diff --git a/addons/web_editor/controllers/main.py b/addons/web_editor/controllers/main.py
index 3a2e0d1d7aa..1b3f03966f3 100644
--- a/addons/web_editor/controllers/main.py
+++ b/addons/web_editor/controllers/main.py
@@ -95,7 +95,7 @@ class Web_Editor(http.Controller):
         font_obj = ImageFont.truetype(addons_path + font, size)
 
         # if received character is not a number, keep old behaviour (icon is character)
-        icon = pycompat.unichr(int(icon)) if icon.isdigit() else icon
+        icon = chr(int(icon)) if icon.isdigit() else icon
 
         # Determine the dimensions of the icon
         image = Image.new("RGBA", (size, size), color=(0, 0, 0, 0))
diff --git a/doc/_extensions/pyjsparser/parser.py b/doc/_extensions/pyjsparser/parser.py
index 28001886a8e..d47a94a8522 100644
--- a/doc/_extensions/pyjsparser/parser.py
+++ b/doc/_extensions/pyjsparser/parser.py
@@ -27,16 +27,8 @@ __all__ = ['PyJsParser', 'parse', 'ENABLE_JS2PY_ERRORS', 'ENABLE_PYIMPORT', 'JsS
 REGEXP_SPECIAL_SINGLE = ('\\', '^', '$', '*', '+', '?', '.', '[', ']', '(', ')', '{', '{', '|', '-')
 ENABLE_PYIMPORT = False
 ENABLE_JS2PY_ERRORS = False
-
-PY3 = sys.version_info >= (3,0)
-
-if PY3:
-    basestring = str
-    long = int
-    xrange = range
-    unicode = str
-
 ESPRIMA_VERSION = '2.2.0'
+
 DEBUG = False
 # Small naming convention changes
 # len -> leng
@@ -179,7 +171,7 @@ class PyJsParser:
     def scanHexEscape(self, prefix):
         code = 0
         leng = 4 if (prefix == 'u') else 2
-        for i in xrange(leng):
+        for i in range(leng):
             if self.index < self.length and isHexDigit(self.source[self.index]):
                 ch = self.source[self.index]
                 self.index += 1
@@ -442,7 +434,7 @@ class PyJsParser:
     def isImplicitOctalLiteral(self):
         # Implicit octal, unless there is a non-octal digit.
         # (Annex B.1.1 on Numeric Literals)
-        for i in xrange(self.index + 1, self.length):
+        for i in range(self.index + 1, self.length):
             ch = self.source[i];
             if (ch == '8' or ch == '9'):
                 return False;
@@ -992,20 +984,20 @@ class PyJsParser:
             except:
                 raise Exception("ENABLE_JS2PY_ERRORS was set to True, but Js2Py was not found!")
             ENABLE_PYIMPORT = old_pyimport
-            error = ERRORS['SyntaxError']('Line ' + unicode(line) + ': ' + unicode(description))
+            error = ERRORS['SyntaxError']('Line ' + str(line) + ': ' + str(description))
             error.put('index', Js(pos))
             error.put('lineNumber', Js(line))
             error.put('column', Js(pos - (self.lineStart if self.scanning else self.lastLineStart) + 1))
             error.put('description', Js(description))
             return JsToPyException(error)
         else:
-            return JsSyntaxError('Line ' + unicode(line) + ': ' + unicode(description))
+            return JsSyntaxError('Line ' + str(line) + ': ' + str(description))
 
 
     # Throw an exception
 
     def throwError(self, messageFormat, *args):
-        msg = messageFormat % tuple(unicode(e) for e in args)
+        msg = messageFormat % tuple(str(e) for e in args)
         raise self.createError(self.lastLineNumber, self.lastIndex, msg);
 
     def tolerateError(self, messageFormat, *args):
@@ -1038,7 +1030,7 @@ class PyJsParser:
             value = token['value']['raw'] if (typ == Token.Template)  else token.get('value')
         else:
             value = 'ILLEGAL'
-        msg = msg.replace('%s', unicode(value))
+        msg = msg.replace('%s', str(value))
 
         return (self.createError(token['lineNumber'], token['start'], msg) if (token and token.get('lineNumber')) else
                 self.createError(self.lineNumber if self.scanning else self.lastLineNumber,
@@ -1445,12 +1437,12 @@ class PyJsParser:
             self.reinterpretExpressionAsPattern(expr.argument)
         elif typ == Syntax.ArrayExpression:
             expr['type'] = Syntax.ArrayPattern
-            for i in xrange(len(expr['elements'])):
+            for i in range(len(expr['elements'])):
                 if (expr['elements'][i] != null):
                     self.reinterpretExpressionAsPattern(expr['elements'][i])
         elif typ == Syntax.ObjectExpression:
             expr['type'] = Syntax.ObjectPattern
-            for i in xrange(len(expr['properties'])):
+            for i in range(len(expr['properties'])):
                 self.reinterpretExpressionAsPattern(expr['properties'][i]['value']);
         elif Syntax.AssignmentExpression:
             expr['type'] = Syntax.AssignmentPattern;
@@ -1526,7 +1518,7 @@ class PyJsParser:
                     if (not self.match('=>')):
                         self.expect('=>');
                     self.isBindingElement = false
-                    for i in xrange(len(expressions)):
+                    for i in range(len(expressions)):
                         self.reinterpretExpressionAsPattern(expressions[i])
                     return {
                         'type': PlaceHolders.ArrowParameterPlaceHolder,
@@ -1539,7 +1531,7 @@ class PyJsParser:
             if (not self.isBindingElement):
                 self.throwUnexpectedToken(self.lookahead);
             if (expr['type'] == Syntax.SequenceExpression):
-                for i in xrange(len(expr.expressions)):
+                for i in range(len(expr.expressions)):
                     self.reinterpretExpressionAsPattern(expr['expressions'][i])
             else:
                 self.reinterpretExpressionAsPattern(expr);
@@ -1880,12 +1872,12 @@ class PyJsParser:
         elif typ == Syntax.AssignmentPattern:
             self.checkPatternParam(options, param.left)
         elif typ == Syntax.ArrayPattern:
-            for i in xrange(len(param.elements)):
+            for i in range(len(param.elements)):
                 if (param.elements[i] != null):
                     self.checkPatternParam(options, param.elements[i]);
         else:
             assert typ == Syntax.ObjectPattern, 'Invalid type'
-            for i in xrange(len(param.properties)):
+            for i in range(len(param.properties)):
                 self.checkPatternParam(options, param.properties[i]['value']);
 
     def reinterpretAsCoverFormalsList(self, expr):
@@ -1902,7 +1894,7 @@ class PyJsParser:
         options = {
             'paramSet': {}}
         le = len(params)
-        for i in xrange(le):
+        for i in range(le):
             param = params[i]
             if param.type == Syntax.AssignmentPattern:
                 params[i] = param.left;
@@ -2870,7 +2862,7 @@ class PyJsParser:
             raise NotImplementedError('Options not implemented! You can only use default settings.')
 
         self.clean()
-        self.source = unicode(code) + ' \n ; //END'  # I have to add it in order not to check for EOF every time
+        self.source = str(code) + ' \n ; //END'  # I have to add it in order not to check for EOF every time
         self.index = 0
         self.lineNumber = 1 if len(self.source) > 0 else 0
         self.lineStart = 0
diff --git a/doc/_extensions/pyjsparser/pyjsparserdata.py b/doc/_extensions/pyjsparser/pyjsparserdata.py
index 6ca6dbc47cc..5db295a34b5 100644
--- a/doc/_extensions/pyjsparser/pyjsparserdata.py
+++ b/doc/_extensions/pyjsparser/pyjsparserdata.py
@@ -23,12 +23,6 @@ import sys
 import unicodedata
 from collections import defaultdict
 
-PY3 = sys.version_info >= (3,0)
-
-if PY3:
-    unichr = chr
-    xrange = range
-    unicode = str
 
 token = {
         'BooleanLiteral': 1,
@@ -221,7 +215,7 @@ LS = u'\u2028'
 PS = u'\u2029'
 
 U_CATEGORIES = defaultdict(list)
-for c in map(unichr, range(sys.maxunicode + 1)):
+for c in map(chr, range(sys.maxunicode + 1)):
     U_CATEGORIES[unicodedata.category(c)].append(c)
 UNICODE_LETTER = set(U_CATEGORIES['Lu']+U_CATEGORIES['Ll']+
                      U_CATEGORIES['Lt']+U_CATEGORIES['Lm']+
@@ -242,33 +236,33 @@ WHITE_SPACE = set((0x20, 0x09, 0x0B, 0x0C, 0xA0, 0x1680,
 LINE_TERMINATORS = set((0x0A, 0x0D, 0x2028, 0x2029))
 
 def isIdentifierStart(ch):
-    return (ch if isinstance(ch, unicode) else unichr(ch))  in IDENTIFIER_START
+    return (ch if isinstance(ch, str) else chr(ch))  in IDENTIFIER_START
 
 def isIdentifierPart(ch):
-    return (ch if isinstance(ch, unicode) else unichr(ch))  in IDENTIFIER_PART
+    return (ch if isinstance(ch, str) else chr(ch))  in IDENTIFIER_PART
 
 def isWhiteSpace(ch):
-    return (ord(ch) if isinstance(ch, unicode) else ch) in WHITE_SPACE
+    return (ord(ch) if isinstance(ch, str) else ch) in WHITE_SPACE
 
 def isLineTerminator(ch):
-    return (ord(ch) if isinstance(ch, unicode) else ch)  in LINE_TERMINATORS
+    return (ord(ch) if isinstance(ch, str) else ch)  in LINE_TERMINATORS
 
 OCTAL = set(('0', '1', '2', '3', '4', '5', '6', '7'))
 DEC = set(('0', '1', '2', '3', '4', '5', '6', '7', '8', '9'))
 HEX = set('0123456789abcdefABCDEF')
-HEX_CONV = dict(('0123456789abcdef'[n],n) for n in xrange(16))
+HEX_CONV = dict(('0123456789abcdef'[n],n) for n in range(16))
 for i,e in enumerate('ABCDEF', 10):
     HEX_CONV[e] = i
 
 
 def isDecimalDigit(ch):
-    return (ch if isinstance(ch, unicode) else unichr(ch)) in DEC
+    return (ch if isinstance(ch, str) else chr(ch)) in DEC
 
 def isHexDigit(ch):
-    return (ch if isinstance(ch, unicode) else unichr(ch))  in HEX
+    return (ch if isinstance(ch, str) else chr(ch))  in HEX
 
 def isOctalDigit(ch):
-    return (ch if isinstance(ch, unicode) else unichr(ch))  in OCTAL
+    return (ch if isinstance(ch, str) else chr(ch))  in OCTAL
 
 def isFutureReservedWord(w):
     return w in ('enum', 'export', 'import', 'super')
diff --git a/doc/python3.rst b/doc/python3.rst
index ac24983ceb7..cfa3389a650 100644
--- a/doc/python3.rst
+++ b/doc/python3.rst
@@ -62,16 +62,6 @@ features whereas:
 * Python 3.2 made ``range`` views more list-like (backported to 2.7)and
   reintroduced ``callable``
 
-.. warning::
-
-    While Python 3 adds plenty of great features (keyword-only parameters,
-    generator delegation, pathlib, ...), you must *not* use them in Odoo
-    until Python 2 support is dropped
-
-.. note::
-
-    In the *very rare* cases where you *need* to differentiate between
-    Python 2 and Python 3, use the :data:`odoo.tools.pycompat.PY2` flag.
 
 Semantics changes
 =================
diff --git a/odoo/addons/base/models/qweb.py b/odoo/addons/base/models/qweb.py
index 071967093d4..d8b1a3cdde2 100644
--- a/odoo/addons/base/models/qweb.py
+++ b/odoo/addons/base/models/qweb.py
@@ -78,10 +78,7 @@ class Contextifier(ast.NodeTransformer):
     def visit_Lambda(self, node):
         args = node.args
         # assume we don't have any tuple parameter, just names
-        if pycompat.PY2:
-            names = [arg.id for arg in args.args]
-        else:
-            names = [arg.arg for arg in args.args]
+        names = [arg.arg for arg in args.args]
         if args.vararg: names.append(args.vararg)
         if args.kwarg: names.append(args.kwarg)
         # remap defaults in case there's any
diff --git a/odoo/fields.py b/odoo/fields.py
index 5a523f28fb9..8da18806f96 100644
--- a/odoo/fields.py
+++ b/odoo/fields.py
@@ -1735,9 +1735,7 @@ class Datetime(Field):
 # http://initd.org/psycopg/docs/usage.html#binary-adaptation
 # Received data is returned as buffer (in Python 2) or memoryview (in Python 3).
 _BINARY = memoryview
-if pycompat.PY2:
-    #pylint: disable=buffer-builtin,undefined-variable
-    _BINARY = buffer
+
 
 class Binary(Field):
     type = 'binary'
diff --git a/odoo/models.py b/odoo/models.py
index bbd4a67d1e1..00e44fcacdc 100644
--- a/odoo/models.py
+++ b/odoo/models.py
@@ -193,7 +193,6 @@ LOG_ACCESS_COLUMNS = ['create_uid', 'create_date', 'write_uid', 'write_date']
 MAGIC_COLUMNS = ['id'] + LOG_ACCESS_COLUMNS
 
 
-@pycompat.implements_to_string
 class BaseModel(MetaModel('DummyModel', (object,), {'_register': False})):
     """ Base class for Odoo models.
 
diff --git a/odoo/modules/migration.py b/odoo/modules/migration.py
index 4081232b5ff..2dec12674de 100644
--- a/odoo/modules/migration.py
+++ b/odoo/modules/migration.py
@@ -5,6 +5,7 @@
 
 from collections import defaultdict
 import glob
+import importlib.util
 import logging
 import os
 from os.path import join as opj
@@ -14,40 +15,17 @@ import odoo.release as release
 import odoo.tools as tools
 from odoo.tools.parse_version import parse_version
 
-if pycompat.PY2:
-    import imp
-    def load_script(path, module_name):
-        fp, fname = tools.file_open(path, pathinfo=True)
-        fp2 = None
-
-        # pylint: disable=file-builtin,undefined-variable
-        if not isinstance(fp, file):
-            # imp.load_source need a real file object, so we create
-            # one from the file-like object we get from file_open
-            fp2 = os.tmpfile()
-            fp2.write(fp.read())
-            fp2.seek(0)
-
-        try:
-            return imp.load_source(module_name, fname, fp2 or fp)
-        finally:
-            if fp:
-                fp.close()
-            if fp2:
-                fp2.close()
-
-else:
-    import importlib.util
-    def load_script(path, module_name):
-        full_path = get_resource_path(*path.split(os.path.sep))
-        spec = importlib.util.spec_from_file_location(module_name, full_path)
-        module = importlib.util.module_from_spec(spec)
-        spec.loader.exec_module(module)
-        return module
-
-
 _logger = logging.getLogger(__name__)
 
+
+def load_script(path, module_name):
+    full_path = get_resource_path(*path.split(os.path.sep))
+    spec = importlib.util.spec_from_file_location(module_name, full_path)
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
 class MigrationManager(object):
     """
         This class manage the migration of modules
diff --git a/odoo/modules/module.py b/odoo/modules/module.py
index 4f7f80ca396..c94df025f9b 100644
--- a/odoo/modules/module.py
+++ b/odoo/modules/module.py
@@ -434,12 +434,8 @@ def get_test_modules(module):
         mod = importlib.import_module('.tests', modpath)
     except ImportError as e:  # will also catch subclass ModuleNotFoundError of P3.6
         # Hide ImportErrors on `tests` sub-module, but display other exceptions
-        if pycompat.PY2:
-            if e.message.startswith('No module named') and e.message.endswith("tests"):
-                return []
-        else:
-            if e.name == modpath + '.tests' and e.msg.startswith('No module named'):
-                return []
+        if e.name == modpath + '.tests' and e.msg.startswith('No module named'):
+            return []
         _logger.exception('Can not `import %s`.', module)
         return []
     except Exception as e:
diff --git a/odoo/tests/common.py b/odoo/tests/common.py
index 89a9473e432..9a03f589200 100644
--- a/odoo/tests/common.py
+++ b/odoo/tests/common.py
@@ -354,10 +354,9 @@ class BaseCase(TreeCase, MetaCase('DummyCase', (object,), {})):
         doc = self._testMethodDoc
         return doc and ' '.join(l.strip() for l in doc.splitlines() if not l.isspace()) or None
 
-    if not pycompat.PY2:
-        # turns out this thing may not be quite as useful as we thought...
-        def assertItemsEqual(self, a, b, msg=None):
-            self.assertCountEqual(a, b, msg=None)
+    # turns out this thing may not be quite as useful as we thought...
+    def assertItemsEqual(self, a, b, msg=None):
+        self.assertCountEqual(a, b, msg=None)
 
 
 class TransactionCase(BaseCase):
diff --git a/odoo/tools/float_utils.py b/odoo/tools/float_utils.py
index c5c2946a5ce..12881503f32 100644
--- a/odoo/tools/float_utils.py
+++ b/odoo/tools/float_utils.py
@@ -2,28 +2,25 @@
 # Part of Odoo. See LICENSE file for full copyright and licensing details.
 
 from __future__ import print_function
+import builtins
 import math
 
 
-if not pycompat.PY2:
-    import builtins
-    def round(f):
-        # P3's builtin round differs from P2 in the following manner:
-        # * it rounds half to even rather than up (away from 0)
-        # * round(-0.) loses the sign (it returns -0 rather than 0)
-        # * round(x) returns an int rather than a float
-        #
-        # this compatibility shim implements Python 2's round in terms of
-        # Python 3's so that important rounding error under P3 can be
-        # trivially fixed, assuming the P2 behaviour to be debugged and
-        # correct.
-        roundf = builtins.round(f)
-        if builtins.round(f + 1) - roundf != 1:
-            return f + math.copysign(0.5, f)
-        # copysign ensures round(-0.) -> -0 *and* result is a float
-        return math.copysign(roundf, f)
-else:
-    round = round
+def round(f):
+    # P3's builtin round differs from P2 in the following manner:
+    # * it rounds half to even rather than up (away from 0)
+    # * round(-0.) loses the sign (it returns -0 rather than 0)
+    # * round(x) returns an int rather than a float
+    #
+    # this compatibility shim implements Python 2's round in terms of
+    # Python 3's so that important rounding error under P3 can be
+    # trivially fixed, assuming the P2 behaviour to be debugged and
+    # correct.
+    roundf = builtins.round(f)
+    if builtins.round(f + 1) - roundf != 1:
+        return f + math.copysign(0.5, f)
+    # copysign ensures round(-0.) -> -0 *and* result is a float
+    return math.copysign(roundf, f)
 
 def _float_check_precision(precision_digits=None, precision_rounding=None):
     assert (precision_digits is not None or precision_rounding is not None) and \
diff --git a/odoo/tools/mail.py b/odoo/tools/mail.py
index 847fcfcf2f3..8bce7e056a0 100644
--- a/odoo/tools/mail.py
+++ b/odoo/tools/mail.py
@@ -545,10 +545,7 @@ def decode_smtp_header(smtp_header):
         smtp_header = ustr(smtp_header)
     if smtp_header:
         text = decode_header(smtp_header.replace('\r', ''))
-        # The joining space will not be needed as of Python 3.3
-        # See https://github.com/python/cpython/commit/07ea53cb218812404cdbde820647ce6e4b2d0f8e
-        sep = ' ' if pycompat.PY2 else ''
-        return sep.join([ustr(x[0], x[1]) for x in text])
+        return ''.join([ustr(x[0], x[1]) for x in text])
     return u''
 
 # was mail_thread.decode_header()
diff --git a/odoo/tools/misc.py b/odoo/tools/misc.py
index b165eb3a130..e1a16b574a3 100644
--- a/odoo/tools/misc.py
+++ b/odoo/tools/misc.py
@@ -1002,7 +1002,6 @@ class Collector(Mapping):
         return len(self._map)
 
 
-@pycompat.implements_to_string
 class StackMap(MutableMapping):
     """ A stack of mappings behaving as a single mapping, and used to implement
         nested scopes. The lookups search the stack from top to bottom, and
diff --git a/odoo/tools/pycompat.py b/odoo/tools/pycompat.py
index 5f25e23e977..d13e706b934 100644
--- a/odoo/tools/pycompat.py
+++ b/odoo/tools/pycompat.py
@@ -6,101 +6,34 @@ import csv
 import codecs
 import collections
 import io
-import sys
 
 
-PY2 = sys.version_info[0] == 2
+def reraise(tp, value, tb=None):
+    if value.__traceback__ != tb:
+        raise value.with_traceback(tb)
+    raise value
 
-_Writer = collections.namedtuple('_Writer', 'writerow writerows')
-if PY2:
-    # pylint: disable=long-builtin,unichr-builtin,unicode-builtin,undefined-variable
-    unichr = unichr
-    text_type = unicode
-    string_types = (str, unicode)
-    def to_native(source, encoding='utf-8', falsy_empty=False):
-        if not source and falsy_empty:
-            return ''
 
-        if isinstance(source, text_type):
-            return source.encode(encoding)
+_reader = codecs.getreader('utf-8')
+_writer = codecs.getwriter('utf-8')
 
-        return str(source)
 
-    integer_types = (int, long)
+def csv_reader(stream, **params):
+    assert not isinstance(stream, io.TextIOBase),\
+        "For cross-compatibility purposes, csv_reader takes a bytes stream"
+    return csv.reader(_reader(stream), **params)
 
-    # noinspection PyUnresolvedReferences
-    from itertools import imap, izip, ifilter
 
-    def implements_to_string(cls):
-        cls.__unicode__ = cls.__str__
-        cls.__str__ = lambda x: x.__unicode__().encode('utf-8')
-        return cls
+def csv_writer(stream, **params):
+    assert not isinstance(stream, io.TextIOBase), \
+        "For cross-compatibility purposes, csv_writer takes a bytes stream"
+    return csv.writer(_writer(stream), **params)
 
-    def implements_iterator(cls):
-        cls.next = cls.__next__
-        del cls.__next__
-        return cls
-
-    exec ('def reraise(tp, value, tb=None):\n raise tp, value, tb')
-
-    def csv_reader(stream, **params):
-        for row in csv.reader(stream, **params):
-            yield [c.decode('utf-8') for c in row]
-    def csv_writer(stream, **params):
-        w = csv.writer(stream, **params)
-        return _Writer(
-            writerow=lambda r: w.writerow([c.encode('utf-8') for c in r]),
-            writerows=lambda rs: w.writerows(
-                [c.encode('utf-8') for c in r]
-                for r in rs
-            )
-        )
-else:
-    # pylint: disable=bad-functions
-    unichr = chr
-    text_type = str
-    string_types = (str,)
-    def to_native(source, encoding='utf-8', falsy_empty=False):
-        if not source and falsy_empty:
-            return ''
-
-        if isinstance(source, bytes):
-            return source.decode(encoding)
-
-        return str(source)
-
-    integer_types = (int,)
-
-    imap = map
-    izip = zip
-    ifilter = filter
-
-    def implements_to_string(cls):
-        return cls
-
-    def implements_iterator(cls):
-        return cls
-
-    def reraise(tp, value, tb=None):
-        if value.__traceback__ != tb:
-            raise value.with_traceback(tb)
-        raise value
-
-    _reader = codecs.getreader('utf-8')
-    _writer = codecs.getwriter('utf-8')
-    def csv_reader(stream, **params):
-        assert not isinstance(stream, io.TextIOBase),\
-            "For cross-compatibility purposes, csv_reader takes a bytes stream"
-        return csv.reader(_reader(stream), **params)
-    def csv_writer(stream, **params):
-        assert not isinstance(stream, io.TextIOBase), \
-            "For cross-compatibility purposes, csv_writer takes a bytes stream"
-        return csv.writer(_writer(stream), **params)
 
 def to_text(source):
-    """ Generates a text value (an instance of text_type) from an arbitrary 
+    """ Generates a text value (an instance of text_type) from an arbitrary
     source.
-    
+
     * False and None are converted to empty strings
     * text is passed through
     * bytes are decoded as UTF-8
@@ -112,4 +45,4 @@ def to_text(source):
     if isinstance(source, bytes):
         return source.decode('utf-8')
 
-    return text_type(source)
+    return str(source)