Merge pull request pandas-dev#9834 from evanpw/skip_rows_blank

BUG: skiprows doesn't handle blank lines properly when engine='c'

Author: jreback

doc/source/whatsnew/v0.16.1.txt

@@ -95,3 +95,4 @@ Bug Fixes
 - Bug in ``FloatArrayFormatter`` where decision boundary for displaying "small" floats in decimal format is off by one order of magnitude for a given display.precision (:issue:`9764`)
 
 - Fixed bug where ``DataFrame.plot()`` raised an error when both ``color`` and ``style`` keywords were passed and there was no color symbol in the style strings (:issue:`9671`)
+- Bug in ``read_csv`` and ``read_table`` when using ``skip_rows`` parameter if blank lines are present. (:issue:`9832`)

pandas/io/tests/test_parsers.py

@@ -839,6 +839,28 @@ def test_deep_skiprows(self):
         condensed_data = self.read_csv(StringIO(condensed_text))
         tm.assert_frame_equal(data, condensed_data)
 
+    def test_skiprows_blank(self):
+        # GH 9832
+        text = """#foo,a,b,c
+#foo,a,b,c
+
+#foo,a,b,c
+#foo,a,b,c
+
+1/1/2000,1.,2.,3.
+1/2/2000,4,5,6
+1/3/2000,7,8,9
+"""
+        data = self.read_csv(StringIO(text), skiprows=6, header=None,
+                              index_col=0, parse_dates=True)
+
+        expected = DataFrame(np.arange(1., 10.).reshape((3, 3)),
+                             columns=[1, 2, 3],
+                             index=[datetime(2000, 1, 1), datetime(2000, 1, 2),
+                                    datetime(2000, 1, 3)])
+        expected.index.name = 0
+        tm.assert_frame_equal(data, expected)
+
     def test_detect_string_na(self):
         data = """A,B
 foo,bar

pandas/src/parser/tokenizer.c

@@ -757,11 +757,9 @@ int tokenize_delimited(parser_t *self, size_t line_limit)
         case START_RECORD:
             // start of record
             if (skip_this_line(self, self->file_lines)) {
+                self->state = SKIP_LINE;
                 if (c == '\n') {
-                    END_LINE()
-                }
-                else {
-                    self->state = SKIP_LINE;
+                    END_LINE();
                 }
                 break;
             }
@@ -1093,11 +1091,9 @@ int tokenize_delim_customterm(parser_t *self, size_t line_limit)
         case START_RECORD:
             // start of record
             if (skip_this_line(self, self->file_lines)) {
+                self->state = SKIP_LINE;
                 if (c == self->lineterminator) {
-                    END_LINE()
-                }
-                else {
-                    self->state = SKIP_LINE;
+                    END_LINE();
                 }
                 break;
             }
@@ -1391,11 +1387,9 @@ int tokenize_whitespace(parser_t *self, size_t line_limit)
         case START_RECORD:
             // start of record
             if (skip_this_line(self, self->file_lines)) {
+                self->state = SKIP_LINE;
                 if (c == '\n') {
-                    END_LINE()
-                }
-                else {
-                    self->state = SKIP_LINE;
+                    END_LINE();
                 }
                 break;
             } else  if (c == '\n') {