kreuzberg 4.0.0.pre.rc.29 → 4.0.0.rc1

data/vendor/kreuzberg/src/extractors/excel.rs

@@ -27,23 +27,50 @@ impl ExcelExtractor {
     ///
     /// Each sheet becomes a table with the first row as headers,
     /// remaining rows as data, and the sheet name as caption.
-    /// Uses pre-extracted cells from ExcelSheet::table_cells to avoid
-    /// expensive markdown re-parsing (40-60% performance improvement).
     fn sheets_to_tables(workbook: &crate::types::ExcelWorkbook) -> Vec<Table> {
         let mut tables = Vec::with_capacity(workbook.sheets.len());
 
         for (sheet_index, sheet) in workbook.sheets.iter().enumerate() {
+            // Skip empty sheets
             if sheet.row_count == 0 || sheet.col_count == 0 {
                 continue;
             }
 
-            if let Some(cells) = &sheet.table_cells
-                && !cells.is_empty()
-            {
+            // We need to re-parse the sheet to get structured cell data
+            // The workbook.sheets only contains markdown, not raw cell data
+            // So we'll extract from the markdown table representation
+
+            // Parse cells from markdown
+            let lines: Vec<&str> = sheet.markdown.lines().collect();
+            let mut cells: Vec<Vec<String>> = Vec::new();
+
+            // Find the table content (skip header line "## Sheet Name" and blank line)
+            let table_start = lines.iter().position(|line| line.starts_with("| "));
+
+            if let Some(start_idx) = table_start {
+                for line in lines.iter().skip(start_idx) {
+                    if line.starts_with("| ") && !line.contains("---") {
+                        // Parse table row
+                        let row: Vec<String> = line
+                            .trim_start_matches("| ")
+                            .trim_end_matches(" |")
+                            .split(" | ")
+                            .map(|cell| {
+                                // Unescape markdown pipes and backslashes
+                                cell.replace("\\|", "|").replace("\\\\", "\\")
+                            })
+                            .collect();
+                        cells.push(row);
+                    }
+                }
+            }
+
+            // Only create table if we have data
+            if !cells.is_empty() {
                 tables.push(Table {
-                    cells: cells.clone(),
+                    cells,
                     markdown: sheet.markdown.clone(),
-                    page_number: sheet_index + 1,
+                    page_number: sheet_index + 1, // 1-indexed
                 });
             }
         }
@@ -72,13 +99,6 @@ impl Plugin for ExcelExtractor {
 
 #[async_trait]
 impl DocumentExtractor for ExcelExtractor {
-    #[cfg_attr(feature = "otel", tracing::instrument(
-        skip(self, content, _config),
-        fields(
-            extractor.name = self.name(),
-            content.size_bytes = content.len(),
-        )
-    ))]
     async fn extract_bytes(
         &self,
         content: &[u8],
@@ -97,17 +117,18 @@ impl DocumentExtractor for ExcelExtractor {
             _ => ".xlsx",
         };
 
+        // Extract workbook
         let workbook = if crate::core::batch_mode::is_batch_mode() {
+            // Batch mode: Use spawn_blocking for parallelism
             let content_owned = content.to_vec();
             let extension_owned = extension.to_string();
-            let span = tracing::Span::current();
             tokio::task::spawn_blocking(move || {
-                let _guard = span.entered();
                 crate::extraction::excel::read_excel_bytes(&content_owned, &extension_owned)
             })
             .await
             .map_err(|e| crate::error::KreuzbergError::parsing(format!("Excel extraction task failed: {}", e)))??
         } else {
+            // Single-file mode: Direct extraction (no spawn overhead)
             crate::extraction::excel::read_excel_bytes(content, extension)?
         };
 
@@ -135,7 +156,6 @@ impl DocumentExtractor for ExcelExtractor {
                 additional,
                 ..Default::default()
             },
-            pages: None,
             tables,
             detected_languages: None,
             chunks: None,
@@ -143,12 +163,6 @@ impl DocumentExtractor for ExcelExtractor {
         })
     }
 
-    #[cfg_attr(feature = "otel", tracing::instrument(
-        skip(self, path, _config),
-        fields(
-            extractor.name = self.name(),
-        )
-    ))]
     async fn extract_file(&self, path: &Path, mime_type: &str, _config: &ExtractionConfig) -> Result<ExtractionResult> {
         let path_str = path
             .to_str()
@@ -179,7 +193,6 @@ impl DocumentExtractor for ExcelExtractor {
                 additional,
                 ..Default::default()
             },
-            pages: None,
             tables,
             detected_languages: None,
             chunks: None,
@@ -231,6 +244,7 @@ mod tests {
         use crate::types::ExcelSheet;
         use std::collections::HashMap;
 
+        // Create a mock workbook with a single sheet
         let sheet = ExcelSheet {
             name: "TestSheet".to_string(),
             markdown: r#"## TestSheet
@@ -244,11 +258,6 @@ mod tests {
             row_count: 3,
             col_count: 3,
             cell_count: 9,
-            table_cells: Some(vec![
-                vec!["Name".to_string(), "Age".to_string(), "City".to_string()],
-                vec!["Alice".to_string(), "30".to_string(), "NYC".to_string()],
-                vec!["Bob".to_string(), "25".to_string(), "LA".to_string()],
-            ]),
         };
 
         let workbook = crate::types::ExcelWorkbook {
@@ -260,7 +269,7 @@ mod tests {
 
         assert_eq!(tables.len(), 1);
         assert_eq!(tables[0].page_number, 1);
-        assert_eq!(tables[0].cells.len(), 3);
+        assert_eq!(tables[0].cells.len(), 3); // Header + 2 data rows
         assert_eq!(tables[0].cells[0], vec!["Name", "Age", "City"]);
         assert_eq!(tables[0].cells[1], vec!["Alice", "30", "NYC"]);
         assert_eq!(tables[0].cells[2], vec!["Bob", "25", "LA"]);
@@ -277,7 +286,6 @@ mod tests {
             row_count: 0,
             col_count: 0,
             cell_count: 0,
-            table_cells: None,
         };
 
         let workbook = crate::types::ExcelWorkbook {
@@ -286,7 +294,7 @@ mod tests {
         };
 
         let tables = ExcelExtractor::sheets_to_tables(&workbook);
-        assert_eq!(tables.len(), 0);
+        assert_eq!(tables.len(), 0); // Empty sheets should not create tables
     }
 
     #[test]
@@ -306,10 +314,6 @@ mod tests {
             row_count: 2,
             col_count: 2,
             cell_count: 4,
-            table_cells: Some(vec![
-                vec!["Col1".to_string(), "Col2".to_string()],
-                vec!["A".to_string(), "B".to_string()],
-            ]),
         };
 
         let sheet2 = ExcelSheet {
@@ -324,10 +328,6 @@ mod tests {
             row_count: 2,
             col_count: 2,
             cell_count: 4,
-            table_cells: Some(vec![
-                vec!["X".to_string(), "Y".to_string()],
-                vec!["1".to_string(), "2".to_string()],
-            ]),
         };
 
         let workbook = crate::types::ExcelWorkbook {
@@ -341,45 +341,4 @@ mod tests {
         assert_eq!(tables[0].page_number, 1);
         assert_eq!(tables[1].page_number, 2);
     }
-
-    #[test]
-    fn test_sheets_to_tables_preserves_cell_content() {
-        use crate::types::ExcelSheet;
-        use std::collections::HashMap;
-
-        let sheet = ExcelSheet {
-            name: "TestSheet".to_string(),
-            markdown: r#"## TestSheet
-
-| Name | Value | Amount |
-| --- | --- | --- |
-| Item\|A | 100 | $1,000 |
-| Item B | 200 | $2,000 |
-"#
-            .to_string(),
-            row_count: 3,
-            col_count: 3,
-            cell_count: 9,
-            table_cells: Some(vec![
-                vec!["Name".to_string(), "Value".to_string(), "Amount".to_string()],
-                vec!["Item|A".to_string(), "100".to_string(), "$1,000".to_string()],
-                vec!["Item B".to_string(), "200".to_string(), "$2,000".to_string()],
-            ]),
-        };
-
-        let workbook = crate::types::ExcelWorkbook {
-            sheets: vec![sheet],
-            metadata: HashMap::new(),
-        };
-
-        let tables = ExcelExtractor::sheets_to_tables(&workbook);
-
-        assert_eq!(tables.len(), 1);
-        assert_eq!(
-            tables[0].cells[1][0], "Item|A",
-            "Escaped characters should be preserved"
-        );
-        assert_eq!(tables[0].cells[1][2], "$1,000");
-        assert_eq!(tables[0].cells[2][0], "Item B");
-    }
 }