* Wed Jan 24 2018 email@example.com
- Version update to 1.6.1:
* Various fixes for python3
* Various pylint issues fixes
* For detailed log see ChangeLog
- Disable tests for now as they seem to fail, upstream knows
* Mon Feb 27 2017 firstname.lastname@example.org
- update for singlespec
- update to 1.4.9
* classes can now know their definition-time arguments
* getitem() now accepts node objects
* explicit namespace package support
* see more details in ChangeLog
- update to 2.0-pre-release git version for python 3.6 support
- enable test suite for all pythons
* Wed Feb 10 2016 email@example.com
- Added missing dependencies on python-lazy_object_proxy
- Update to version 1.4.4
* unpack_infer raises InferenceError if it can't operate
with the given sequences of nodes.
* Support accessing properties with super().
* Enforce strong updates per frames.
* Wed Jan 13 2016 firstname.lastname@example.org
- Update to version 1.4.3
* pkg_resources brain tips are a bit more specific,
by specifiying proper returns.
* Standard library modules are properly detected by is_standard_module.
This should fix issues such as https://github.com/PyCQA/pylint/issues/725.
- Update to version 1.4.2
* The slots() method conflates all the slots from the ancestors
into a list of current and parent slots.
We're doing this because this is the right semantics of slots,
they get inherited, as long as each parent defines a __slots__
* Revert to using printf-style formatting in as_string, in order
to avoid a potential problem with encodings when using .format.
Closes issue #273.
* assigned_stmts methods have the same signature from now on.
They used to have different signatures and each one made
assumptions about what could be passed to other implementations,
leading to various possible crashes when one or more arguments
weren't given. Closes issue #277.
- update to version 1.4.1:
* Add support for handling Uninferable nodes when calling as_string
Some object, for instance List or Tuple can have, after inference,
Uninferable as their elements, happening when their components
weren't couldn't be inferred properly. This means that as_string
needs to cope with expecting Uninferable nodes part of the other
nodes coming for a string transformation. The patch adds a visit
method in AsString and "accept" on Yes / Uninferable nodes.
Closes issue #270.
- update to version 1.4.0:
* Class.getattr('__mro__') returns the actual MRO. Closes issue #128.
* The logilab-common dependency is not needed anymore as the needed code
was integrated into astroid.
* Add 'assert_equals' method in nose.tools's brain plugin.
* Generated enum member stubs now support IntEnum and multiple
* Add brain tips for multiprocessing.Manager and
* Add brain tips for multiprocessing post Python 3.4+,
where the module level functions are retrieved with getattr
from a context object, leading to many no-member errors
* The Generator objects inferred with `infer_call_result`
from functions have as parent the function from which they
* Understand partially the 3-argument form of `type`.
The only change is that astroid understands members
passed in as dictionaries as the third argument.
* Improve the inference of Getattr nodes when dealing with
abstract properties from the abc module.
In astroid.bases.Instance._wrap_attr we had a detection
code for properties, which basically inferred whatever
a property returned, passing the results up the stack,
to the igetattr() method. It handled only the builtin property
but the new patch also handles a couple of other properties,
such as abc.abstractproperty.
* UnboundMethod.getattr calls the getattr of its _proxied object
and doesn't call super(...) anymore.
It previously crashed, since the first ancestor in its mro was
bases.Proxy and bases.Proxy doesn't implement the .getattr method.
Closes issue #91.
* Don't hard fail when calling .mro() on a class which has
combined both newstyle and old style classes. The class
in question is actually newstyle (and the __mro__ can be
retrieved using Python).
.mro() fallbacks to using .ancestors() in that case.
* Class.local_attr and Class.local_attr_ancestors uses internally
a mro lookup, using .mro() method, if they can.
That means for newstyle classes, when trying to lookup a member
using one of these functions, the first one according to the
mro will be returned. This reflects nicely the reality,
but it can have as a drawback the fact that it is a behaviour
change (the previous behaviour was incorrect though). Also,
having bases which can return multiple values when inferred
will not work with the new approach, because .mro() only
retrieves the first value inferred from a base.
* Expose a implicit_metaclass() method in Class. This will return
a builtins.type instance for newstyle classes.
* Add two new exceptions for handling MRO error cases. DuplicateBasesError
is emitted when duplicate bases are found in a class,
InconsistentMroError is raised when the method resolution is determined
to be inconsistent. They share a common class, MroError, which
is a subclass of ResolveError, meaning that this change is backwards
* Classes aren't marked as interfaces anymore, in the `type` attribute.
* Class.has_dynamic_getattr doesn't return True for special methods
which aren't implemented in pure Python, as it is the case for extension modules.
Since most likely the methods were coming from a live object, this implies
that all of them will have __getattr__ and __getattribute__ present and it
is wrong to consider that those methods were actually implemented.
* Add `igetattr` method to scoped_nodes.Function.
* Add support for Python 3.5's MatMul operation: see PEP 465 for more
* NotImplemented is detected properly now as being part of the
builtins module. Previously trying to infer the Name(NotImplemented)
returned an YES object.
* Add proper grammatical names for `infered` and `ass_type` methods,
namely `inferred` and `assign_type`.
The old methods will raise PendingDeprecationWarning, being slated
for removal in astroid 2.0.
* Add new AST names in order to be similar to the ones
from the builtin ast module.
With this change, Getattr becomes Attributes, Backquote becomes
Repr, Class is ClassDef, Function is FunctionDef, Discard is Expr,
CallFunc is Call, From is ImportFrom, AssName is AssignName
and AssAttr is AssignAttr. The old names are maintained for backwards
compatibility and they are interchangeable, in the sense that using
Discard will use Expr under the hood and the implemented visit_discard
in checkers will be called with Expr nodes instead. The AST does not
contain the old nodes, only the interoperability between them hides this
fact. Recommandations to move to the new nodes are emitted accordingly,
the old names will be removed in astroid 2.0.
* Star unpacking in assignments returns properly a list,
not the individual components. Closes issue #138.
* Lambdas found at class level, which have a `self` argument, are considered
* Add annotation support for function.as_string(). Closes issue #37.
* Add support for indexing bytes on Python 3.
BoundMethods when accessing them from instances of their class.
* Add support for inferring subscript on instances, which will
use __getitem__. Closes issue #124.
* Understand metaclasses added with six.add_metaclass decorator. Closes issue #129.
* Add a new convenience API, `astroid.parse`, which can be used to retrieve
an astroid AST from a source code string, similar to how ast.parse can be
used to obtain a Python AST from a source string. This is the test_utils.build_module
promoted to a public API.
* do_import_module passes the proper relative_only flag if the level is higher
than 1. This has the side effect that using `from .something import something`
in a non-package will finally result in an import-error on Pylint's side.
Until now relative_only was ignored, leading to the import of `something`,
if it was globally available.
* There's a new separate step for transforms.
Until now, the transforms were applied at the same time the tree was
being built. This was problematic if the transform functions were
using inference, since the inference was executed on a partially
constructed tree, which led to failures when post-building
information was needed (such as setting the _from_names
for the From imports).
Now there's a separate step for transforms, which are applied
There's a couple of other related changes:
+ astroid.parse and AstroidBuilder gained a new parameter
`apply_transforms`, which is a boolean flag, which will
control if the transforms are applied. We do this because
there are uses when the vanilla tree is wanted, without
any implicit modification.
+ the transforms are also applied for builtin modules,
as a side effect of the fact that transform visiting
was moved in AstroidBuilder._post_build from
Closes issue #116.
* Add a new node, DictUnpack, which is used to represent the unpacking
of a dictionary into another dictionary, using PEP 448 specific syntax
This is a different approach than what the builtin ast module does,
since it just uses None to represent this kind of operation,
which seems conceptually wrong, due to the fact the AST contains
non-AST nodes. Closes issue #206.
* Add a new type of node, called *inference objects*. Inference objects are similar with
AST nodes, but they can be obtained only after inference, so they can't be found
inside the AST tree. Their purpose is to handle at astroid level
some operations which can't be handled when using brain transforms.
For instance, the first object added is FrozenSet, which can be manipulated
at astroid's level (inferred, itered etc). Code such as this 'frozenset((1,2))'
will not return an Instance of frozenset, without having access to its
content, but a new objects.FrozenSet, which can be used just as a nodes.Set.
* Add a new *inference object* called Super, which also adds support for understanding
super calls. astroid understands the zero-argument form of super, specific to
Python 3, where the interpreter fills itself the arguments of the call. Also, we
are understanding the 2-argument form of super, both for bounded lookups
(super(X, instance)) as well as for unbounded lookups (super(X, Y)),
having as well support for validating that the object-or-type is a subtype
of the first argument. The unbounded form of super (one argument) is not
understood, since it's useless in practice and should be removed from
Python's specification. Closes issue #89.
* astroid.utils.ASTWalker and astroid.utils.LocalsVisitor
were moved to pylint.pyreverse.utils.
- update to version 1.3.8:
* Backport of 40e3176, which fixes issue #84.
- update to version 1.3.7:
* Improve the inference of six.moves, especially when using `from
... import ...` syntax. Also, we added a new fail import hook for
six.moves, which fixes the import-error false positive from
pylint. Closes issue #107.
* Mon Mar 16 2015 email@example.com
- Update to 1.3.6:
* Class.slots raises NotImplementedError for old style classes.
Closes issue #67.
* Add a new option to AstroidManager, `optimize_ast`, which
controls if peephole optimizer should be enabled or not.
This prevents a regression, where the visit_binop method
wasn't called anymore with astroid 1.3.5, due to the differences
in the resulting AST. Closes issue #82.
* Add the ability to optimize small ast subtrees,
with the first use in the optimization of multiple
BinOp nodes. This removes recursivity in the rebuilder
when dealing with a lot of small strings joined by the
addition operator. Closes issue #59.
* Obtain the methods for the nose brain tip through an
unittest.TestCase instance. Closes Pylint issue #457.
* Fix a crash which occurred when a class was the ancestor
of itself. Closes issue #78.
* Improve the scope_lookup method for Classes regarding qualified
objects, with an attribute name exactly as one provided in the
For example, a class containing an attribute 'first',
which was also an import and which had, as a base, a qualified name
or a Gettattr node, in the form 'module.first', then Pylint would
have inferred the `first` name as the function from the Class,
not the import. Closes Pylint issue #466.
* Implement the assigned_stmts operation for Starred nodes,
which was omitted when support for Python 3 was added in astroid.
Closes issue #36.
* Wed Feb 18 2015 firstname.lastname@example.org
- Update to 1.3.4:
* Get the first element from the method list when obtaining
the functions from nose.tools.trivial. Closes Pylint issue #448.
* Restore file_stream to a property, but deprecate it in favour of
the newly added method Module.stream. By using a method instead of a
property, it will be easier to properly close the file right
after it is used, which will ensure that no file descriptors are
leaked. Until now, due to the fact that a module was cached,
it was not possible to close the file_stream anywhere.
file_stream will start emitting PendingDeprecationWarnings in
astroid 1.4, DeprecationWarnings in astroid 1.5 and it will
be finally removed in astroid 1.6.
* Add inference tips for 'tuple', 'list', 'dict' and 'set' builtins.
* Add brain definition for most string and unicode methods
* Changed the API for Class.slots. It returns None when the class
doesn't define any slots. Previously, for both the cases where
the class didn't have slots defined and when it had an empty list
of slots, Class.slots returned an empty list.
* Add a new method to Class nodes, 'mro', for obtaining the
the method resolution order of the class.
* Add brain tips for six.moves. Closes issue #63.
* Improve the detection for functions decorated with decorators
which returns static or class methods.
* .slots() can contain unicode strings on Python 2.
* Add inference tips for nose.tools.
* Mon Dec 08 2014 email@example.com
- Update to 1.3.2:
* Fixed a crash with invalid subscript index.
* Implement proper base class semantics for Python 3, where
every class derives from object.
* Allow more fine-grained control over C extension loading
in the manager.
* Fixed a crash issue with the pytest brain module.
* Fri Nov 21 2014 firstname.lastname@example.org
- Update to 1.3.0:
* Fix a maximum recursion error occured during the inference,
where statements with the same name weren't filtered properly.
Closes pylint issue #295.
* Check that EmptyNode has an underlying object in
* Simplify the understanding of enum members.
* Fix an infinite loop with decorator call chain inference,
where the decorator returns itself. Closes issue #50.
* Various speed improvements. Patch by Alex Munroe.
* Add pytest brain plugin. Patch by Robbie Coomber.
* Support for Python versions < 2.7 has been dropped, and the
source has been made compatible with Python 2 and 3. Running
2to3 on installation for Python 3 is not needed anymore.
* astroid now depends on six.
* modutils._module_file opens __init__.py in binary mode.
Closes issues #51 and #13.
* Only C extensions from trusted sources (the standard library)
are loaded into the examining Python process to build an AST
from the live module.
* Path names on case-insensitive filesystems are now properly
handled. This fixes the stdlib detection code on Windows.
* Metaclass-generating functions like six.with_metaclass
are now supported via some explicit detection code.
* astroid.register_module_extender has been added to generalize
the support for module extenders as used by many brain plugins.
* brain plugins can now register hooks to handle failed imports,
as done by the gobject-introspection plugin.
* The modules have been moved to a separate package directory,
`setup.py develop` now works correctly.
* Tue Nov 04 2014 email@example.com
- Update to 1.2.1:
* Fix a crash occurred when inferring decorator call chain.
Closes issue #42.
* Set the parent of vararg and kwarg nodes when inferring them.
Closes issue #43.
* namedtuple inference knows about '_fields' attribute.
* enum members knows about the methods from the enum class.
* Name inference will lookup in the parent function
of the current scope, in case searching in the current scope
* Inference of the functional form of the enums takes into
consideration the various inputs that enums accepts.
* The inference engine handles binary operations (add, mul etc.)
* Fix an infinite loop in the inference, by returning a copy
of instance attributes, when calling 'instance_attr'.
Closes issue #34 (patch by Emile Anclin).
* Don't crash when trying to infer unbound object.__new__ call.
Closes issue #11.
- Use upstream zip for packaging as .tar.gz is no longer provided
* Tue Aug 12 2014 firstname.lastname@example.org
- Update to 1.2.0
* Function nodes can detect decorator call chain and see if they are
decorated with builtin descriptors (`classmethod` and `staticmethod`).
* infer_call_result called on a subtype of the builtin type will now
return a new `Class` rather than an `Instance`.
* `Class.metaclass()` now handles module-level __metaclass__ declaration
on python 2, and no longer looks at the __metaclass__ class attribute on
* Function nodes can detect if they are decorated with subclasses
of builtin descriptors when determining their type
(`classmethod` and `staticmethod`).
* Add `slots` method to `Class` nodes, for retrieving
the list of valid slots it defines.
* Expose function annotation to astroid: `Arguments` node
exposes 'varargannotation', 'kwargannotation' and 'annotations'
attributes, while `Function` node has the 'returns' attribute.
* Backported most of the logilab.common.modutils module there, as
most things there are for pylint/astroid only and we want to be
able to fix them without requiring a new logilab.common release
* Fix names grabed using wildcard import in "absolute import mode"
(ie with absolute_import activated from the __future__ or with
python 3). Fix pylint issue #58.
* Add support in pylint-brain for understanding enum classes.
* Thu Jul 10 2014 email@example.com
- Update to 1.1.1
* `Class.metaclass()` looks in ancestors when the current class
does not define explicitly a metaclass.
* Do not cache modules if a module with the same qname is already
known, and only return cached modules if both name and filepath
match. Fixes pylint Bitbucket issue #136.
- Update to 1.1.0
* All class nodes are marked as new style classes for Py3k.
* Add a `metaclass` function to `Class` nodes to
retrieve their metaclass.
* Add a new YieldFrom node.
* Add support for inferring arguments to namedtuple invocations.
* Make sure that objects returned for namedtuple
inference have parents.
* Don't crash when inferring nodes from `with` clauses
with multiple context managers. Closes #18.
* Don't crash when a class has some __call__ method that is not
inferable. Closes #17.
* Unwrap instances found in `.ancestors()`, by using their _proxied
* Thu Nov 28 2013 firstname.lastname@example.org
- Rename to python-astroid to match new upstream name.
- Initial version 1.0.1
* Wed Jun 19 2013 email@example.com
- Use upstream URL
* Fri May 31 2013 firstname.lastname@example.org
- update to 0.24.3:
* #124360 [py3.3]: Don't crash on 'yield from' nodes
* #123062 [pylint-brain]: Use correct names for keywords for urlparse
* #123056 [pylint-brain]: Add missing methods for hashlib
* #123068: Fix inference for generator methods to correctly handle yields
* #123068: Make sure .as_string() returns valid code for yields in
* #47957: Set literals are now correctly treated as inference leaves.
* #123074: Add support for inference of subscript operations on dict
* pylint-brain: more subprocess.Popen faking (see #46273)
* #109562 [jython]: java modules have no __doc__, causing crash
* #120646 [py3]: fix for python3.3 _ast changes which may cause crash
* #109988 [py3]: test fixes
* #106191: fix __future__ absolute import w/ From node
* #50395: fix function fromlineno when some decorator is splited on
multiple lines (patch by Mark Gius)
* #92362: fix pyreverse crash on relative import
* #104041: fix crash 'module object has no file_encoding attribute'
* #4294 (pylint-brain): bad inference on mechanize.Browser.open
* #46273 (pylint-brain): bad inference subprocess.Popen.communicate
* Thu Apr 26 2012 email@example.com
- Add python 3 package
- Change license to LGPL 2.1+
* Sat Feb 18 2012 firstname.lastname@example.org
- update to version 0.23.1
* #62295: avoid "OSError: Too many open files" by moving
.file_stream as a Module property opening the file only when
* Lambda nodes should have a `name` attribute
* only call transformers if modname specified
* Thu Oct 06 2011 email@example.com
- make it build on old distros
* Sun Sep 25 2011 firstname.lastname@example.org
- Don't remove logilab/__init__.py on SLE, where it is not existing.
* Wed Sep 21 2011 email@example.com
- Update to version 0.22.0:
* added column offset information on nodes (patch by fawce)
* #70497: Crash on AttributeError: 'NoneType' object has no attribute '_infer_name'
* #70381: IndendationError in import causes crash
* #70565: absolute imports treated as relative (patch by Jacek Konieczny)
* #70494: fix file encoding detection with python2.x
* py3k: __builtin__ module renamed to builtins, we should consider this to properly
build ast for builtin objects
- Run testsuite
- Removed outdated stuff from spec file (%clean section, authors, ...)
- Set license to GPL-2.0+ (SPDX style)
* Tue Apr 12 2011 firstname.lastname@example.org
- update to 0.21.1:
* python3: handle file encoding; fix a lot of tests
* fix #52006: "True" and "False" can be assigned as variable in Python2x
* fix #8847: pylint doesn't understand function attributes at all
* fix #8774: iterator / generator / next method