Merge pull request #171 – Add bytes_strictness

encukou · web-flow · commit deb411c65541 · 2018-03-02T21:07:56.000+01:00
#171
diff --git a/Doc/bytes_mode.rst b/Doc/bytes_mode.rst
@@ -43,37 +43,51 @@ Encoding/decoding to other formats – text, images, etc. – is left to the cal
 The bytes mode
 --------------
 
-The behavior of python-ldap 3.0 in Python 2 is influenced by a ``bytes_mode``
-argument to :func:`ldap.initialize`.
-The argument can take these values:
+In Python 3, text values are represented as ``str``, the Unicode text type.
 
-``bytes_mode=True``: backwards-compatible
+In Python 2, the behavior of python-ldap 3.0 is influenced by a ``bytes_mode``
+argument to :func:`ldap.initialize`:
 
-    Text values returned from python-ldap are always bytes (``str``).
-    Text values supplied to python-ldap may be either bytes or Unicode.
-    The encoding for bytes is always assumed to be UTF-8.
+``bytes_mode=True`` (backwards compatible):
+    Text values are represented as bytes (``str``) encoded using UTF-8.
 
-    Not available in Python 3.
+``bytes_mode=False`` (future compatible):
+    Text values are represented as ``unicode``.
 
-``bytes_mode=False``: strictly future-compatible
+If not given explicitly, python-ldap will default to ``bytes_mode=True``,
+but if an ``unicode`` value supplied to it, if will warn and use that value.
 
-    Text values must be represented as ``unicode``.
-    An error is raised if python-ldap receives a text value as bytes (``str``).
+Backwards-compatible behavior is not scheduled for removal until Python 2
+itself reaches end of life.
 
-Unspecified: relaxed mode with warnings
 
-    Causes a warning on Python 2.
+Errors, warnings, and automatic encoding
+----------------------------------------
 
-    Text values returned from python-ldap are always ``unicode``.
-    Text values supplied to python-ldap should be ``unicode``;
-    warnings are emitted when they are not.
+While the type of values *returned* from python-ldap is always given by
+``bytes_mode``, for Python 2 the behavior for “wrong-type” values *passed in*
+can be controlled by the ``bytes_strictness`` argument to
+:func:`ldap.initialize`:
 
-    The warnings are of type :class:`~ldap.LDAPBytesWarning`, which
-    is a subclass of :class:`BytesWarning` designed to be easily
-    :ref:`filtered out <filter-bytes-warning>` if needed.
+``bytes_strictness='error'`` (default if ``bytes_mode`` is specified):
+  A ``TypeError`` is raised.
 
-Backwards-compatible behavior is not scheduled for removal until Python 2
-itself reaches end of life.
+``bytes_strictness='warn'`` (default when ``bytes_mode`` is not given explicitly):
+  A warning is raised, and the value is encoded/decoded
+  using the UTF-8 encoding.
+
+  The warnings are of type :class:`~ldap.LDAPBytesWarning`, which
+  is a subclass of :class:`BytesWarning` designed to be easily
+  :ref:`filtered out <filter-bytes-warning>` if needed.
+
+``bytes_strictness='silent'``:
+  The value is automatically encoded/decoded using the UTF-8 encoding.
+
+On Python 3, ``bytes_strictness`` is ignored and a ``TypeError`` is always
+raised.
+
+When setting ``bytes_strictness``, an explicit value for ``bytes_mode`` needs
+to be given as well.
 
 
 Porting recommendations
diff --git a/Doc/reference/ldap.rst b/Doc/reference/ldap.rst
@@ -29,7 +29,7 @@ Functions
 
 This module defines the following functions:
 
-.. py:function:: initialize(uri [, trace_level=0 [, trace_file=sys.stdout [, trace_stack_limit=None, [bytes_mode=None]]]]) -> LDAPObject object
+.. py:function:: initialize(uri [, trace_level=0 [, trace_file=sys.stdout [, trace_stack_limit=None, [bytes_mode=None, [bytes_strictness=None]]]]]) -> LDAPObject object
 
    Initializes a new connection object for accessing the given LDAP server,
    and return an LDAP object (see :ref:`ldap-objects`) used to perform operations
@@ -53,7 +53,8 @@ This module defines the following functions:
    *trace_file* specifies a file-like object as target of the debug log and
    *trace_stack_limit* specifies the stack limit of tracebacks in debug log.
 
-   The *bytes_mode* argument specifies text/bytes behavior under Python 2.
+   The *bytes_mode* and *bytes_strictness* arguments specify text/bytes
+   behavior under Python 2.
    See :ref:`text-bytes` for a complete documentation.
 
    Possible values for *trace_level* are
@@ -696,6 +697,9 @@ and wait for and return with the server's result, or with
 
    *serverctrls* and *clientctrls* like described in section :ref:`ldap-controls`.
 
+   The *dn* argument, and mod_type (second item) of *modlist* are text strings;
+   see :ref:`bytes_mode`.
+
 
 .. py:method:: LDAPObject.bind(who, cred, method) -> int
 
@@ -737,6 +741,8 @@ and wait for and return with the server's result, or with
 
    *serverctrls* and *clientctrls* like described in section :ref:`ldap-controls`.
 
+   The *dn* and *attr* arguments are text strings; see :ref:`bytes_mode`.
+
    .. note::
 
       A design fault in the LDAP API prevents *value*
@@ -757,6 +763,8 @@ and wait for and return with the server's result, or with
 
    *serverctrls* and *clientctrls* like described in section :ref:`ldap-controls`.
 
+   The *dn* argument is text string; see :ref:`bytes_mode`.
+
 
 .. py:method:: LDAPObject.extop(extreq[,serverctrls=None[,clientctrls=None]]]) -> int
 
@@ -810,6 +818,9 @@ and wait for and return with the server's result, or with
    You might want to look into sub-module :py:mod:`ldap.modlist` for
    generating *modlist*.
 
+   The *dn* argument, and mod_type (second item) of *modlist* are text strings;
+   see :ref:`bytes_mode`.
+
 
 .. py:method:: LDAPObject.modrdn(dn, newrdn [, delold=1]) -> int
 
@@ -826,6 +837,8 @@ and wait for and return with the server's result, or with
    This operation is emulated by :py:meth:`rename()` and :py:meth:`rename_s()` methods
    since the modrdn2* routines in the C library are deprecated.
 
+   The *dn* and *newrdn* arguments are text strings; see :ref:`bytes_mode`.
+
 
 .. py:method:: LDAPObject.passwd(user, oldpw, newpw [, serverctrls=None [, clientctrls=None]]) -> int
 
@@ -844,6 +857,8 @@ and wait for and return with the server's result, or with
 
    The asynchronous version returns the initiated message id.
 
+   The *user*, *oldpw* and *newpw* arguments are text strings; see :ref:`bytes_mode`.
+
    .. seealso::
 
       :rfc:`3062` - LDAP Password Modify Extended Operation
@@ -865,6 +880,8 @@ and wait for and return with the server's result, or with
 
    *serverctrls* and *clientctrls* like described in section :ref:`ldap-controls`.
 
+   The *dn* and *newdn* arguments are text strings; see :ref:`bytes_mode`.
+
 
 .. py:method:: LDAPObject.result([msgid=RES_ANY [, all=1 [, timeout=None]]]) -> 2-tuple
 
@@ -1015,12 +1032,13 @@ and wait for and return with the server's result, or with
 
    *serverctrls* and *clientctrls* like described in section :ref:`ldap-controls`.
 
+   The *who* and *cred* arguments are text strings; see :ref:`bytes_mode`.
+
    .. versionchanged:: 3.0
 
       :meth:`~LDAPObject.simple_bind` and :meth:`~LDAPObject.simple_bind_s`
       now accept ``None`` for *who* and *cred*, too.
 
-
 .. py:method:: LDAPObject.search(base, scope [,filterstr='(objectClass=*)' [, attrlist=None [, attrsonly=0]]]) ->int
 
 .. py:method:: LDAPObject.search_s(base, scope [,filterstr='(objectClass=*)' [, attrlist=None [, attrsonly=0]]]) ->list|None
@@ -1073,6 +1091,9 @@ and wait for and return with the server's result, or with
    or :py:meth:`search_ext_s()` (client-side search limit). If non-zero
    not more than *sizelimit* results are returned by the server.
 
+   The *base* and *filterstr* arguments, and *attrlist* contents,
+   are text strings; see :ref:`bytes_mode`.
+
    .. versionchanged:: 3.0
 
       ``filterstr=None`` is equivalent to ``filterstr='(objectClass=*)'``.
diff --git a/Lib/ldap/ldapobject.py b/Lib/ldap/ldapobject.py
@@ -93,7 +93,8 @@ class SimpleLDAPObject:
 
   def __init__(
     self,uri,
-    trace_level=0,trace_file=None,trace_stack_limit=5,bytes_mode=None
+    trace_level=0,trace_file=None,trace_stack_limit=5,bytes_mode=None,
+    bytes_strictness=None,
   ):
     self._trace_level = trace_level
     self._trace_file = trace_file or sys.stdout
@@ -107,20 +108,26 @@ def __init__(
     # Bytes mode
     # ----------
 
-    # By default, raise a TypeError when receiving invalid args
-    self.bytes_mode_hardfail = True
-    if bytes_mode is None and PY2:
-      _raise_byteswarning(
-        "Under Python 2, python-ldap uses bytes by default. "
-        "This will be removed in Python 3 (no bytes for DN/RDN/field names). "
-        "Please call initialize(..., bytes_mode=False) explicitly.")
-      bytes_mode = True
-      # Disable hard failure when running in backwards compatibility mode.
-      self.bytes_mode_hardfail = False
-    elif bytes_mode and not PY2:
-      raise ValueError("bytes_mode is *not* supported under Python 3.")
-    # On by default on Py2, off on Py3.
+    if PY2:
+        if bytes_mode is None:
+            bytes_mode = True
+            if bytes_strictness is None:
+                _raise_byteswarning(
+                  "Under Python 2, python-ldap uses bytes by default. "
+                  "This will be removed in Python 3 (no bytes for "
+                  "DN/RDN/field names). "
+                  "Please call initialize(..., bytes_mode=False) explicitly.")
+                bytes_strictness = 'warn'
+        else:
+            if bytes_strictness is None:
+                bytes_strictness = 'error'
+    else:
+        if bytes_mode:
+            raise ValueError("bytes_mode is *not* supported under Python 3.")
+        bytes_mode = False
+        bytes_strictness = 'error'
     self.bytes_mode = bytes_mode
+    self.bytes_strictness = bytes_strictness
 
   def _bytesify_input(self, arg_name, value):
     """Adapt a value following bytes_mode in Python 2.
@@ -130,38 +137,46 @@ def _bytesify_input(self, arg_name, value):
     With bytes_mode ON, takes bytes or None and returns bytes or None.
     With bytes_mode OFF, takes unicode or None and returns bytes or None.
 
-    This function should be applied on all text inputs (distinguished names
-    and attribute names in modlists) to convert them to the bytes expected
-    by the C bindings.
+    For the wrong argument type (unicode or bytes, respectively),
+    behavior depends on the bytes_strictness setting.
+    In all cases, bytes or None are returned (or an exception is raised).
     """
     if not PY2:
       return value
-
     if value is None:
       return value
+
     elif self.bytes_mode:
       if isinstance(value, bytes):
         return value
+      elif self.bytes_strictness == 'silent':
+        pass
+      elif self.bytes_strictness == 'warn':
+        _raise_byteswarning(
+            "Received non-bytes value for '{}' in bytes mode; "
+            "please choose an explicit "
+            "option for bytes_mode on your LDAP connection".format(arg_name))
       else:
-        if self.bytes_mode_hardfail:
           raise TypeError(
             "All provided fields *must* be bytes when bytes mode is on; "
             "got type '{}' for '{}'.".format(type(value).__name__, arg_name)
           )
-        else:
-          _raise_byteswarning(
-            "Received non-bytes value for '{}' with default (disabled) bytes mode; "
-            "please choose an explicit "
-            "option for bytes_mode on your LDAP connection".format(arg_name))
-          return value.encode('utf-8')
+      return value.encode('utf-8')
     else:
-      if not isinstance(value, text_type):
+      if isinstance(value, unicode):
+        return value.encode('utf-8')
+      elif self.bytes_strictness == 'silent':
+        pass
+      elif self.bytes_strictness == 'warn':
+        _raise_byteswarning(
+            "Received non-text value for '{}' with bytes_mode off and "
+            "bytes_strictness='warn'".format(arg_name))
+      else:
         raise TypeError(
           "All provided fields *must* be text when bytes mode is off; "
           "got type '{}' for '{}'.".format(type(value).__name__, arg_name)
         )
-      assert not isinstance(value, bytes)
-      return value.encode('utf-8')
+      return value
 
   def _bytesify_modlist(self, arg_name, modlist, with_opcode):
     """Adapt a modlist according to bytes_mode.
@@ -1064,7 +1079,7 @@ class ReconnectLDAPObject(SimpleLDAPObject):
   def __init__(
     self,uri,
     trace_level=0,trace_file=None,trace_stack_limit=5,bytes_mode=None,
-    retry_max=1,retry_delay=60.0
+    bytes_strictness=None, retry_max=1, retry_delay=60.0
   ):
     """
     Parameters like SimpleLDAPObject.__init__() with these
@@ -1078,7 +1093,9 @@ def __init__(
     self._uri = uri
     self._options = []
     self._last_bind = None
-    SimpleLDAPObject.__init__(self,uri,trace_level,trace_file,trace_stack_limit,bytes_mode)
+    SimpleLDAPObject.__init__(self, uri, trace_level, trace_file,
+                              trace_stack_limit, bytes_mode,
+                              bytes_strictness=bytes_strictness)
     self._reconnect_lock = ldap.LDAPLock(desc='reconnect lock within %s' % (repr(self)))
     self._retry_max = retry_max
     self._retry_delay = retry_delay
@@ -1097,6 +1114,11 @@ def __getstate__(self):
 
   def __setstate__(self,d):
     """set up the object from pickled data"""
+    hardfail = d.get('bytes_mode_hardfail')
+    if hardfail:
+        d.setdefault('bytes_strictness', 'error')
+    else:
+        d.setdefault('bytes_strictness', 'warn')
     self.__dict__.update(d)
     self._last_bind = getattr(SimpleLDAPObject, self._last_bind[0]), self._last_bind[1], self._last_bind[2]
     self._ldap_object_lock = self._ldap_lock()
diff --git a/Tests/t_ldapobject.py b/Tests/t_ldapobject.py
