Detect CSV delimiter when one is not supplied (fixes #622)

src/tablib/formats/_csv.py

@@ -34,12 +34,53 @@ def export_set(cls, dataset, **kwargs):
         stream = cls.export_stream_set(dataset, **kwargs)
         return stream.getvalue()
 
+    #: Delimiters considered when detecting the dialect of a CSV stream.
+    CANDIDATE_DELIMITERS = (',', ';', '	', '|', ':')
+
+    @classmethod
+    def sniff_delimiter(cls, in_stream):
+        """Detect the delimiter used by a CSV stream.
+
+        Only a small set of common delimiters is considered, and the
+        default comma is used whenever the dialect cannot be determined
+        or only one column is present.
+        """
+        try:
+            position = in_stream.tell()
+        except (AttributeError, OSError):
+            position = None
+
+        try:
+            sample = in_stream.read(2048)
+        except Exception:
+            return cls.DEFAULT_DELIMITER
+
+        if position is not None:
+            in_stream.seek(position)
+
+        for candidate in cls.CANDIDATE_DELIMITERS:
+            if candidate not in sample:
+                continue
+            try:
+                dialect = csv.Sniffer().sniff(sample, delimiters=candidate)
+            except Exception:
+                continue
+            if dialect.delimiter == candidate:
+                return candidate
+
+        return cls.DEFAULT_DELIMITER
+
     @classmethod
-    def import_set(cls, dset, in_stream, headers=True, skip_lines=0, **kwargs):
+    def import_set(
+            cls, dset, in_stream, headers=True, skip_lines=0, **kwargs
+    ):
         """Returns dataset from CSV stream."""
 
         dset.wipe()
 
+        if 'delimiter' not in kwargs and not skip_lines:
+            kwargs['delimiter'] = cls.sniff_delimiter(in_stream)
+
         kwargs.setdefault('delimiter', cls.DEFAULT_DELIMITER)
 
         rows = csv.reader(in_stream, **kwargs)

tests/test_tablib.py

@@ -1165,14 +1165,79 @@ def test_csv_formatter_support_kwargs(self):
         _csv = data.export('csv', **kwargs)
         self.assertEqual(expected, _csv)
 
-        # the import works but consider default delimiter=','
+        # the delimiter is now detected, so no explicit delimiter is needed
         d1 = tablib.import_set(_csv, format="csv")
-        self.assertEqual(1, len(d1.headers))
+        self.assertEqual(3, len(d1.headers))
 
         d2 = tablib.import_set(_csv, format="csv", **kwargs)
         self.assertEqual(3, len(d2.headers))
 
 
+    def test_csv_import_set_sniffs_colon_delimiter(self):
+        """Colon-separated CSV is detected without an explicit delimiter."""
+        csv_text = (
+            'test:test2\n'
+            'value1:value2\n'
+        )
+        data.csv = csv_text
+        self.assertEqual(('test', 'test2'), tuple(data.headers))
+        self.assertEqual(('value1', 'value2'), tuple(data[0]))
+
+    def test_csv_import_set_sniffs_semicolon_delimiter(self):
+        """Semicolon-separated CSV is detected without an explicit
+        delimiter."""
+        csv_text = (
+            'a;b;c\n'
+            '1;2;3\n'
+            '4;5;6\n'
+        )
+        data.csv = csv_text
+        self.assertEqual(('a', 'b', 'c'), tuple(data.headers))
+        self.assertEqual(('1', '2', '3'), tuple(data[0]))
+        self.assertEqual(('4', '5', '6'), tuple(data[1]))
+
+    def test_csv_import_set_sniffs_pipe_delimiter(self):
+        """Pipe-separated CSV is detected without an explicit delimiter."""
+        csv_text = (
+            'a|b|c\n'
+            '1|2|3\n'
+        )
+        data.csv = csv_text
+        self.assertEqual(('a', 'b', 'c'), tuple(data.headers))
+        self.assertEqual(('1', '2', '3'), tuple(data[0]))
+
+    def test_csv_import_set_sniffs_tab_delimiter(self):
+        """Tab-separated CSV is detected without an explicit delimiter."""
+        csv_text = (
+            'a\tb\tc\n'
+            '1\t2\t3\n'
+        )
+        data.csv = csv_text
+        self.assertEqual(('a', 'b', 'c'), tuple(data.headers))
+        self.assertEqual(('1', '2', '3'), tuple(data[0]))
+
+    def test_csv_import_set_single_column_falls_back_to_comma(self):
+        """A single-column CSV has no delimiter to detect, so it falls back
+        to the default without raising."""
+        csv_text = (
+            'name\n'
+            'Alice\n'
+            'Bob\n'
+        )
+        data.csv = csv_text
+        self.assertEqual(('name',), tuple(data.headers))
+        self.assertEqual(('Alice',), tuple(data[0]))
+
+    def test_csv_import_set_explicit_delimiter_still_wins(self):
+        """An explicitly supplied delimiter takes precedence over
+        detection."""
+        csv_text = (
+            'a;b;c\n'
+            '1;2;3\n'
+        )
+        data.set_csv(csv_text, delimiter='|')
+        self.assertEqual(('a;b;c',), tuple(data.headers))
+
 class TSVTests(BaseTestCase):
     def test_tsv_import_set(self):
         """Generate and import TSV set serialization."""