fix(import): preserve empty strings in Excel by default

Closes #5329

apps/desktop/src/components/import/TableImportDialog.vue

@@ -15,6 +15,7 @@ import { useToast } from "@/composables/useToast";
 import {
   autoMapImportColumns,
   buildTableImportParseOptions,
+  defaultTableImportEmptyStringAsNull,
   formatTableImportElapsed,
   nextTableImportWizardStep,
   previousTableImportWizardStep,
@@ -90,7 +91,7 @@ const titleRow = ref(1);
 const dataStartRow = ref(2);
 const lastDataRow = ref(0);
 const trimValues = ref(false);
-const emptyStringAsNull = ref(true);
+const emptyStringAsNull = ref(defaultTableImportEmptyStringAsNull(sourceFormat.value));
 const selectedSheet = ref("");
 const jsonShape = ref<api.TableImportJsonShape>("auto");
 const previewLimit = ref(50);
@@ -250,7 +251,7 @@ function resetState() {
   dataStartRow.value = 2;
   lastDataRow.value = 0;
   trimValues.value = false;
-  emptyStringAsNull.value = true;
+  emptyStringAsNull.value = defaultTableImportEmptyStringAsNull(sourceFormat.value);
   selectedSheet.value = "";
   jsonShape.value = "auto";
   previewLimit.value = 50;
@@ -519,6 +520,7 @@ function assignSelectedSource(source: string | File) {
   errorMessage.value = "";
   const name = typeof source === "string" ? source : source.name;
   sourceFormat.value = detectFormat(name);
+  emptyStringAsNull.value = defaultTableImportEmptyStringAsNull(sourceFormat.value);
   if (!newTableName.value.trim()) {
     newTableName.value = suggestedTableName(name);
   }
@@ -554,9 +556,11 @@ async function prepareBatchSources(sources: ImportSource[]) {
   errorMessage.value = "";
   const tasks: BatchImportTask[] = [];
   const usedNames = new Set<string>();
+  const formats = sources.map((source) => detectFormat(sourceName(source)));
+  emptyStringAsNull.value = formats.length && formats.every((format) => format === formats[0]) ? defaultTableImportEmptyStringAsNull(formats[0]!) : true;
   try {
-    for (const source of sources) {
-      const format = detectFormat(sourceName(source));
+    for (const [index, source] of sources.entries()) {
+      const format = formats[index]!;
       const initialPreview = await api.previewTableImportFile(source, {
         sourceFormat: format,
         parseOptions: taskParseOptions(format),

apps/desktop/src/lib/__tests__/table/tableImport.spec.ts

@@ -2,6 +2,7 @@ import { describe, expect, it } from "vitest";
 import {
   autoMapImportColumns,
   buildTableImportParseOptions,
+  defaultTableImportEmptyStringAsNull,
   formatTableImportElapsed,
   nextTableImportWizardStep,
   previousTableImportWizardStep,
@@ -13,6 +14,14 @@ import {
 } from "@/lib/table/tableImport";
 
 describe("tableImport", () => {
+  it("preserves explicit empty strings by default for Excel only", () => {
+    expect(defaultTableImportEmptyStringAsNull("excel")).toBe(false);
+    expect(defaultTableImportEmptyStringAsNull("csv")).toBe(true);
+    expect(defaultTableImportEmptyStringAsNull("tsv")).toBe(true);
+    expect(defaultTableImportEmptyStringAsNull("delimited")).toBe(true);
+    expect(defaultTableImportEmptyStringAsNull("json")).toBe(true);
+  });
+
   it("formats import elapsed time for progress and terminal summaries", () => {
     expect(formatTableImportElapsed(0)).toBe("0 ms");
     expect(formatTableImportElapsed(999)).toBe("999 ms");

apps/desktop/src/lib/table/tableImport.ts

@@ -88,6 +88,10 @@ export interface TableImportParseSettings {
   jsonShape: TableImportJsonShape;
 }
 
+export function defaultTableImportEmptyStringAsNull(format: TableImportSourceFormat): boolean {
+  return format !== "excel";
+}
+
 export function buildTableImportParseOptions(settings: TableImportParseSettings): TableImportParseOptions {
   const isDelimited = settings.format === "csv" || settings.format === "tsv" || settings.format === "delimited";
   return {

crates/dbx-core/src/table_import.rs

@@ -6719,6 +6719,42 @@ mod tests {
         assert_xlsx_empty_string_option(TableImportParseOptions::default(), vec![serde_json::Value::Null; 5]);
     }
 
+    #[test]
+    fn dbx_exported_xlsx_round_trip_preserves_empty_strings_when_configured() {
+        let path =
+            std::env::temp_dir().join(format!("dbx-table-import-empty-round-trip-{}.xlsx", uuid::Uuid::new_v4()));
+        let workbook = build_xlsx_workbook_multi(&[XlsxWorksheetData {
+            sheet_name: Some("Data".to_string()),
+            columns: vec!["empty_text".to_string(), "missing_value".to_string()],
+            column_types: vec!["VARCHAR(255)".to_string(), "VARCHAR(255)".to_string()],
+            column_comments: vec![],
+            rows: vec![vec![serde_json::json!(""), serde_json::Value::Null]],
+            numeric_column_right_align: false,
+        }])
+        .unwrap();
+        std::fs::write(&path, workbook).unwrap();
+        let options =
+            TableImportParseOptions { empty_string_as_null: Some(false), ..TableImportParseOptions::default() };
+
+        let parsed = parse_xlsx_file_with_options(&path.to_string_lossy(), &options, 10).unwrap();
+        let (preview, _) = parse_xlsx_preview_file_with_options(&path.to_string_lossy(), &options, 10).unwrap();
+        let (sender, mut receiver) = tokio::sync::mpsc::channel(16);
+        stream_xlsx_rows_to_channel(&path.to_string_lossy(), &options, 500, None, HashSet::new(), false, sender)
+            .unwrap();
+        let mut streamed_rows = Vec::new();
+        while let Some(message) = receiver.blocking_recv() {
+            if let XlsxStreamMessage::Rows(rows) = message.unwrap() {
+                streamed_rows.extend(rows);
+            }
+        }
+
+        let expected = vec![vec![serde_json::json!(""), serde_json::Value::Null]];
+        assert_eq!(parsed.rows, expected);
+        assert_eq!(preview.rows, parsed.rows);
+        assert_eq!(streamed_rows, parsed.rows);
+        let _ = std::fs::remove_file(path);
+    }
+
     #[test]
     fn retains_only_temporal_and_text_target_xlsx_styles() {
         let styles = vec![