Along the way, dismantle importlib._bootstrap._SpecMethods as it was
no longer relevant and constructing the new function required
partially dismantling the class anyway.
__loader__ is not set on a module. This brings the exception in line
with when __loader__ is None (which is equivalent to not having the
attribute defined).
state of the import system. Also make importlib.invalidate_caches()
work with sys.meta_path instead of sys.path_importer_cache to
completely separate the path-based import system from the overall
import system.
Patch by Eric Snow.
The long-term goal is to deprecate imp.find_module() in favour of this
API, but it will take some time as some APIs explicitly return/use what
imp.find_module() returns.
This required updating the code to use posix instead of os. This is
all being done to make bootstrapping easier to removing dependencies
that are kept in importlib.__init__ and thus outside of the single
file to bootstrap from.
This required moving the class from importlib/abc.py into
importlib/_bootstrap.py and jiggering some code to work better with the class.
This included changing how the file finder worked to better meet import
semantics. This also led to fixing importlib to handle the empty string from
sys.path as import currently does (and making me wish we didn't support that
instead just required people to insert '.' instead to represent cwd).
It also required making the new set_data abstractmethod create
any needed subdirectories implicitly thanks to __pycache__ (it was either this
or grow the SourceLoader ABC to gain an 'exists' method and either a mkdir
method or have set_data with no data arg mean to create a directory).
Lastly, as an optimization the file loaders cache the file path where the
finder found something to use for loading (this is thanks to having a
sourceless loader separate from the source loader to simplify the code and
cut out stat calls).
Unfortunately test_runpy assumed a loader would always work for a module, even
if you changed from underneath it what it was expected to work with. By simply
dropping the previous loader in test_runpy so the proper loader can be returned
by the finder fixed the failure.
At this point importlib deviates from import on two points:
1. The exception raised when trying to import a file is different (import does
an explicit file check to print a special message, importlib just says the path
cannot be imported as if it was just some module name).
2. the co_filename on a code object is not being set to where bytecode was
actually loaded from instead of where the marshalled code object originally
came from (a solution for this has already been agreed upon on python-dev but has
not been implemented yet; issue8611).
The main io module now uses the C implementation. The Python one still exists
in Lib/_pyio.py for ease of testing new features and usefulness to other
implementers.
The rewrite was done by Antoine Pitrou and Amaury Forgeot d'Arc. I was slightly
helpful at the end. :)
Following are the log messages from the io-c branch:
Merged revisions 68683-68685,68687-68689,68693,68704,68741-68743,68745,68747,68752-68754,68756,68758,68812,68816-68817,68820-68822,68824-68825,68828,68876-68877,69037,69044,69104,69115,69194,69626-69629,69636,69638,69641-69642,69644-69654,69656-69661,69671,69677,69812-69815,69817,69827-69830,69839,69841-69845,69848,69850,69852,69854,69860,69865-69866,69868,69872-69873,69885,69888,69891-69893,69911,69913-69916,69963,70033,70035,70038,70041-70048,70067-70070,70075,70112,70133,70135,70140 via svnmerge from
svn+ssh://pythondev@svn.python.org/python/branches/io-c
........
r68683 | antoine.pitrou | 2009-01-17 17:13:48 -0600 (Sat, 17 Jan 2009) | 3 lines
Merge in changes from the io-c sandbox. Tests will follow in separate commits.
........
r68684 | antoine.pitrou | 2009-01-17 17:17:26 -0600 (Sat, 17 Jan 2009) | 3 lines
Fixes and additions to test_io.py
........
r68685 | antoine.pitrou | 2009-01-17 17:22:04 -0600 (Sat, 17 Jan 2009) | 1 line
Fix test_fileio
........
r68687 | antoine.pitrou | 2009-01-17 17:35:11 -0600 (Sat, 17 Jan 2009) | 3 lines
Add dependency to _iomodule.h for the various C sources
........
r68688 | antoine.pitrou | 2009-01-17 17:38:18 -0600 (Sat, 17 Jan 2009) | 3 lines
These precautions are not needed anymore!
........
r68689 | antoine.pitrou | 2009-01-17 17:41:48 -0600 (Sat, 17 Jan 2009) | 3 lines
Fix another test
........
r68693 | antoine.pitrou | 2009-01-17 17:49:58 -0600 (Sat, 17 Jan 2009) | 3 lines
Fix test_uu (which was using private attributes of TextIOWrapper)
........
r68704 | antoine.pitrou | 2009-01-17 18:45:29 -0600 (Sat, 17 Jan 2009) | 3 lines
Most io sources are Py_ssize_t-clean (I don't know about bytesio and stringio)
........
r68741 | antoine.pitrou | 2009-01-18 15:20:30 -0600 (Sun, 18 Jan 2009) | 3 lines
Check return type in TextIOWrapper.__next__
........
r68742 | antoine.pitrou | 2009-01-18 15:28:48 -0600 (Sun, 18 Jan 2009) | 4 lines
Make binary buffered readline and iteration much faster
(8x as fast as the IOBase generic implementation)
........
r68743 | antoine.pitrou | 2009-01-18 15:47:47 -0600 (Sun, 18 Jan 2009) | 3 lines
Reinsert test_io_after_close (was removed by mistake)
........
r68745 | antoine.pitrou | 2009-01-18 16:16:06 -0600 (Sun, 18 Jan 2009) | 3 lines
Add read, read1 and write methods to BufferedIOBase
........
r68747 | antoine.pitrou | 2009-01-18 16:35:58 -0600 (Sun, 18 Jan 2009) | 3 lines
Kill test failure
........
r68752 | amaury.forgeotdarc | 2009-01-18 17:05:43 -0600 (Sun, 18 Jan 2009) | 3 lines
Fix a segfault when e.g a BufferedReader is created with a FileIO in
read mode.
........
r68753 | antoine.pitrou | 2009-01-18 17:13:09 -0600 (Sun, 18 Jan 2009) | 3 lines
Add truncate() to text IO objects
........
r68754 | antoine.pitrou | 2009-01-18 17:51:08 -0600 (Sun, 18 Jan 2009) | 3 lines
Remove IOBase.__del__ and replace it with custom code with tp_dealloc
........
r68756 | antoine.pitrou | 2009-01-18 18:10:16 -0600 (Sun, 18 Jan 2009) | 3 lines
Remove irrelevant comment.
........
r68758 | antoine.pitrou | 2009-01-18 18:36:16 -0600 (Sun, 18 Jan 2009) | 3 lines
in importlib:_fileio._FileIO -> _io.FileIO
........
r68812 | antoine.pitrou | 2009-01-20 14:15:51 -0600 (Tue, 20 Jan 2009) | 3 lines
Add garbage collection support to FileIO objects
........
r68816 | antoine.pitrou | 2009-01-20 14:56:28 -0600 (Tue, 20 Jan 2009) | 3 lines
Add GC support to Buffered and Text IO objects
........
r68817 | antoine.pitrou | 2009-01-20 15:19:45 -0600 (Tue, 20 Jan 2009) | 3 lines
Add some file headers
........
r68820 | antoine.pitrou | 2009-01-20 15:29:59 -0600 (Tue, 20 Jan 2009) | 3 lines
Add class TextIOBase
........
r68821 | antoine.pitrou | 2009-01-20 15:36:16 -0600 (Tue, 20 Jan 2009) | 3 lines
Add properties to TextIOBase
........
r68822 | antoine.pitrou | 2009-01-20 15:41:19 -0600 (Tue, 20 Jan 2009) | 3 lines
Disable the pure Python TextIOBase class, and inject C the implementation instead
........
r68824 | antoine.pitrou | 2009-01-20 16:36:28 -0600 (Tue, 20 Jan 2009) | 3 lines
Fix two leaks
........
r68825 | antoine.pitrou | 2009-01-20 16:38:29 -0600 (Tue, 20 Jan 2009) | 3 lines
FileIO.name is just a plain attribute, we can set it directly
........
r68828 | antoine.pitrou | 2009-01-20 17:06:33 -0600 (Tue, 20 Jan 2009) | 3 lines
Speed up closed checks on text IO objects. Good for a 25% speedup on small ops.
........
r68876 | antoine.pitrou | 2009-01-23 17:01:25 -0600 (Fri, 23 Jan 2009) | 3 lines
Two typos
........
r68877 | antoine.pitrou | 2009-01-23 18:13:20 -0600 (Fri, 23 Jan 2009) | 3 lines
Remove two unused functions
........
r69037 | amaury.forgeotdarc | 2009-01-27 17:10:25 -0600 (Tue, 27 Jan 2009) | 2 lines
Update the win32 project files
........
r69044 | antoine.pitrou | 2009-01-27 18:51:07 -0600 (Tue, 27 Jan 2009) | 3 lines
Improve heuristic in IncrementalNewlineDecoder + some micro-optimizations
........
r69104 | antoine.pitrou | 2009-01-29 15:23:42 -0600 (Thu, 29 Jan 2009) | 3 lines
Fix some crashers found by Victor
........
r69115 | hirokazu.yamamoto | 2009-01-29 20:36:28 -0600 (Thu, 29 Jan 2009) | 1 line
Updated VC6 project file.
........
r69194 | antoine.pitrou | 2009-02-01 16:57:18 -0600 (Sun, 01 Feb 2009) | 3 lines
Fix downcasting warnings in 32-bit mode with 64-bit offsets (Windows)
........
r69626 | benjamin.peterson | 2009-02-14 17:33:34 -0600 (Sat, 14 Feb 2009) | 1 line
only catch AttributeError and UnsupportedOperation
........
r69627 | benjamin.peterson | 2009-02-14 21:35:28 -0600 (Sat, 14 Feb 2009) | 1 line
give the IO module its own state and store the os and locale modules in it
........
r69628 | benjamin.peterson | 2009-02-14 22:08:32 -0600 (Sat, 14 Feb 2009) | 1 line
put interned strings in the module state structure
........
r69629 | benjamin.peterson | 2009-02-14 22:15:29 -0600 (Sat, 14 Feb 2009) | 1 line
put UnsupportedOperation in the module state
........
r69636 | benjamin.peterson | 2009-02-15 08:31:42 -0600 (Sun, 15 Feb 2009) | 1 line
dealloc unsupported_operation
........
r69638 | benjamin.peterson | 2009-02-15 09:24:45 -0600 (Sun, 15 Feb 2009) | 1 line
actually test the C implementation
........
r69641 | benjamin.peterson | 2009-02-15 10:12:37 -0600 (Sun, 15 Feb 2009) | 5 lines
make interned strings globals again ;(
putting them in the module state was asking for trouble when the module
was dealloced before the classes in it were
........
r69642 | benjamin.peterson | 2009-02-15 10:19:45 -0600 (Sun, 15 Feb 2009) | 1 line
actually test the python implementations
........
r69644 | antoine.pitrou | 2009-02-15 11:59:30 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix memory leak in destructor when a Python class inherits from IOBase (or an IOBase-derived type)
........
r69645 | antoine.pitrou | 2009-02-15 12:23:26 -0600 (Sun, 15 Feb 2009) | 3 lines
Add a warning about the embarassing state of IOBase finalization
........
r69646 | antoine.pitrou | 2009-02-15 13:14:42 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix opening of 8-bit filenames with FileIO
........
r69647 | antoine.pitrou | 2009-02-15 13:20:22 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix leak in FileIO constructor
........
r69648 | antoine.pitrou | 2009-02-15 13:58:16 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix some refleaks
........
r69649 | antoine.pitrou | 2009-02-15 14:05:13 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix a leak in IOBase.writelines
........
r69650 | antoine.pitrou | 2009-02-15 14:11:56 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix memory leak in BufferedWriter.truncate
........
r69651 | antoine.pitrou | 2009-02-15 14:25:34 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix a leak in TextIOWrapper.seek
........
r69652 | antoine.pitrou | 2009-02-15 14:26:28 -0600 (Sun, 15 Feb 2009) | 3 lines
Unify implementations of truncate for buffered objects
........
r69653 | antoine.pitrou | 2009-02-15 15:15:15 -0600 (Sun, 15 Feb 2009) | 3 lines
Fix more leaks in TextIOWrapper
........
r69654 | antoine.pitrou | 2009-02-15 15:21:57 -0600 (Sun, 15 Feb 2009) | 3 lines
Smaller chunk size for a faster test
........
r69656 | benjamin.peterson | 2009-02-15 17:29:48 -0600 (Sun, 15 Feb 2009) | 1 line
braces make this much clearer
........
r69657 | benjamin.peterson | 2009-02-15 17:46:07 -0600 (Sun, 15 Feb 2009) | 1 line
use the correct macro
........
r69658 | antoine.pitrou | 2009-02-15 19:38:59 -0600 (Sun, 15 Feb 2009) | 5 lines
Fix crash in test_urllib2_localnet in debug mode. It was due to an HTTPResponse
object being revived when calling its close() method in IOBase's tp_dealloc.
_PyIOBase_finalize() starts looking scary...
........
r69659 | benjamin.peterson | 2009-02-15 20:55:48 -0600 (Sun, 15 Feb 2009) | 1 line
fix segfault on initialization failing
........
r69660 | benjamin.peterson | 2009-02-15 21:09:31 -0600 (Sun, 15 Feb 2009) | 1 line
apparently locale.getprefferedencoding() can raise a ImportError, too
........
r69661 | benjamin.peterson | 2009-02-15 21:54:15 -0600 (Sun, 15 Feb 2009) | 1 line
it's amazing this worked at all; I was using the wrong structs!
........
r69671 | benjamin.peterson | 2009-02-16 08:38:27 -0600 (Mon, 16 Feb 2009) | 1 line
add garbage collection support to bytesio
........
r69677 | benjamin.peterson | 2009-02-16 10:31:03 -0600 (Mon, 16 Feb 2009) | 5 lines
reduce ImportError catching code duplication
I'm not sure this makes the code clearer with its new gotos, but
at least I added a big fat comment
........
r69812 | antoine.pitrou | 2009-02-20 13:50:16 -0600 (Fri, 20 Feb 2009) | 3 lines
_StringIO now belongs to the _io modules, rather to its own _stringio module
........
r69813 | antoine.pitrou | 2009-02-20 13:58:22 -0600 (Fri, 20 Feb 2009) | 3 lines
Add a test for StringIO properties
........
r69814 | antoine.pitrou | 2009-02-20 14:06:03 -0600 (Fri, 20 Feb 2009) | 3 lines
Reimplement a few trivial StringIO functions and properties in C
........
r69815 | antoine.pitrou | 2009-02-20 14:13:11 -0600 (Fri, 20 Feb 2009) | 3 lines
Add the line_buffering property to TextIOWrapper, and test for it
........
r69817 | antoine.pitrou | 2009-02-20 14:45:50 -0600 (Fri, 20 Feb 2009) | 4 lines
Allow IncrementalNewlineDecoder to take unicode objects as decoding input if the decoder parameter is None
This will help rewriting StringIO to C
........
r69827 | antoine.pitrou | 2009-02-20 19:00:30 -0600 (Fri, 20 Feb 2009) | 3 lines
Rewrite most of StringIO in C. Some almost empty stubs remain to be converted.
........
r69828 | antoine.pitrou | 2009-02-20 19:09:25 -0600 (Fri, 20 Feb 2009) | 3 lines
Plug a leak, and remove an unused string
........
r69829 | benjamin.peterson | 2009-02-20 20:02:28 -0600 (Fri, 20 Feb 2009) | 1 line
this assertions makes more sense here
........
r69830 | benjamin.peterson | 2009-02-20 20:03:04 -0600 (Fri, 20 Feb 2009) | 1 line
PyModule_AddObject can fail; simplify this code with a macro
........
r69839 | antoine.pitrou | 2009-02-21 12:54:01 -0600 (Sat, 21 Feb 2009) | 3 lines
StringIO is now written entirely in C (and blazingly fast)
........
r69841 | benjamin.peterson | 2009-02-21 14:05:40 -0600 (Sat, 21 Feb 2009) | 1 line
split the Python implementation of io into another module and rewrite the tests to test both implementations
........
r69842 | benjamin.peterson | 2009-02-21 14:10:00 -0600 (Sat, 21 Feb 2009) | 1 line
closed is not a function
........
r69843 | benjamin.peterson | 2009-02-21 14:13:04 -0600 (Sat, 21 Feb 2009) | 1 line
fix __all__ test
........
r69844 | benjamin.peterson | 2009-02-21 14:21:24 -0600 (Sat, 21 Feb 2009) | 1 line
fix the rest of the Misc tests
........
r69845 | benjamin.peterson | 2009-02-21 14:26:59 -0600 (Sat, 21 Feb 2009) | 1 line
RawIOBase is better for FileIO
........
r69848 | benjamin.peterson | 2009-02-21 15:33:53 -0600 (Sat, 21 Feb 2009) | 1 line
fix some more tests broken by bag argument validation
........
r69850 | benjamin.peterson | 2009-02-21 16:16:42 -0600 (Sat, 21 Feb 2009) | 1 line
make the python IncrementalNewineDecoder support a None decoder
........
r69852 | benjamin.peterson | 2009-02-21 16:36:09 -0600 (Sat, 21 Feb 2009) | 1 line
fix a BlockingIOError.characters_written bug
........
r69854 | benjamin.peterson | 2009-02-21 16:49:02 -0600 (Sat, 21 Feb 2009) | 1 line
check whence
........
r69860 | benjamin.peterson | 2009-02-21 17:42:50 -0600 (Sat, 21 Feb 2009) | 1 line
fix some of these Misbehaving io tests
........
r69865 | benjamin.peterson | 2009-02-21 18:59:52 -0600 (Sat, 21 Feb 2009) | 1 line
don't use super here()
........
r69866 | benjamin.peterson | 2009-02-21 19:05:28 -0600 (Sat, 21 Feb 2009) | 1 line
use implementation specific classes
........
r69868 | benjamin.peterson | 2009-02-21 22:12:05 -0600 (Sat, 21 Feb 2009) | 1 line
use a more DRY friendly approach to injecting module contents into test classes
........
r69872 | antoine.pitrou | 2009-02-22 13:39:45 -0600 (Sun, 22 Feb 2009) | 3 lines
Sanitize destructor behaviour of IOBase. Now Python-defined attributes can be accessed from close().
........
r69873 | antoine.pitrou | 2009-02-22 13:50:14 -0600 (Sun, 22 Feb 2009) | 4 lines
Only set the internal fd after it has been checked to be valid
(otherwise, the destructor will attempt to close it)
........
r69885 | benjamin.peterson | 2009-02-22 15:30:14 -0600 (Sun, 22 Feb 2009) | 1 line
convert some other tests to use both io implementations
........
r69888 | antoine.pitrou | 2009-02-22 17:03:16 -0600 (Sun, 22 Feb 2009) | 3 lines
Silence all exceptions when finalizing
........
r69891 | benjamin.peterson | 2009-02-22 17:27:24 -0600 (Sun, 22 Feb 2009) | 1 line
convert another test to test both io implementations
........
r69892 | benjamin.peterson | 2009-02-22 17:32:15 -0600 (Sun, 22 Feb 2009) | 1 line
help poor people like me to find their io tests (did I miss any?)
........
r69893 | benjamin.peterson | 2009-02-22 17:37:56 -0600 (Sun, 22 Feb 2009) | 1 line
put a big note in the test telling people to write tests for both implementations now
........
r69911 | antoine.pitrou | 2009-02-23 13:57:18 -0600 (Mon, 23 Feb 2009) | 3 lines
expose DEFAULT_BUFFER_SIZE again (fixes a bunch of test failures)
........
r69913 | antoine.pitrou | 2009-02-23 14:10:30 -0600 (Mon, 23 Feb 2009) | 4 lines
Do the cyclic garbage collection tests only on the C version.
The Python version is helpless as it uses __del__.
........
r69914 | antoine.pitrou | 2009-02-23 14:21:41 -0600 (Mon, 23 Feb 2009) | 3 lines
Adapt test_largefile to test both implementations
........
r69915 | antoine.pitrou | 2009-02-23 14:25:14 -0600 (Mon, 23 Feb 2009) | 3 lines
One small failure
........
r69916 | antoine.pitrou | 2009-02-23 14:28:33 -0600 (Mon, 23 Feb 2009) | 3 lines
Add a comment, at BP's request
........
r69963 | antoine.pitrou | 2009-02-25 09:42:59 -0600 (Wed, 25 Feb 2009) | 3 lines
Add a test of ABC inheritance
........
r70033 | antoine.pitrou | 2009-02-27 15:49:50 -0600 (Fri, 27 Feb 2009) | 3 lines
The base classes now are ABCs.
........
r70035 | benjamin.peterson | 2009-02-27 15:57:41 -0600 (Fri, 27 Feb 2009) | 1 line
good house keeping
........
r70038 | antoine.pitrou | 2009-02-27 17:05:23 -0600 (Fri, 27 Feb 2009) | 4 lines
Make the buffer allocation overflow tests specific to the C implementation, since the Python implementation resizes its buffers when needed rather than allocating them up front.
........
r70041 | benjamin.peterson | 2009-02-27 18:26:12 -0600 (Fri, 27 Feb 2009) | 1 line
kill java naming for sanity
........
r70042 | benjamin.peterson | 2009-02-27 18:28:53 -0600 (Fri, 27 Feb 2009) | 2 lines
timingTest is superseded by iobench
........
r70043 | antoine.pitrou | 2009-02-27 19:13:50 -0600 (Fri, 27 Feb 2009) | 3 lines
Remove the last traces of java naming in test_io
........
r70044 | antoine.pitrou | 2009-02-27 19:18:34 -0600 (Fri, 27 Feb 2009) | 3 lines
Better resource cleanup
........
r70045 | antoine.pitrou | 2009-02-27 19:29:00 -0600 (Fri, 27 Feb 2009) | 3 lines
Remove dubious uses of super(), and fix one test
........
r70046 | antoine.pitrou | 2009-02-27 19:31:00 -0600 (Fri, 27 Feb 2009) | 3 lines
Bump up CHUNK_SIZE (no need to make the Python version look slower than it is)
........
r70047 | benjamin.peterson | 2009-02-27 20:03:26 -0600 (Fri, 27 Feb 2009) | 1 line
fix typo
........
r70048 | benjamin.peterson | 2009-02-27 21:35:11 -0600 (Fri, 27 Feb 2009) | 1 line
move code to a better place
........
r70067 | benjamin.peterson | 2009-02-28 10:43:20 -0600 (Sat, 28 Feb 2009) | 4 lines
1. make sure to undo buffered read aheads in BufferedRandom.seek()
2. refill the buffer if have <= 0
3. fix the last failing test_io test!
........
r70068 | benjamin.peterson | 2009-02-28 10:57:50 -0600 (Sat, 28 Feb 2009) | 1 line
define read1() on the python implementation's BufferedIOBase
........
r70069 | benjamin.peterson | 2009-02-28 11:01:17 -0600 (Sat, 28 Feb 2009) | 1 line
document read1() in BufferedIOBase
........
r70070 | benjamin.peterson | 2009-02-28 11:06:42 -0600 (Sat, 28 Feb 2009) | 1 line
give credit where credit is due
........
r70075 | antoine.pitrou | 2009-02-28 13:34:59 -0600 (Sat, 28 Feb 2009) | 3 lines
Amaury's name
........
r70112 | antoine.pitrou | 2009-03-02 17:11:55 -0600 (Mon, 02 Mar 2009) | 4 lines
Looks like this is necessary in order to build cleanly under Windows
(someone correct this if it's wrong, I'm no Windows user)
........
r70133 | benjamin.peterson | 2009-03-03 15:23:32 -0600 (Tue, 03 Mar 2009) | 1 line
fix test_newline_property on _pyio.StringIO
........
r70135 | benjamin.peterson | 2009-03-03 15:47:30 -0600 (Tue, 03 Mar 2009) | 1 line
fix typos and inconsistencies. thanks to Daniel Diniz
........
r70140 | benjamin.peterson | 2009-03-03 16:21:10 -0600 (Tue, 03 Mar 2009) | 1 line
add the test from #5266
........
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).
Garvit Khatri
Mandeep Bhutani
Matthias Bussonnier
Eric Snow
Serhiy Storchaka
Brett Cannon
Ezio Melotti
Nick Coghlan
Antoine Pitrou
Florent Xicluna
Brett Cannon
Benjamin Peterson