From 0d4b418ddf1c8e4c00c63bbdd719781a7f507052 Mon Sep 17 00:00:00 2001
From: Liam <liam@cpan.org>
Date: Sat, 28 Oct 2017 18:22:00 -0500
Subject: [PATCH 1/4] io.html.read_html(): rewind seekable io objects when
 parsers fail

If lxml has read to the end of a file and then errored,
bs4/html5lib won't rewind it before trying to parse again,
and will throw a `ValueError: No text parsed from document`.

This patch fixes this issue, by rewinding the file object
when a parser fails. If the object was IO-ish but not seekable,
we throw an error notifying the user and asking them to try
a different flavor.
---
 pandas/io/html.py | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/pandas/io/html.py b/pandas/io/html.py
index 6f98683a1bff1..e1636d8007345 100644
--- a/pandas/io/html.py
+++ b/pandas/io/html.py
@@ -742,6 +742,18 @@ def _parse(flavor, io, match, attrs, encoding, **kwargs):
         try:
             tables = p.parse_tables()
         except Exception as caught:
+            # if `io` is an io-like object, check if it's seekable
+            # and try to rewind it before trying the next parser
+            if hasattr(io, 'seekable') and io.seekable():
+                io.seek(0)
+            elif hasattr(io, 'seekable') and not io.seekable():
+                # if we couldn't rewind it, let the user know
+                raise ValueError('The flavor {} failed to parse your input. '
+                                 'Since you passed a non-rewindable file '
+                                 'object, we can\'t rewind it to try '
+                                 'another parser. Try read_html() with a '
+                                 'different flavor.'.format(flav))
+
             retained = caught
         else:
             break

From 9e813173336127dca19611f9d2d22c0c99b48e52 Mon Sep 17 00:00:00 2001
From: Liam <liam@cpan.org>
Date: Sat, 28 Oct 2017 19:00:49 -0500
Subject: [PATCH 2/4] test: read_html() parse error interaction with unseekable
 IO

---
 pandas/tests/io/test_html.py | 24 ++++++++++++++++++++++++
 1 file changed, 24 insertions(+)

diff --git a/pandas/tests/io/test_html.py b/pandas/tests/io/test_html.py
index 8dfae2733ef20..c481870d5edd0 100644
--- a/pandas/tests/io/test_html.py
+++ b/pandas/tests/io/test_html.py
@@ -967,3 +967,27 @@ def test_importcheck_thread_safety():
     while helper_thread1.is_alive() or helper_thread2.is_alive():
         pass
     assert None is helper_thread1.err is helper_thread2.err
+
+
+def test_parse_failure_unseekable():
+    # Issue #17975
+    _skip_if_no('lxml')
+
+    class UnseekableStringIO(StringIO):
+        def seekable(self):
+            return False
+
+    good = UnseekableStringIO('''
+        <table><tr><td>spam<br />eggs</td></tr></table>''')
+    bad = UnseekableStringIO('''
+        <table><tr><td>spam<foobr />eggs</td></tr></table>''')
+
+    assert read_html(good)
+    assert read_html(bad, flavor='bs4')
+
+    bad.seek(0)
+
+    with pytest.raises(ValueError,
+                       match='passed a non-rewindable file object'):
+        read_html(bad)
+

From 58b46474f7fa2afa31c4d0a7dbf33c52c1758d4c Mon Sep 17 00:00:00 2001
From: Liam <liam@cpan.org>
Date: Sat, 28 Oct 2017 19:55:02 -0500
Subject: [PATCH 3/4] test: read_html() rewinding after parse errors

---
 pandas/tests/io/test_html.py | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)

diff --git a/pandas/tests/io/test_html.py b/pandas/tests/io/test_html.py
index c481870d5edd0..399cac905967e 100644
--- a/pandas/tests/io/test_html.py
+++ b/pandas/tests/io/test_html.py
@@ -991,3 +991,29 @@ def seekable(self):
                        match='passed a non-rewindable file object'):
         read_html(bad)
 
+
+def test_parse_failure_rewinds():
+    # Issue #17975
+    _skip_if_no('lxml')
+
+    class MockFile(object):
+        def __init__(self, data):
+            self.data = data
+            self.at_end = False
+
+        def read(self, size=None):
+            data = '' if self.at_end else self.data
+            self.at_end = True
+            return data
+
+        def seek(self, offset):
+            self.at_end = False
+
+        def seekable(self):
+            return True
+
+    good = MockFile('<table><tr><td>spam<br />eggs</td></tr></table>')
+    bad = MockFile('<table><tr><td>spam<foobr />eggs</td></tr></table>')
+
+    assert read_html(good)
+    assert read_html(bad)

From a85cd4217cc18e5e684ae9e8f75a2acaa948e739 Mon Sep 17 00:00:00 2001
From: Liam <liam@cpan.org>
Date: Sat, 28 Oct 2017 20:07:08 -0500
Subject: [PATCH 4/4] doc: whatsnew for read_html() rewinding fixes

---
 doc/source/whatsnew/v0.22.0.txt | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/doc/source/whatsnew/v0.22.0.txt b/doc/source/whatsnew/v0.22.0.txt
index cbd094ec4ef49..1a08d48f3948b 100644
--- a/doc/source/whatsnew/v0.22.0.txt
+++ b/doc/source/whatsnew/v0.22.0.txt
@@ -103,7 +103,7 @@ Indexing
 I/O
 ^^^
 
--
+- :func:`read_html` now rewinds seekable IO objects after parse failure, before attempting to parse with a new parser. If a parser errors and the object is non-seekable, an informative error is raised suggesting the use of a different parser (:issue:`17975`)
 -
 -