
| 133 | + source: &str, |
| 134 | + ) -> Value { |
| 135 | + let text = |
| 136 | + render_layout_json(layout, replacements, tables_written, source).expect("renders"); |
| 137 | + serde_json::from_str(&text).expect("output is valid JSON") |
| 138 | + } |
| 139 | + |
| 140 | + #[test] |
| 141 | + fn full_render_links_image_and_table_and_preserves_latex() { |
| 142 | + // One page: title, a LaTeX formula, an image region (remote URL), a table. |
| 143 | + let layout = vec![json!([ |
| 144 | + {"label": "title", "bbox_2d": [72, 41, 928, 96], "content": "Paper title"}, |
| 145 | + { |
| 146 | + "label": "formula", |
| 147 | + "bbox_2d": [10, 10, 200, 50], |
| 148 | + "content": r"$\int_0^\infty e^{-x^2}\,dx = \frac{\sqrt{\pi}}{2}$" |
| 149 | + }, |
| 150 | + { |
| 151 | + "label": "image", |
| 152 | + "bbox_2d": [90, 120, 910, 390], |
| 153 | + "content": "https://example.com/fig.png" |
| 154 | + }, |
| 155 | + { |
| 156 | + "label": "table", |
| 157 | + "bbox_2d": [80, 530, 920, 780], |
| 158 | + "content": "<table>x</table>" |
| 159 | + } |
| 160 | + ])]; |
| 161 | + // The figure URL resolves through the replacement map to a local file. |
| 162 | + let replacements = HashMap::from([( |
| 163 | + "https://example.com/fig.png".to_string(), |
| 164 | + "figures/fig-001-001.png".to_string(), |
| 165 | + )]); |
| 166 | + |
| 167 | + // One table region equals the number of tables actually written -> order linking. |
| 168 | + let doc = render(&layout, &replacements, 1, "paper.pdf"); |
| 169 | + |
| 170 | + assert_eq!(doc["schema"], "paperdown.layout.v1"); |
| 171 | + assert_eq!(doc["source_pdf"], "paper.pdf"); |
| 172 | + assert_eq!(doc["ocr_model"], "glm-ocr"); |
| 173 | + assert_eq!(doc["bbox_format"]["order"], "x1,y1,x2,y2"); |
| 174 | + assert_eq!(doc["bbox_format"]["origin"], "top-left"); |
| 175 | + assert_eq!(doc["table_artifact_match"], "order"); |
| 176 | + |
| 177 | + let regions = doc["pages"][0]["regions"] |
| 178 | + .as_array() |
| 179 | + .expect("regions array"); |
| 180 | + assert_eq!(doc["pages"][0]["page"], 1); |
| 181 | + assert_eq!(regions.len(), 4); |
| 182 | + |
| 183 | + // Title: bbox copied verbatim from bbox_2d. |
| 184 | + assert_eq!(regions[0]["label"], "title"); |
| 185 | + assert_eq!(regions[0]["bbox"].clone(), json!([72, 41, 928, 96])); |
| 186 | + assert_eq!(regions[0]["content"], "Paper title"); |
| 187 | + assert_eq!(regions[0]["artifact"], Value::Null); |
| 188 | + |
| 189 | + // Formula: raw LaTeX content is carried through unchanged. |
| 190 | + assert_eq!(regions[1]["label"], "formula"); |
| 191 | + assert_eq!( |
| 192 | + regions[1]["content"].as_str(), |
| 193 | + Some(r"$\int_0^\infty e^{-x^2}\,dx = \frac{\sqrt{\pi}}{2}$"), |
| 194 | + ); |
| 195 | + assert_eq!(regions[1]["artifact"], Value::Null); |
| 196 | + |
| 197 | + // Image: artifact resolved via the figure replacement map. |
| 198 | + assert_eq!(regions[2]["label"], "image"); |
| 199 | + assert_eq!(regions[2]["artifact"], "figures/fig-001-001.png"); |
| 200 | + |
| 201 | + // Table: first table region -> tables/table_001.html by document order. |
| 202 | + assert_eq!(regions[3]["label"], "table"); |
| 203 | + assert_eq!(regions[3]["artifact"], "tables/table_001.html"); |
| 204 | + } |
| 205 | + |
| 206 | + #[test] |
| 207 | + fn table_count_mismatch_skips_table_links() { |
| 208 | + // Two table regions, but three tables written: counts disagree. |
| 209 | + let layout = vec![json!([ |
| 210 | + {"label": "table", "bbox_2d": [1, 2, 3, 4], "content": "<table>a</table>"}, |
| 211 | + {"label": "table", "bbox_2d": [5, 6, 7, 8], "content": "<table>b</table>"} |
| 212 | + ])]; |
| 213 | + let replacements = HashMap::new(); |
| 214 | + |
| 215 | + let doc = render(&layout, &replacements, 3, "paper.pdf"); |
| 216 | + |
| 217 | + assert_eq!(doc["table_artifact_match"], "none"); |
| 218 | + let regions = doc["pages"][0]["regions"] |
| 219 | + .as_array() |
| 220 | + .expect("regions array"); |
| 221 | + assert_eq!(regions.len(), 2); |
| 222 | + assert!( |
| 223 | + regions.iter().all(|r| r["artifact"].is_null()), |
| 224 | + "every table artifact must be null when region count != tables written" |
| 225 | + ); |
| 226 | + } |
| 227 | + |
| 228 | + #[test] |
| 229 | + fn missing_or_non_string_fields_become_null() { |
| 230 | + // Region 0: non-string label, no bbox keys, no content. |
| 231 | + // Region 1: entirely absent label/bbox/content. |
| 232 | + let layout = vec![json!([{"label": 123}, {"weird": "block"}])]; |
| 233 | + let replacements = HashMap::new(); |
| 234 | + |
| 235 | + let doc = render(&layout, &replacements, 0, "paper.pdf"); |
| 236 | + let regions = doc["pages"][0]["regions"] |
| 237 | + .as_array() |
| 238 | + .expect("regions array"); |
| 239 | + |
| 240 | + for region in regions { |
| 241 | + assert_eq!(region["label"], Value::Null); |
| 242 | + assert_eq!(region["bbox"], Value::Null); |
| 243 | + assert_eq!(region["content"], Value::Null); |
| 244 | + assert_eq!(region["artifact"], Value::Null); |
| 245 | + } |
| 246 | + } |
| 247 | + |
| 248 | + #[test] |
| 249 | + fn non_array_page_emits_empty_regions() { |
| 250 | + // A page whose value is not an array must not panic; it yields no regions. |
| 251 | + let layout = vec![json!({"not": "an array"})]; |
| 252 | + let replacements = HashMap::new(); |
| 253 | + |
| 254 | + let doc = render(&layout, &replacements, 0, "paper.pdf"); |
| 255 | + |
| 256 | + assert_eq!(doc["pages"][0]["page"], 1); |
| 257 | + assert_eq!(doc["pages"][0]["regions"].clone(), json!([])); |
| 258 | + } |
| 259 | +} |
0 commit comments