diff --git a/README.md b/README.md
index 64fa64f7..7ca64778 100644
--- a/README.md
+++ b/README.md
@@ -299,8 +299,9 @@ Changelog
 Unreleased
 -----
 
-- ...
-
+- Refactor parser to fix parsing inconsistencies ([@bbc2])([#170]).
+  - Interpret escapes as control characters only in double-quoted strings.
+  - Interpret `#` as start of comment only if preceded by whitespace.
 
 0.10.2
 -----
@@ -428,6 +429,7 @@ Unreleased
 [#172]: https://github.com/theskumar/python-dotenv/issues/172
 [#121]: https://github.com/theskumar/python-dotenv/issues/121
 [#176]: https://github.com/theskumar/python-dotenv/issues/176
+[#170]: https://github.com/theskumar/python-dotenv/issues/170
 
 [@asyncee]: https://github.com/asyncee
 [@greyli]: https://github.com/greyli
diff --git a/setup.cfg b/setup.cfg
index 7f784591..f0847b32 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -5,6 +5,9 @@ universal = 1
 max-line-length = 120
 exclude = .tox,.git,docs,venv,.venv
 
+[mypy]
+ignore_missing_imports = true
+
 [metadata]
 description-file = README.rst
 
diff --git a/src/dotenv/compat.py b/src/dotenv/compat.py
index 99ffb39b..1a145345 100644
--- a/src/dotenv/compat.py
+++ b/src/dotenv/compat.py
@@ -1,4 +1,5 @@
 import sys
+from typing import Text
 
 if sys.version_info >= (3, 0):
     from io import StringIO  # noqa
@@ -6,3 +7,17 @@
     from StringIO import StringIO  # noqa
 
 PY2 = sys.version_info[0] == 2  # type: bool
+
+
+def to_text(string):
+    # type: (str) -> Text
+    """
+    Make a string Unicode if it isn't already.
+
+    This is useful for defining raw unicode strings because `ur"foo"` isn't valid in
+    Python 3.
+    """
+    if PY2:
+        return string.decode("utf-8")
+    else:
+        return string
diff --git a/src/dotenv/main.py b/src/dotenv/main.py
index 08122825..5b619b11 100644
--- a/src/dotenv/main.py
+++ b/src/dotenv/main.py
@@ -1,7 +1,6 @@
 # -*- coding: utf-8 -*-
 from __future__ import absolute_import, print_function, unicode_literals
 
-import codecs
 import io
 import os
 import re
@@ -9,13 +8,14 @@
 import sys
 from subprocess import Popen
 import tempfile
-from typing import (Any, Dict, Iterator, List, Match, NamedTuple, Optional,  # noqa
-                    Pattern, Union, TYPE_CHECKING, Text, IO, Tuple)  # noqa
+from typing import (Dict, Iterator, List, Match, Optional,  # noqa
+                    Pattern, Union, TYPE_CHECKING, Text, IO, Tuple)
 import warnings
 from collections import OrderedDict
 from contextlib import contextmanager
 
 from .compat import StringIO, PY2
+from .parser import parse_stream
 
 if TYPE_CHECKING:  # pragma: no cover
     if sys.version_info >= (3, 6):
@@ -30,84 +30,6 @@
 
 __posix_variable = re.compile(r'\$\{[^\}]*\}')  # type: Pattern[Text]
 
-_binding = re.compile(
-    r"""
-        (
-            \s*                     # leading whitespace
-            (?:export{0}+)?         # export
-
-            ( '[^']+'               # single-quoted key
-            | [^=\#\s]+             # or unquoted key
-            )?
-
-            (?:
-                (?:{0}*={0}*)       # equal sign
-
-                ( '(?:\\'|[^'])*'   # single-quoted value
-                | "(?:\\"|[^"])*"   # or double-quoted value
-                | [^\#\r\n]*        # or unquoted value
-                )
-            )?
-
-            \s*                     # trailing whitespace
-            (?:\#[^\r\n]*)?         # comment
-            (?:\r|\n|\r\n)?         # newline
-        )
-    """.format(r'[^\S\r\n]'),
-    re.MULTILINE | re.VERBOSE,
-)  # type: Pattern[Text]
-
-_escape_sequence = re.compile(r"\\[\\'\"abfnrtv]")  # type: Pattern[Text]
-
-
-Binding = NamedTuple("Binding", [("key", Optional[Text]),
-                                 ("value", Optional[Text]),
-                                 ("original", Text)])
-
-
-def decode_escapes(string):
-    # type: (Text) -> Text
-    def decode_match(match):
-        # type: (Match[Text]) -> Text
-        return codecs.decode(match.group(0), 'unicode-escape')  # type: ignore
-
-    return _escape_sequence.sub(decode_match, string)
-
-
-def is_surrounded_by(string, char):
-    # type: (Text, Text) -> bool
-    return (
-        len(string) > 1
-        and string[0] == string[-1] == char
-    )
-
-
-def parse_binding(string, position):
-    # type: (Text, int) -> Tuple[Binding, int]
-    match = _binding.match(string, position)
-    assert match is not None
-    (matched, key, value) = match.groups()
-    if key is None or value is None:
-        key = None
-        value = None
-    else:
-        value_quoted = is_surrounded_by(value, "'") or is_surrounded_by(value, '"')
-        if value_quoted:
-            value = decode_escapes(value[1:-1])
-        else:
-            value = value.strip()
-    return (Binding(key=key, value=value, original=matched), match.end())
-
-
-def parse_stream(stream):
-    # type:(IO[Text]) -> Iterator[Binding]
-    string = stream.read()
-    position = 0
-    length = len(string)
-    while position < length:
-        (binding, position) = parse_binding(string, position)
-        yield binding
-
 
 def to_env(text):
     # type: (Text) -> str
diff --git a/src/dotenv/parser.py b/src/dotenv/parser.py
new file mode 100644
index 00000000..b63cb3a0
--- /dev/null
+++ b/src/dotenv/parser.py
@@ -0,0 +1,147 @@
+import codecs
+import re
+from typing import (IO, Iterator, Match, NamedTuple, Optional, Pattern,  # noqa
+                    Sequence, Text)
+
+from .compat import to_text
+
+
+def make_regex(string, extra_flags=0):
+    # type: (str, int) -> Pattern[Text]
+    return re.compile(to_text(string), re.UNICODE | extra_flags)
+
+
+_whitespace = make_regex(r"\s*", extra_flags=re.MULTILINE)
+_export = make_regex(r"(?:export[^\S\r\n]+)?")
+_single_quoted_key = make_regex(r"'([^']+)'")
+_unquoted_key = make_regex(r"([^=\#\s]+)")
+_equal_sign = make_regex(r"[^\S\r\n]*=[^\S\r\n]*")
+_single_quoted_value = make_regex(r"'((?:\\'|[^'])*)'")
+_double_quoted_value = make_regex(r'"((?:\\"|[^"])*)"')
+_unquoted_value_part = make_regex(r"([^ \r\n]*)")
+_comment = make_regex(r"(?:\s*#[^\r\n]*)?")
+_end_of_line = make_regex(r"[^\S\r\n]*(?:\r\n|\n|\r)?")
+_rest_of_line = make_regex(r"[^\r\n]*(?:\r|\n|\r\n)?")
+_double_quote_escapes = make_regex(r"\\[\\'\"abfnrtv]")
+_single_quote_escapes = make_regex(r"\\[\\']")
+
+Binding = NamedTuple("Binding", [("key", Optional[Text]),
+                                 ("value", Optional[Text]),
+                                 ("original", Text)])
+
+
+class Error(Exception):
+    pass
+
+
+class Reader:
+    def __init__(self, stream):
+        # type: (IO[Text]) -> None
+        self.string = stream.read()
+        self.position = 0
+        self.mark = 0
+
+    def has_next(self):
+        # type: () -> bool
+        return self.position < len(self.string)
+
+    def set_mark(self):
+        # type: () -> None
+        self.mark = self.position
+
+    def get_marked(self):
+        # type: () -> Text
+        return self.string[self.mark:self.position]
+
+    def peek(self, count):
+        # type: (int) -> Text
+        return self.string[self.position:self.position + count]
+
+    def read(self, count):
+        # type: (int) -> Text
+        result = self.string[self.position:self.position + count]
+        if len(result) < count:
+            raise Error("read: End of string")
+        self.position += count
+        return result
+
+    def read_regex(self, regex):
+        # type: (Pattern[Text]) -> Sequence[Text]
+        match = regex.match(self.string, self.position)
+        if match is None:
+            raise Error("read_regex: Pattern not found")
+        self.position = match.end()
+        return match.groups()
+
+
+def decode_escapes(regex, string):
+    # type: (Pattern[Text], Text) -> Text
+    def decode_match(match):
+        # type: (Match[Text]) -> Text
+        return codecs.decode(match.group(0), 'unicode-escape')  # type: ignore
+
+    return regex.sub(decode_match, string)
+
+
+def parse_key(reader):
+    # type: (Reader) -> Text
+    char = reader.peek(1)
+    if char == "'":
+        (key,) = reader.read_regex(_single_quoted_key)
+    else:
+        (key,) = reader.read_regex(_unquoted_key)
+    return key
+
+
+def parse_unquoted_value(reader):
+    # type: (Reader) -> Text
+    value = u""
+    while True:
+        (part,) = reader.read_regex(_unquoted_value_part)
+        value += part
+        after = reader.peek(2)
+        if len(after) < 2 or after[0] in u"\r\n" or after[1] in u" #\r\n":
+            return value
+        value += reader.read(2)
+
+
+def parse_value(reader):
+    # type: (Reader) -> Text
+    char = reader.peek(1)
+    if char == u"'":
+        (value,) = reader.read_regex(_single_quoted_value)
+        return decode_escapes(_single_quote_escapes, value)
+    elif char == u'"':
+        (value,) = reader.read_regex(_double_quoted_value)
+        return decode_escapes(_double_quote_escapes, value)
+    elif char in (u"", u"\n", u"\r"):
+        return u""
+    else:
+        return parse_unquoted_value(reader)
+
+
+def parse_binding(reader):
+    # type: (Reader) -> Binding
+    reader.set_mark()
+    try:
+        reader.read_regex(_whitespace)
+        reader.read_regex(_export)
+        key = parse_key(reader)
+        reader.read_regex(_equal_sign)
+        value = parse_value(reader)
+        reader.read_regex(_comment)
+        reader.read_regex(_end_of_line)
+        return Binding(key=key, value=value, original=reader.get_marked())
+    except Error:
+        reader.read_regex(_rest_of_line)
+        return Binding(key=None, value=None, original=reader.get_marked())
+
+
+def parse_stream(stream):
+    # type:(IO[Text]) -> Iterator[Binding]
+    reader = Reader(stream)
+    while reader.has_next():
+        try:
+            yield parse_binding(reader)
+        except Error:
+            return
diff --git a/tests/test_core.py b/tests/test_core.py
index daf0f59e..349c58b8 100644
--- a/tests/test_core.py
+++ b/tests/test_core.py
@@ -13,7 +13,6 @@
 
 from dotenv import dotenv_values, find_dotenv, load_dotenv, set_key
 from dotenv.compat import PY2, StringIO
-from dotenv.main import Binding, parse_stream
 
 
 @contextlib.contextmanager
@@ -26,81 +25,6 @@ def restore_os_environ():
         os.environ.update(environ)
 
 
-@pytest.mark.parametrize("test_input,expected", [
-    ("", []),
-    ("a=b", [Binding(key="a", value="b", original="a=b")]),
-    ("'a'=b", [Binding(key="'a'", value="b", original="'a'=b")]),
-    ("[=b", [Binding(key="[", value="b", original="[=b")]),
-    (" a = b ", [Binding(key="a", value="b", original=" a = b ")]),
-    ("export a=b", [Binding(key="a", value="b", original="export a=b")]),
-    (" export 'a'=b", [Binding(key="'a'", value="b", original=" export 'a'=b")]),
-    (" export 'a'=b", [Binding(key="'a'", value="b", original=" export 'a'=b")]),
-    ("# a=b", [Binding(key=None, value=None, original="# a=b")]),
-    ('a=b # comment', [Binding(key="a", value="b", original="a=b # comment")]),
-    ("a=b space ", [Binding(key="a", value="b space", original="a=b space ")]),
-    ("a='b space '", [Binding(key="a", value="b space ", original="a='b space '")]),
-    ('a="b space "', [Binding(key="a", value="b space ", original='a="b space "')]),
-    ("export export_a=1", [Binding(key="export_a", value="1", original="export export_a=1")]),
-    ("export port=8000", [Binding(key="port", value="8000", original="export port=8000")]),
-    ('a="b\nc"', [Binding(key="a", value="b\nc", original='a="b\nc"')]),
-    ("a='b\nc'", [Binding(key="a", value="b\nc", original="a='b\nc'")]),
-    ('a="b\nc"', [Binding(key="a", value="b\nc", original='a="b\nc"')]),
-    ('a="b\\nc"', [Binding(key="a", value='b\nc', original='a="b\\nc"')]),
-    ('a="b\\"c"', [Binding(key="a", value='b"c', original='a="b\\"c"')]),
-    ("a='b\\'c'", [Binding(key="a", value="b'c", original="a='b\\'c'")]),
-    ("a=à", [Binding(key="a", value="à", original="a=à")]),
-    ('a="à"', [Binding(key="a", value="à", original='a="à"')]),
-    ('garbage', [Binding(key=None, value=None, original="garbage")]),
-    (
-        "a=b\nc=d",
-        [
-            Binding(key="a", value="b", original="a=b\n"),
-            Binding(key="c", value="d", original="c=d"),
-        ],
-    ),
-    (
-        "a=b\r\nc=d",
-        [
-            Binding(key="a", value="b", original="a=b\r\n"),
-            Binding(key="c", value="d", original="c=d"),
-        ],
-    ),
-    (
-        'a=\nb=c',
-        [
-            Binding(key="a", value='', original='a=\n'),
-            Binding(key="b", value='c', original="b=c"),
-        ]
-    ),
-    (
-        'a="\nb=c',
-        [
-            Binding(key="a", value='"', original='a="\n'),
-            Binding(key="b", value='c', original="b=c"),
-        ]
-    ),
-    (
-        '# comment\na="b\nc"\nd=e\n',
-        [
-            Binding(key=None, value=None, original="# comment\n"),
-            Binding(key="a", value="b\nc", original='a="b\nc"\n'),
-            Binding(key="d", value="e", original="d=e\n"),
-        ],
-    ),
-    (
-        'garbage[%$#\na=b',
-        [
-            Binding(key=None, value=None, original="garbage[%$#\n"),
-            Binding(key="a", value="b", original='a=b'),
-        ],
-    ),
-])
-def test_parse_stream(test_input, expected):
-    result = parse_stream(StringIO(test_input))
-
-    assert list(result) == expected
-
-
 def test_warns_if_file_does_not_exist():
     with warnings.catch_warnings(record=True) as w:
         load_dotenv('.does_not_exist', verbose=True)
diff --git a/tests/test_parser.py b/tests/test_parser.py
new file mode 100644
index 00000000..f191f902
--- /dev/null
+++ b/tests/test_parser.py
@@ -0,0 +1,88 @@
+# -*- coding: utf-8 -*-
+import pytest
+
+from dotenv.compat import StringIO
+from dotenv.parser import Binding, parse_stream
+
+
+@pytest.mark.parametrize("test_input,expected", [
+    (u"", []),
+    (u"a=b", [Binding(key=u"a", value=u"b", original=u"a=b")]),
+    (u"'a'=b", [Binding(key=u"a", value=u"b", original=u"'a'=b")]),
+    (u"[=b", [Binding(key=u"[", value=u"b", original=u"[=b")]),
+    (u" a = b ", [Binding(key=u"a", value=u"b", original=u" a = b ")]),
+    (u"export a=b", [Binding(key=u"a", value=u"b", original=u"export a=b")]),
+    (u" export 'a'=b", [Binding(key=u"a", value=u"b", original=u" export 'a'=b")]),
+    (u"# a=b", [Binding(key=None, value=None, original=u"# a=b")]),
+    (u"a=b#c", [Binding(key=u"a", value=u"b#c", original=u"a=b#c")]),
+    (u'a=b # comment', [Binding(key=u"a", value=u"b", original=u"a=b # comment")]),
+    (u"a=b space ", [Binding(key=u"a", value=u"b space", original=u"a=b space ")]),
+    (u"a='b space '", [Binding(key=u"a", value=u"b space ", original=u"a='b space '")]),
+    (u'a="b space "', [Binding(key=u"a", value=u"b space ", original=u'a="b space "')]),
+    (u"export export_a=1", [Binding(key=u"export_a", value=u"1", original=u"export export_a=1")]),
+    (u"export port=8000", [Binding(key=u"port", value=u"8000", original=u"export port=8000")]),
+    (u'a="b\nc"', [Binding(key=u"a", value=u"b\nc", original=u'a="b\nc"')]),
+    (u"a='b\nc'", [Binding(key=u"a", value=u"b\nc", original=u"a='b\nc'")]),
+    (u'a="b\nc"', [Binding(key=u"a", value=u"b\nc", original=u'a="b\nc"')]),
+    (u'a="b\\nc"', [Binding(key=u"a", value=u'b\nc', original=u'a="b\\nc"')]),
+    (u"a='b\\nc'", [Binding(key=u"a", value=u'b\\nc', original=u"a='b\\nc'")]),
+    (u'a="b\\"c"', [Binding(key=u"a", value=u'b"c', original=u'a="b\\"c"')]),
+    (u"a='b\\'c'", [Binding(key=u"a", value=u"b'c", original=u"a='b\\'c'")]),
+    (u"a=à", [Binding(key=u"a", value=u"à", original=u"a=à")]),
+    (u'a="à"', [Binding(key=u"a", value=u"à", original=u'a="à"')]),
+    (u'garbage', [Binding(key=None, value=None, original=u"garbage")]),
+    (
+        u"a=b\nc=d",
+        [
+            Binding(key=u"a", value=u"b", original=u"a=b\n"),
+            Binding(key=u"c", value=u"d", original=u"c=d"),
+        ],
+    ),
+    (
+        u"a=b\r\nc=d",
+        [
+            Binding(key=u"a", value=u"b", original=u"a=b\r\n"),
+            Binding(key=u"c", value=u"d", original=u"c=d"),
+        ],
+    ),
+    (
+        u'a=\nb=c',
+        [
+            Binding(key=u"a", value=u'', original=u'a=\n'),
+            Binding(key=u"b", value=u'c', original=u"b=c"),
+        ]
+    ),
+    (
+        u'a=b\n\nc=d',
+        [
+            Binding(key=u"a", value=u"b", original=u"a=b\n"),
+            Binding(key=u"c", value=u"d", original=u"\nc=d"),
+        ]
+    ),
+    (
+        u'a="\nb=c',
+        [
+            Binding(key=None, value=None, original=u'a="\n'),
+            Binding(key=u"b", value=u"c", original=u"b=c"),
+        ]
+    ),
+    (
+        u'# comment\na="b\nc"\nd=e\n',
+        [
+            Binding(key=None, value=None, original=u"# comment\n"),
+            Binding(key=u"a", value=u"b\nc", original=u'a="b\nc"\n'),
+            Binding(key=u"d", value=u"e", original=u"d=e\n"),
+        ],
+    ),
+    (
+        u'garbage[%$#\na=b',
+        [
+            Binding(key=None, value=None, original=u"garbage[%$#\n"),
+            Binding(key=u"a", value=u"b", original=u'a=b'),
+        ],
+    ),
+])
+def test_parse_stream(test_input, expected):
+    result = parse_stream(StringIO(test_input))
+
+    assert list(result) == expected
diff --git a/tox.ini b/tox.ini
index 56c8732a..077780f4 100644
--- a/tox.ini
+++ b/tox.ini
@@ -19,11 +19,11 @@ deps =
   mypy
 commands =
   flake8 src tests
-  mypy --python-version=3.7 src
-  mypy --python-version=3.6 src
-  mypy --python-version=3.5 src
-  mypy --python-version=3.4 src
-  mypy --python-version=2.7 src
+  mypy --python-version=3.7 src tests
+  mypy --python-version=3.6 src tests
+  mypy --python-version=3.5 src tests
+  mypy --python-version=3.4 src tests
+  mypy --python-version=2.7 src tests
 
 [testenv:manifest]
 deps = check-manifest