Required updating code relying on other modules to switch to _bootstrap's
unique module requirements. This led to the realization that
get_code was being too liberal in its exception catching when calling set_data
by blindly grabbing IOError. Shifted the responsibility of safely ignoring
writes to a read-only path to set_data.
Importlib is still not relying on SourceLoader yet; requires creating a
SourcelessLoader and updating the source finder.
SourceLoader is a simplification of both PyLoader and PyPycLoader. If one only
wants to use source, then they need to only implement get_data and
get_filename. To also use bytecode -- sourceless loading is not supported --
then two abstract methods -- path_mtime and set_data -- need to be implemented.
Compared to PyLoader and PyPycLoader, there are less abstract methods
introduced and bytecode files become an optimization controlled by the ABC and
hidden from the user (this need came about as PEP 3147 showed that not treating
bytecode as an optimization can cause problems for compatibility).
PyLoader is deprecated in favor of SourceLoader. To be compatible from Python
3.1 onwards, a subclass need only use simple methods for source_path and
is_package. Otherwise conditional subclassing based on whether Python 3.1 or
Python 3.2 is being is the only change. The documentation and docstring for
PyLoader explain what is exactly needed.
PyPycLoader is deprecated also in favor of SourceLoader. Because PEP 3147
shifted bytecode path details so much, there is no foolproof way to provide
backwards-compatibility with SourceLoader. Because of this the class is simply
deprecated and users should move to SourceLoader (and optionally PyLoader for
Python 3.1). This does lead to a loss of support for sourceless loading
unfortunately.
At some point before Python 3.2 is released, SourceLoader will be moved over to
importlib._bootstrap so that the core code of importlib relies on the new code
instead of the old PyPycLoader code. This commit is being done now so that
there is no issue in having the API in Python 3.1a1.
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r78242 | brett.cannon | 2010-02-19 11:01:06 -0500 (Fri, 19 Feb 2010) | 5 lines
Importlib was not matching import's handling of .pyc files where it had less
then 8 bytes total in the file.
Fixes issues 7361 & 7875.
........
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r76146 | brett.cannon | 2009-11-07 15:55:05 -0800 (Sat, 07 Nov 2009) | 6 lines
When trying to write new bytecode, importlib was not catching the IOError
thrown if the file happened to be read-only to keep the failure silent.
Fixes issue #7187. Thanks, Dave Malcolm for the report and analysis of the
problem.
........
thrown if the file happened to be read-only to keep the failure silent.
Fixes issue #7187. Thanks, Dave Malcolm for the report and analysis of the
problem.
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r76113 | brett.cannon | 2009-11-04 17:17:22 -0800 (Wed, 04 Nov 2009) | 3 lines
importlib.test.source.util referenced variables in the 'finally' part of a
try/finally which may not have been set.
........
r76114 | brett.cannon | 2009-11-04 17:26:57 -0800 (Wed, 04 Nov 2009) | 6 lines
Use tempfile.mkdtemp() instead of tempfile.tempdir for where importlib places
source files for tests. Allows for concurrent execution of the tests by
preventing various executions from trampling each other.
Closes issue #7248.
........
second instead of some fixed number.
Keeps benchmark faster by putting a cap on total execution time. Before a run
using importlib took longer by some factor, but now it takes roughly the
same amount of time as using the built-in __import__.
__package__, it was used. This was incorrect since it could be set to None to
represent the fact that a proper value was unknown. Now None will trigger the
calculation for __package__.
Discovered when running importlib against test_importhooks.
attribute. Was throwing AttributeError before. Discovered when running
test_builtin against importlib.
This exception change is specific to importlib.__import__() and does not apply to
import_module() as it is being done for compatibility reasons only.
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r74584 | brett.cannon | 2009-08-29 20:47:36 -0700 (Sat, 29 Aug 2009) | 3 lines
Have importlib raise ImportError if None is found in sys.modules. This matches
current import semantics.
........
The file must be run using runpy. Certain tests are currently excluded from
being run as they have known failures based on golden value checks that fail
for various reasons (typically because __loader__ is not expected to be set on
modules). Running the tests with this file does discover some incompatibilites
in importlib that will be fixed in the near future (as noted currently in the
docstring).
importlib.abc.ExecutionLoader. PyLoader now inherits from this ABC instead of
InspectLoader directly. Both PyLoader and PyPycLoader provide concrete
implementations of get_filename in terms of source_path and bytecode_path.
svn+ssh://pythondev@svn.python.org/python/branches/py3k
........
r74107 | brett.cannon | 2009-07-19 20:19:18 -0700 (Sun, 19 Jul 2009) | 8 lines
Importlib's documentation said that importlib.abc.PyLoader inherited from
importlib.abc.ResourceLoader, when in fact it did not. Fixed the ABC to inherit
as documented.
This does in introduce an backwards-incompatiblity as the code in PyLoader
already required the single method ResourceLoader defined as an abstract
method.
........
importlib.abc.ResourceLoader, when in fact it did not. Fixed the ABC to inherit
as documented.
This doesn't introduce an backwards-incompatiblity as the code in PyLoader
already required the single method ResourceLoader defined as an abstract
method.
when a loader is given missing or bad code object bytecode. Unfortunately an
exception related to source paths was masking what the proper exception to test
should be. Making the test explicitly set the environment fixed the test.
The code being test was not affected.
case-sensitive filesystems -- which is not the default case. Along the way also
fixed the skipping of tests when sys.dont_write_bytecode is true.
Closes issue #5442 again.
entries in sys.path_importer_cache. While this differs from semantics in how
__import__ works, it prevents any implicit semantics from taking hold with
users.
(mostly stuff specified by PEP 302). There are two ABCs, PyLoader and
PyPycLoader, which help with implementing source and source/bytecode loaders by
implementing load_module in terms of other methods. This removes a lot of
gritty details loaders typically have to worry about.
+ Ditch using arguments to super().
+ Ditch subclassing from object directly.
+ Move directory check out of chaining path hook to file path hook/finder.
+ Rename some classes to better reflect they are finders, not importers.
load failure in relation to reloads. Also expose
importlib.util.module_for_loader to handle all of the details of this along
with making sure all current loaders behave nicely.
and relies much more on meta path finders to abstract out various parts of
import.
As part of this the semantics for import_module tightened up and now follow
__import__ much more closely (biggest thing is that the 'package' argument must
now already be imported, else a SystemError is raised).
FrozenImporter. Docs forthcoming.
I plan on all finders and loaders (and most likely hooks) to live
in imoprtlib.machinery. Utility stuff will end up in importlib.util.
Higher-level API stuff will stay on imoprtlib directly (e.g. import_module).
Brett Cannon
Benjamin Peterson
Barry Warsaw
Georg Brandl
R. David Murray
Raymond Hettinger