Add formatter option for comma first notation (fixes andialbrecht#141).

andialbrecht · andialbrecht · commit 4d05b441fcb8 · 2016-09-14T18:42:22.000+02:00
diff --git a/CHANGELOG b/CHANGELOG
@@ -1,6 +1,11 @@
 Development Version
 -------------------
 
+Enhancements
+
+* Add comma_first option: When splitting list "comma first" notation
+  is used (issue141).
+
 Bug Fixes
 
 * Fix parsing of incomplete AS (issue284, by vmuriart).
diff --git a/sqlparse/cli.py b/sqlparse/cli.py
@@ -118,6 +118,13 @@ def create_parser():
         type=int,
         help='Column after which lists should be wrapped')
 
+    group.add_argument(
+        '--comma_first',
+        dest='comma_first',
+        default=False,
+        type=bool,
+        help='Insert linebreak before comma (default False)')
+
     return parser
 
 
diff --git a/sqlparse/filters/reindent.py b/sqlparse/filters/reindent.py
@@ -11,13 +11,15 @@
 
 
 class ReindentFilter(object):
-    def __init__(self, width=2, char=' ', wrap_after=0, n='\n'):
+    def __init__(self, width=2, char=' ', wrap_after=0, n='\n',
+                 comma_first=False):
         self.n = n
         self.width = width
         self.char = char
         self.indent = 0
         self.offset = 0
         self.wrap_after = wrap_after
+        self.comma_first = comma_first
         self._curr_stmt = None
         self._last_stmt = None
 
@@ -123,7 +125,20 @@ def _process_identifierlist(self, tlist):
                     # Add 1 for the "," separator
                     position += len(token.value) + 1
                     if position > (self.wrap_after - self.offset):
+                        if self.comma_first:
+                            _, comma = tlist.token_prev(
+                                tlist.token_index(token))
+                            if comma is None:
+                                continue
+                            token = comma
                         tlist.insert_before(token, self.nl())
+                        if self.comma_first:
+                            _, ws = tlist.token_next(
+                                tlist.token_index(token), skip_ws=False)
+                            if (ws is not None
+                                and not ws.ttype is T.Text.Whitespace):
+                                tlist.insert_after(
+                                    token, sql.Token(T.Whitespace, ' '))
                         position = 0
         self._process_default(tlist)
 
diff --git a/sqlparse/formatter.py b/sqlparse/formatter.py
@@ -97,6 +97,11 @@ def validate_options(options):
         raise SQLParseError('wrap_after requires a positive integer')
     options['wrap_after'] = wrap_after
 
+    comma_first = options.get('comma_first', False)
+    if comma_first not in [True, False]:
+        raise SQLParseError('comma_first requires a boolean value')
+    options['comma_first'] = comma_first
+
     right_margin = options.get('right_margin')
     if right_margin is not None:
         try:
@@ -148,7 +153,8 @@ def build_filter_stack(stack, options):
         stack.stmtprocess.append(
             filters.ReindentFilter(char=options['indent_char'],
                                    width=options['indent_width'],
-                                   wrap_after=options['wrap_after']))
+                                   wrap_after=options['wrap_after'],
+                                   comma_first=options['comma_first']))
 
     if options.get('reindent_aligned', False):
         stack.enable_grouping()
diff --git a/tests/test_format.py b/tests/test_format.py
@@ -322,6 +322,8 @@ def test_option(self):
             sqlparse.format('foo', reindent=True, wrap_after='foo')
         with pytest.raises(SQLParseError):
             sqlparse.format('foo', reindent=True, wrap_after=-12)
+        with pytest.raises(SQLParseError):
+            sqlparse.format('foo', reindent=True, comma_first='foo')
 
     def test_stmts(self):
         f = lambda sql: sqlparse.format(sql, reindent=True)
@@ -428,6 +430,19 @@ def test_identifier_list_with_wrap_after(self):
             'from table1, table2',
             'where 1 = 2'])
 
+    def test_identifier_list_comment_first(self):
+        f = lambda sql: sqlparse.format(sql, reindent=True, comma_first=True)
+        # not the 3: It cleans up whitespace too!
+        s = 'select foo, bar, baz from table where foo in (1, 2,3)'
+        assert f(s) == '\n'.join([
+            'select foo',
+            '       , bar',
+            '       , baz',
+            'from table',
+            'where foo in (1',
+            '              , 2',
+            '              , 3)'])
+
     def test_identifier_list_with_functions(self):
         f = lambda sql: sqlparse.format(sql, reindent=True)
         s = ("select 'abc' as foo, coalesce(col1, col2)||col3 as bar,"
