collections.abc.Awaitable and collections.abc.Coroutine no longer
use __instancecheck__ hook to detect generator-based coroutines.
inspect.isawaitable() can be used to detect generator-based coroutines
and to distinguish them from regular generator objects.
isawaitable() was added before collections.abc.Awaitable; now,
with Awaitable, it is no longer needed (we don't have ishashable()
or isiterable() methods in the inspect module either).
Summary of changes:
1. Coroutines now have a distinct, separate from generators
type at the C level: PyGen_Type, and a new typedef PyCoroObject.
PyCoroObject shares the initial segment of struct layout with
PyGenObject, making it possible to reuse existing generators
machinery. The new type is exposed as 'types.CoroutineType'.
As a consequence of having a new type, CO_GENERATOR flag is
no longer applied to coroutines.
2. Having a separate type for coroutines made it possible to add
an __await__ method to the type. Although it is not used by the
interpreter (see details on that below), it makes coroutines
naturally (without using __instancecheck__) conform to
collections.abc.Coroutine and collections.abc.Awaitable ABCs.
[The __instancecheck__ is still used for generator-based
coroutines, as we don't want to add __await__ for generators.]
3. Add new opcode: GET_YIELD_FROM_ITER. The opcode is needed to
allow passing native coroutines to the YIELD_FROM opcode.
Before this change, 'yield from o' expression was compiled to:
(o)
GET_ITER
LOAD_CONST
YIELD_FROM
Now, we use GET_YIELD_FROM_ITER instead of GET_ITER.
The reason for adding a new opcode is that GET_ITER is used
in some contexts (such as 'for .. in' loops) where passing
a coroutine object is invalid.
4. Add two new introspection functions to the inspec module:
getcoroutinestate(c) and getcoroutinelocals(c).
5. inspect.iscoroutine(o) is updated to test if 'o' is a native
coroutine object. Before this commit it used abc.Coroutine,
and it was requested to update inspect.isgenerator(o) to use
abc.Generator; it was decided, however, that inspect functions
should really be tailored for checking for native types.
6. sys.set_coroutine_wrapper(w) API is updated to work with only
native coroutines. Since types.coroutine decorator supports
any type of callables now, it would be confusing that it does
not work for all types of coroutines.
7. Exceptions logic in generators C implementation was updated
to raise clearer messages for coroutines:
Before: TypeError("generator raised StopIteration")
After: TypeError("coroutine raised StopIteration")
For details, see:
PEP 0485 -- A Function for testing approximate equality
Functions added: math.isclose() and cmath.isclose().
Original code by Chris Barker. Patch by Tal Einat.
(merge 3.5)
For details, see:
PEP 0485 -- A Function for testing approximate equality
Functions added: math.isclose() and cmath.isclose().
Original code by Chris Barker. Patch by Tal Einat.
Known limitations of the current implementation:
- documentation changes are incomplete
- there's a reference leak I haven't tracked down yet
The leak is most visible by running:
./python -m test -R3:3 test_importlib
However, you can also see it by running:
./python -X showrefcount
Importing the array or _testmultiphase modules, and
then deleting them from both sys.modules and the local
namespace shows significant increases in the total
number of active references each cycle. By contrast,
with _testcapi (which continues to use single-phase
initialisation) the global refcounts stabilise after
a couple of cycles.
Also, deprecate formatargspec, formatargvalues, and getargvalues
functions. Since we are deprecating 'getfullargspec' function in
3.5 (documentation only, no DeprecationWarning), it makes sense
to also deprecate functions designed to be directly used with it.
In 3.6 we will remove 'getargsspec' function (was deprecated since
Python 3.0), and start raising DeprecationWarnings in other
'getarg*' family of functions. We can remove them in 3.7 or later.
Also, it is worth noting, that Signature API does not provide 100%
of functionality that deprecated APIs have. It is important to do
a soft deprecation of outdated APIs in 3.5 to gather users feedback,
and improve Signature object.
This defaults to True in the compat32 policy for backward compatibility,
but to False for all new policies.
Patch by Milan Oberkirch, with a few tweaks.
This could use more edge case tests, but the basic functionality is tested.
(Note that this changeset does not add tailored support for the RFC 6532
message/global MIME type, but the email package generic facilities will handle
it.)
Reviewed by Maciej Szulik.
Patch by Milan Oberkirch, with a few updates. This changeset also
tweaks the smtpd and whatsnew docs for smtpd into what should be
the final form for the 3.5 release.
Some applications (e.g. traditional Unix diff, version control
systems) neither know nor care about the encodings of the files they
are comparing. They are textual, but to the diff utility they are just
bytes. This worked fine under Python 2, because all of the hardcoded
strings in difflib.py are ASCII, so could safely be combined with
old-style u'' strings. But it stopped working in 3.x.
The solution is to use surrogate escapes for a lossless
bytes->str->bytes roundtrip. That means {unified,context}_diff() can
continue to just handle strings without worrying about bytes. Callers
who have to deal with bytes will need to change to using diff_bytes().
Use case: Mercurial's test runner uses difflib to compare current hg
output with known good output. But Mercurial's output is just bytes,
since it can contain:
* file contents (arbitrary unknown encoding)
* filenames (arbitrary unknown encoding)
* usernames and commit messages (usually UTF-8, but not guaranteed
because old versions of Mercurial did not enforce it)
* user messages (locale encoding)
Since the output of any given hg command can include text in multiple
encodings, it is hopeless to try to treat it as decodable Unicode
text. It's just bytes, all the way down.
This is an elaboration of a patch by Terry Reedy.
This fix is a superset of the functionality introduced by the issue #19494
enhancement, and supersedes that fix. Instead of a new handler, we have a new
password manager that tracks whether we should send the auth for a given uri.
This allows us to say "always send", satisfying #19494, or track that we've
succeeded in auth and send the creds right away on every *subsequent* request.
The support for using the password manager is added to AbstractBasicAuth,
which means the proxy handler also now can handle prior auth if passed
the new password manager.
Patch by Akshit Khurana, docs mostly by me.
The concept of .pyo files no longer exists. Now .pyc files have an
optional `opt-` tag which specifies if any extra optimizations beyond
the peepholer were applied.
instead of raising InterruptedError if the connection is interrupted by
signals, signal handlers don't raise an exception and the socket is blocking or
has a timeout.
socket.socket.connect() still raise InterruptedError for non-blocking sockets.
timeout when interrupted by a signal, except if the signal handler raises an
exception. This change is part of the PEP 475.
The asyncore and selectors module doesn't catch the InterruptedError exception
anymore when calling select.select(), since this function should not raise
InterruptedError anymore.
Setting attributes key, value and coded_value directly now is deprecated.
update() and setdefault() now transform and check keys. Comparing for
equality now takes into account attributes key, value and coded_value.
copy() now returns a Morsel, not a dict. repr() now contains all attributes.
Optimized checking keys and quoting values. Added new tests.
Original patch by Demian Brecht.
importlib.abc.Loader.exec_module() is also defined.
Before this change, create_module() was optional **and** could return
None to trigger default semantics. This change now reduces the
options for choosing default semantics to one and in the most
backporting-friendly way (define create_module() to return None).
This auth handler adds the Authorization header to the first
HTTP request rather than waiting for a HTTP 401 Unauthorized
response from the server as the default HTTPBasicAuthHandler
does.
This allows working with websites like https://api.github.com which do
not follow the strict interpretation of RFC, but more the dicta in the
end of section 2 of RFC 2617:
> A client MAY preemptively send the corresponding Authorization
> header with requests for resources in that space without receipt
> of another challenge from the server. Similarly, when a client
> sends a request to a proxy, it may reuse a userid and password in
> the Proxy-Authorization header field without receiving another
> challenge from the proxy server. See section 4 for security
> considerations associated with Basic authentication.
Patch by Matej Cepl.
Some time ago we changed the docs to consistently use the term 'bytes-like
object' in all the contexts where bytes, bytearray, memoryview, etc are used.
This patch (by Ezio Melotti) completes that work by changing the error
messages that previously reported that certain types did "not support the
buffer interface" to instead say that a bytes-like object is required. (The
glossary entry for bytes-like object references the discussion of the buffer
protocol in the docs.)
A debian code search (by Tshepang Lekhonkhobe) turned up only one package
checking email.__version__...and it was the 2.7-only mailman package. Since
Barry approves this change, it seems safe enough to make it...
Both compileall.compile_dir() and the CLI for compileall now allow for
specifying how many workers to use (or 0 to use all CPUs).
Thanks to Claudiu Popa for the patch.
threading.Lock.acquire(), threading.RLock.acquire() and socket operations now
use a monotonic clock, instead of the system clock, when a timeout is used.
Patch by Milan Oberkirch, developed as part of his 2014 GSOC project.
Note that this also fixes a bug in mock_socket ('getpeername' was returning a
simple string instead of the tuple required for IPvX protocols), a bug in
DebugServer with respect to handling binary data (should have been fixed when
decode_data was introduced, but wasn't found until this patch was written),
and a long-standing bug in DebugServer (it was printing an extra blank line at
the end of the displayed message text).
name, and use it in the representation of a generator (``repr(gen)``). The
default name of the generator (``__name__`` attribute) is now get from the
function instead of the code. Use ``gen.gi_code.co_name`` to get the name of
the code.
Otherwise smtpd is restricted to 7bit clean data, since even if the
incoming data is actually utf-8, it will often break things to decode
it before parsing the message.
Patch by Maciej Szulik, with some adjustments (mostly the warning
support).
Along the way, dismantle importlib._bootstrap._SpecMethods as it was
no longer relevant and constructing the new function required
partially dismantling the class anyway.
PyObject_Calloc(), _PyObject_GC_Calloc(). bytes(int) and bytearray(int) are now
using ``calloc()`` instead of ``malloc()`` for large objects which is faster
and use less memory (until the bytearray buffer is filled with data).
This makes doctest work like unittest: if the test case is empty, that
just means there are zero tests run, it's not an error. The existing
behavior was broken, since it only gave an error if there were *no*
docstrings, and zero tests run if there were docstrings but none of them
contained tests. So this makes it self-consistent as well.
Patch by Glenn Jones.
Yury Selivanov
Berker Peksag
Victor Stinner
Steve Dower
Zachary Ware
Nick Coghlan
Benjamin Peterson
Senthil Kumaran
Ned Deily
Tal Einat
Serhiy Storchaka
Guido van Rossum
Terry Jan Reedy
Gregory P. Smith
R David Murray
Raymond Hettinger
Greg Ward
Eric V. Smith
Larry Hastings
Brett Cannon
Charles-François Natali
Stefan Krah
Georg Brandl
Mark Dickinson
Ezio Melotti
Giampaolo Rodola'
Donald Stufft
Antoine Pitrou
Jesus Cea