From 89d11e9e37d4f68da025e752fae36855c0c32595 Mon Sep 17 00:00:00 2001
From: Oliver <oliver.henry.walters@gmail.com>
Date: Sat, 11 Jul 2026 12:05:14 +1000
Subject: [PATCH] Fix data import bug (#12355)

- Fixes bug where imported data headers get out of sync
- Additional unit tests
---
 src/backend/InvenTree/importer/models.py     |  2 +-
 src/backend/InvenTree/importer/operations.py | 28 ++++++--
 src/backend/InvenTree/importer/tests.py      | 74 ++++++++++++++++++++
 3 files changed, 97 insertions(+), 7 deletions(-)

diff --git a/src/backend/InvenTree/importer/models.py b/src/backend/InvenTree/importer/models.py
index ce64e0677d..8c62ac5d3e 100644
--- a/src/backend/InvenTree/importer/models.py
+++ b/src/backend/InvenTree/importer/models.py
@@ -341,7 +341,7 @@ class DataImportSession(models.Model):
             logger.error('Failed to load data file')
             return
 
-        headers = df.headers
+        headers = importer.operations.normalize_headers(df.headers)
 
         imported_rows = []
 
diff --git a/src/backend/InvenTree/importer/operations.py b/src/backend/InvenTree/importer/operations.py
index 0efca9017c..7334f2badf 100644
--- a/src/backend/InvenTree/importer/operations.py
+++ b/src/backend/InvenTree/importer/operations.py
@@ -72,17 +72,33 @@ def extract_column_names(data_file) -> list:
     """
     data = load_data_file(data_file)
 
-    headers = []
+    return normalize_headers(data.headers)
 
-    for idx, header in enumerate(data.headers):
+
+def normalize_headers(headers) -> list:
+    """Normalize a list of raw column headers extracted from a data file.
+
+    Strips whitespace from each header, and generates a default header
+    for any column that does not have one. Must be used consistently
+    wherever column headers are extracted, so that column names used for
+    field mapping match the column names used when extracting row data.
+
+    Args:
+        headers: List of raw header values (as returned by tablib)
+
+    Returns:
+        List of normalized column names
+    """
+    result = []
+
+    for idx, header in enumerate(headers):
         if header:
-            header = str(header).strip()
-            headers.append(header)
+            result.append(str(header).strip())
         else:
             # If the header is empty, generate a default header
-            headers.append(f'Column {idx + 1}')
+            result.append(f'Column {idx + 1}')
 
-    return headers
+    return result
 
 
 def get_field_label(field) -> Optional[str]:
diff --git a/src/backend/InvenTree/importer/tests.py b/src/backend/InvenTree/importer/tests.py
index bc0fb147b5..1b91baaa8a 100644
--- a/src/backend/InvenTree/importer/tests.py
+++ b/src/backend/InvenTree/importer/tests.py
@@ -77,6 +77,80 @@ class ImporterTest(ImporterMixin, InvenTreeTestCase):
         # Check that the new companies have been created
         self.assertEqual(n + 12, Company.objects.count())
 
+    def test_import_header_whitespace(self):
+        """Test that column headers with leading/trailing whitespace are handled correctly.
+
+        Regression test: column mappings are built from *stripped* header names, but
+        row data was previously extracted using the *raw* (unstripped) headers. If a
+        header had surrounding whitespace, the mapped column name would never match a
+        key in the row data, so that field was silently skipped during import.
+        """
+        from company.models import Company
+
+        n = Company.objects.count()
+
+        # Pad each header in the source file with extra whitespace
+        raw = self.helper_file('companies.csv').read()
+        lines = raw.splitlines()
+        headers = [f'  {header}  ' for header in lines[0].split(',')]
+        padded_data = '\n'.join([','.join(headers), *lines[1:]])
+
+        data_file = ContentFile(padded_data, 'companies_whitespace.csv')
+
+        session = DataImportSession.objects.create(
+            data_file=data_file, model_type='company'
+        )
+
+        session.extract_columns()
+
+        # Extracted column names should have whitespace stripped
+        for col in session.columns:
+            self.assertEqual(col, col.strip())
+
+        # Field mappings should be created correctly, against the *stripped* names
+        for field, col in [
+            ('website', 'Website'),
+            ('is_customer', 'Is customer'),
+            ('phone', 'Phone number'),
+            ('description', 'Company description'),
+            ('active', 'Active'),
+        ]:
+            self.assertTrue(
+                session.column_mappings.filter(field=field, column=col).exists()
+            )
+
+        # Run the data import
+        session.import_data()
+        self.assertEqual(session.rows.count(), 12)
+
+        rows = list(session.rows.all())
+
+        # Row data keys must be stripped, to match the mapped column names
+        for row in rows:
+            for key in row.row_data:
+                self.assertEqual(key, key.strip())
+
+        # Find the 'Arrow' row, and check that mapped fields were actually extracted
+        arrow_row = next(row for row in rows if row.data.get('name') == 'Arrow')
+        self.assertEqual(arrow_row.data.get('website'), 'https://www.arrow.com/')
+        self.assertEqual(arrow_row.data.get('description'), 'Arrow Electronics')
+        self.assertEqual(arrow_row.data.get('is_customer'), False)
+
+        for row in rows:
+            self.assertIsNotNone(row.data.get('name', None))
+            self.assertTrue(row.valid)
+
+            row.validate(commit=True)
+            self.assertTrue(row.complete)
+
+        # All rows accepted: rows and mappings are cleared, session is retained
+        session.refresh_from_db()
+        self.assertEqual(session.rows.count(), 0)
+        self.assertEqual(session.column_mappings.count(), 0)
+
+        # Check that the new companies have been created
+        self.assertEqual(n + 12, Company.objects.count())
+
     def test_field_defaults(self):
         """Test default field values."""