Support multi-line text labels by splitting on newlines (#301)

Author: teunbrand

src/writer/vegalite/encoding.rs

@@ -415,7 +415,7 @@ fn apply_title_to_encoding(
             .and_then(|labels| labels.labels.get(primary));
 
         if let Some(label) = explicit_label {
-            encoding["title"] = json!(label);
+            encoding["title"] = super::split_label_on_newlines(label);
             titled_families.insert(primary.to_string());
         } else if let Some(orig) = original_name {
             // Use original column name as default title when available
@@ -429,7 +429,7 @@ fn apply_title_to_encoding(
         // Variant without primary: allow first variant to claim title (for explicit labels)
         if let Some(ref labels) = spec.labels {
             if let Some(label) = labels.labels.get(primary) {
-                encoding["title"] = json!(label);
+                encoding["title"] = super::split_label_on_newlines(label);
                 titled_families.insert(primary.to_string());
             }
         }

src/writer/vegalite/layer.rs

@@ -653,6 +653,34 @@ impl TextRenderer {
         Ok((result_df, run_lengths))
     }
 
+    /// Split label values containing newlines into arrays of strings
+    ///
+    /// Uses the shared split_label_on_newlines function to ensure consistent
+    /// newline handling across all label types (text data, axis labels, titles, etc.)
+    fn split_label_newlines(values: &mut [Value]) -> Result<()> {
+        let label_col = naming::aesthetic_column("label");
+
+        for row in values.iter_mut() {
+            // Get the object, skip if not an object
+            let Some(obj) = row.as_object_mut() else {
+                continue;
+            };
+
+            // Get the label value, skip if not present
+            let Some(label_value) = obj.get(&label_col) else {
+                continue;
+            };
+            // Get the string value, skip if not a string
+            let Some(label_str) = label_value.as_str() else {
+                continue;
+            };
+
+            // Use shared function for consistent newline splitting
+            obj.insert(label_col.clone(), super::split_label_on_newlines(label_str));
+        }
+        Ok(())
+    }
+
     /// Convert typeface to Vega-Lite font value
     /// Prefers literal over column value
     fn convert_typeface(
@@ -1032,12 +1060,15 @@ impl GeomRenderer for TextRenderer {
             // Slice the contiguous run from the DataFrame (more efficient than boolean masking)
             let sliced = df.slice(position as i64, length);
 
-            let values = if binned_columns.is_empty() {
+            let mut values = if binned_columns.is_empty() {
                 dataframe_to_values(&sliced)?
             } else {
                 dataframe_to_values_with_bins(&sliced, binned_columns)?
             };
 
+            // Post-process label values to split on newlines
+            Self::split_label_newlines(&mut values)?;
+
             components.insert(suffix, values);
             position += length;
         }
@@ -2784,6 +2815,81 @@ mod tests {
         assert!(labels.contains(&"$21.00"));
     }
 
+    #[test]
+    fn test_text_label_newline_splitting() {
+        use crate::execute;
+        use crate::reader::DuckDBReader;
+        use crate::writer::vegalite::VegaLiteWriter;
+        use crate::writer::Writer;
+
+        // Test that labels containing newlines are split into arrays
+
+        let reader = DuckDBReader::from_connection_string("duckdb://memory").unwrap();
+
+        // Query with labels containing newlines in DRAW and PLACE
+        let query = r#"
+            SELECT
+                n::INTEGER as x,
+                n::INTEGER as y,
+                CASE
+                    WHEN n = 0 THEN 'First Line\nSecond Line'
+                    WHEN n = 1 THEN 'Single Line'
+                    ELSE 'Line 1\nLine 2\nLine 3'
+                END as label
+            FROM generate_series(0, 2) as t(n)
+            VISUALISE x, y, label
+            DRAW text
+            PLACE text SETTING x => 5, y => 15, label => 'Annotation\nWith Newline'
+        "#;
+
+        let prepared = execute::prepare_data_with_reader(query, &reader).unwrap();
+        let spec = &prepared.specs[0];
+
+        let writer = VegaLiteWriter::new();
+        let json_str = writer.write(spec, &prepared.data).unwrap();
+        let vl_spec: serde_json::Value = serde_json::from_str(&json_str).unwrap();
+
+        let data_values = vl_spec["data"]["values"].as_array().unwrap();
+        let label_col = crate::naming::aesthetic_column("label");
+
+        // Check first label (contains newline - should be array)
+        let label_0 = &data_values[0][&label_col];
+        assert!(label_0.is_array(), "Label with newline should be an array");
+        let lines_0 = label_0.as_array().unwrap();
+        assert_eq!(lines_0.len(), 2);
+        assert_eq!(lines_0[0].as_str().unwrap(), "First Line");
+        assert_eq!(lines_0[1].as_str().unwrap(), "Second Line");
+
+        // Check second label (no newline - should be string)
+        let label_1 = &data_values[1][&label_col];
+        assert!(
+            label_1.is_string(),
+            "Label without newline should be a string"
+        );
+        assert_eq!(label_1.as_str().unwrap(), "Single Line");
+
+        // Check third label (multiple newlines - should be array with 3 elements)
+        let label_2 = &data_values[2][&label_col];
+        assert!(label_2.is_array(), "Label with newlines should be an array");
+        let lines_2 = label_2.as_array().unwrap();
+        assert_eq!(lines_2.len(), 3);
+        assert_eq!(lines_2[0].as_str().unwrap(), "Line 1");
+        assert_eq!(lines_2[1].as_str().unwrap(), "Line 2");
+        assert_eq!(lines_2[2].as_str().unwrap(), "Line 3");
+
+        // Check PLACE annotation layer (index 3, after the 3 DRAW data rows)
+        assert!(data_values.len() > 3, "Should have annotation data");
+        let annotation_label = &data_values[3][&label_col];
+        assert!(
+            annotation_label.is_array(),
+            "Annotation label with newline should be an array"
+        );
+        let annotation_lines = annotation_label.as_array().unwrap();
+        assert_eq!(annotation_lines.len(), 2);
+        assert_eq!(annotation_lines[0].as_str().unwrap(), "Annotation");
+        assert_eq!(annotation_lines[1].as_str().unwrap(), "With Newline");
+    }
+
     #[test]
     fn test_text_setting_fontweight() {
         use crate::execute;

src/writer/vegalite/mod.rs

@@ -49,6 +49,24 @@ const POINTS_TO_PIXELS: f64 = 96.0 / 72.0;
 /// So: area_px^2 = pi * (r_pt * POINTS_TO_PIXELS)^2 = pi * r_pt^2 * (96/72)^2
 const POINTS_TO_AREA: f64 = std::f64::consts::PI * POINTS_TO_PIXELS * POINTS_TO_PIXELS;
 
+/// Split a label string on newlines and return appropriate JSON value
+///
+/// Returns a JSON array if the string contains multiple lines, or a JSON string
+/// if it's a single line. Handles both actual newlines (from database columns,
+/// CHAR(10), imported data) and literal \\n (from SQL string literals).
+fn split_label_on_newlines(label: &str) -> Value {
+    // Normalize literal \\n to actual \n, then split on any newline type
+    let normalized = label.replace("\\n", "\n");
+    let lines: Vec<&str> = normalized.lines().collect();
+
+    // Return array if multiple lines, string if single line
+    if lines.len() > 1 {
+        json!(lines)
+    } else {
+        json!(label)
+    }
+}
+
 /// Result of preparing layer data for rendering
 ///
 /// Contains the datasets, renderers, and prepared data needed to build Vega-Lite layers.
@@ -1097,13 +1115,20 @@ impl Writer for VegaLiteWriter {
             match (title, subtitle) {
                 (Some(t), Some(st)) => {
                     // Vega-Lite uses an object for title + subtitle
-                    vl_spec["title"] = json!({"text": t, "subtitle": st});
+                    // Split both title and subtitle on newlines
+                    vl_spec["title"] = json!({
+                        "text": split_label_on_newlines(t),
+                        "subtitle": split_label_on_newlines(st)
+                    });
                 }
                 (Some(t), None) => {
-                    vl_spec["title"] = json!(t);
+                    vl_spec["title"] = split_label_on_newlines(t);
                 }
                 (None, Some(st)) => {
-                    vl_spec["title"] = json!({"text": "", "subtitle": st});
+                    vl_spec["title"] = json!({
+                        "text": "",
+                        "subtitle": split_label_on_newlines(st)
+                    });
                 }
                 (None, None) => {}
             }
@@ -1558,6 +1583,94 @@ mod tests {
         assert_eq!(vl_spec["layer"][0]["mark"]["clip"], true);
     }
 
+    #[test]
+    fn test_labels_newline_splitting() {
+        use crate::execute;
+        use crate::reader::DuckDBReader;
+
+        // Test that LABEL clause values with newlines are split into arrays
+
+        let reader = DuckDBReader::from_connection_string("duckdb://memory").unwrap();
+
+        let query = r#"
+            SELECT
+                n::INTEGER as x,
+                n::INTEGER as y,
+                CASE WHEN n % 2 = 0 THEN 'Group A' ELSE 'Group B' END as category
+            FROM generate_series(0, 2) as t(n)
+            VISUALISE x, y, category AS stroke
+            DRAW point
+            LABEL
+                title => 'Multi-line\nChart Title',
+                subtitle => 'Line 1\nLine 2\nLine 3',
+                x => 'X Axis\nWith Newline',
+                y => 'Single Line',
+                stroke => 'Category\nMulti-line Legend'
+        "#;
+
+        let prepared = execute::prepare_data_with_reader(query, &reader).unwrap();
+        let spec = &prepared.specs[0];
+
+        let writer = VegaLiteWriter::new();
+        let json_str = writer.write(spec, &prepared.data).unwrap();
+        let vl_spec: Value = serde_json::from_str(&json_str).unwrap();
+
+        // Check title (should be object with text and subtitle)
+        assert!(vl_spec["title"].is_object(), "Title should be an object");
+        let title_obj = vl_spec["title"].as_object().unwrap();
+
+        // Check title text (multi-line, should be array)
+        assert!(
+            title_obj["text"].is_array(),
+            "Title with newline should be an array"
+        );
+        let title_lines = title_obj["text"].as_array().unwrap();
+        assert_eq!(title_lines.len(), 2);
+        assert_eq!(title_lines[0].as_str().unwrap(), "Multi-line");
+        assert_eq!(title_lines[1].as_str().unwrap(), "Chart Title");
+
+        // Check subtitle (multi-line, should be array)
+        assert!(
+            title_obj["subtitle"].is_array(),
+            "Subtitle with newlines should be an array"
+        );
+        let subtitle_lines = title_obj["subtitle"].as_array().unwrap();
+        assert_eq!(subtitle_lines.len(), 3);
+        assert_eq!(subtitle_lines[0].as_str().unwrap(), "Line 1");
+        assert_eq!(subtitle_lines[1].as_str().unwrap(), "Line 2");
+        assert_eq!(subtitle_lines[2].as_str().unwrap(), "Line 3");
+
+        // Check x axis label (multi-line, should be array)
+        let x_encoding = &vl_spec["layer"][0]["encoding"]["x"];
+        assert!(
+            x_encoding["title"].is_array(),
+            "X axis label with newline should be an array"
+        );
+        let x_label_lines = x_encoding["title"].as_array().unwrap();
+        assert_eq!(x_label_lines.len(), 2);
+        assert_eq!(x_label_lines[0].as_str().unwrap(), "X Axis");
+        assert_eq!(x_label_lines[1].as_str().unwrap(), "With Newline");
+
+        // Check y axis label (single line, should be string)
+        let y_encoding = &vl_spec["layer"][0]["encoding"]["y"];
+        assert!(
+            y_encoding["title"].is_string(),
+            "Y axis label without newline should be a string"
+        );
+        assert_eq!(y_encoding["title"].as_str().unwrap(), "Single Line");
+
+        // Check stroke legend label (multi-line, should be array)
+        let stroke_encoding = &vl_spec["layer"][0]["encoding"]["stroke"];
+        assert!(
+            stroke_encoding["title"].is_array(),
+            "Stroke legend title with newline should be an array"
+        );
+        let stroke_label_lines = stroke_encoding["title"].as_array().unwrap();
+        assert_eq!(stroke_label_lines.len(), 2);
+        assert_eq!(stroke_label_lines[0].as_str().unwrap(), "Category");
+        assert_eq!(stroke_label_lines[1].as_str().unwrap(), "Multi-line Legend");
+    }
+
     #[test]
     fn test_fontsize_linear_scaling() {
         use crate::plot::{ArrayElement, OutputRange, Scale, ScaleType};