patch for sharing single character Unicode objects.
Martin's patch had to be reworked in a number of ways to take Unicode
resizing into consideration as well. Here's what the updated patch
implements:
* Single character Unicode strings in the Latin-1 range are shared
(not only ASCII chars as in Martin's original patch).
* The ASCII and Latin-1 codecs make use of this optimization,
providing a noticable speedup for single character strings. Most
Unicode methods can use the optimization as well (by virtue
of using PyUnicode_FromUnicode()).
* Some code cleanup was done (replacing memcpy with Py_UNICODE_COPY)
* The PyUnicode_Resize() can now also handle the case of resizing
unicode_empty which previously resulted in an error.
* Modified the internal API _PyUnicode_Resize() and
the public PyUnicode_Resize() API to handle references to
shared objects correctly. The _PyUnicode_Resize() signature
changed due to this.
* Callers of PyUnicode_FromUnicode() may now only modify the Unicode
object contents of the returned object in case they called the API
with NULL as content template.
Note that even though this patch passes the regression tests, there
may still be subtle bugs in the sharing code.
sees it (test_iter.py is unchanged).
- Added a tp_iternext slot, which calls the iterator's next() method;
this is much faster for built-in iterators over built-in types
such as lists and dicts, speeding up pybench's ForLoop with about
25% compared to Python 2.1. (Now there's a good argument for
iterators. ;-)
- Renamed the built-in sequence iterator SeqIter, affecting the C API
functions for it. (This frees up the PyIter prefix for generic
iterator operations.)
- Added PyIter_Check(obj), which checks that obj's type has a
tp_iternext slot and that the proper feature flag is set.
- Added PyIter_Next(obj) which calls the tp_iternext slot. It has a
somewhat complex return condition due to the need for speed: when it
returns NULL, it may not have set an exception condition, meaning
the iterator is exhausted; when the exception StopIteration is set
(or a derived exception class), it means the same thing; any other
exception means some other error occurred.
new slot tp_iter in type object, plus new flag Py_TPFLAGS_HAVE_ITER
new C API PyObject_GetIter(), calls tp_iter
new builtin iter(), with two forms: iter(obj), and iter(function, sentinel)
new internal object types iterobject and calliterobject
new exception StopIteration
new opcodes for "for" loops, GET_ITER and FOR_ITER (also supported by dis.py)
new magic number for .pyc files
new special method for instances: __iter__() returns an iterator
iteration over dictionaries: "for x in dict" iterates over the keys
iteration over files: "for x in file" iterates over lines
TODO:
documentation
test suite
decide whether to use a different way to spell iter(function, sentinal)
decide whether "for key in dict" is a good idea
use iterators in map/filter/reduce, min/max, and elsewhere (in/not in?)
speed tuning (make next() a slot tp_next???)
I know some people don't like this -- if it's really controversial,
I'll take it out again. (If it's only Alex Martelli who doesn't like
it, that doesn't count as "real controversial" though. :-)
That's why this is a separate checkin from the iterators stuff I'm
about to check in next.
PyTuple_New() could *conceivably* clear the dict, so move the test for
an empty dict after the tuple allocation. It means that we waste time
allocating and deallocating a 2-tuple when the dict is empty, but who
cares. It also means that when the dict is empty *and* there's no
memory to allocate a 2-tuple, we raise MemoryError, not KeyError --
but that may actually a good idea: if there's no room for a lousy
2-tuple, what are the chances that there's room for a KeyError
instance?
and reported to python-dev: because we were calling dict_resize() in
PyDict_Next(), and because GC's dict_traverse() uses PyDict_Next(),
and because PyTuple_New() can cause GC, and because dict_items() calls
PyTuple_New(), it was possible for dict_items() to have the dict
resized right under its nose.
The solution is convoluted, and touches several places: keys(),
values(), items(), popitem(), PyDict_Next(), and PyDict_SetItem().
There are two parts to it. First, we no longer call dict_resize() in
PyDict_Next(), which seems to solve the immediate problem. But then
PyDict_SetItem() must have a different policy about when *it* calls
dict_resize(), because we want to guarantee (e.g. for an algorithm
that Jeremy uses in the compiler) that you can loop over a dict using
PyDict_Next() and make changes to the dict as long as those changes
are only value replacements for existing keys using PyDict_SetItem().
This is done by resizing *after* the insertion instead of before, and
by remembering the size before we insert the item, and if the size is
still the same, we don't bother to even check if we might need to
resize. An additional detail is that if the dict starts out empty, we
must still resize it before the insertion.
That was the first part. :-)
The second part is to make keys(), values(), items(), and popitem()
safe against side effects on the dict caused by allocations, under the
assumption that if the GC can cause arbitrary Python code to run, it
can cause other threads to run, and it's not inconceivable that our
dict could be resized -- it would be insane to write code that relies
on this, but not all code is sane.
Now, I have this nagging feeling that the loops in lookdict probably
are blissfully assuming that doing a simple key comparison does not
change the dict's size. This is not necessarily true (the keys could
be class instances after all). But that's a battle for another day.
"%#x" % 0
blew up, at heart because C sprintf supplies a base marker if and only if
the value is not 0. I then fixed that, by tolerating C's inconsistency
when it does %#x, and taking away that *Python* produced 0x0 when
formatting 0L (the "long" flavor of 0) under %#x itself. But after talking
with Guido, we agreed it would be better to supply 0x for the short int
case too, despite that it's inconsistent with C, because C is inconsistent
with itself and with Python's hex(0) (plus, while "%#x" % 0 didn't work
before, "%#x" % 0L *did*, and returned "0x0"). Similarly for %#X conversion.
http://sourceforge.net/tracker/index.php?func=detail&aid=415514&group_id=5470&atid=105470
For short ints, Python defers to the platform C library to figure out what
%#x should do. The code asserted that the platform C returned a string
beginning with "0x". However, that's not true when-- and only when --the
*value* being formatted is 0. Changed the code to live with C's inconsistency
here. In the meantime, the problem does not arise if you format a long 0 (0L)
instead. However, that's because the code *we* wrote to do %#x conversions on
longs produces a leading "0x" regardless of value. That's probably wrong too:
we should drop leading "0x", for consistency with C, when (& only when) formatting
0L. So I changed the long formatting code to do that too.
must now initialize the extra field used by the weak-ref machinery to
NULL themselves, to avoid having to require PyObject_INIT() to check
if the type supports weak references and do it there. This causes less
work to be done for all objects (the type object does not need to be
consulted to check for the Py_TPFLAGS_HAVE_WEAKREFS bit).
frees. Note there doesn't seem to be any way to test LocalsToFast(),
because the instructions that trigger it are illegal in nested scopes
with free variables.
Fix allocation strategy for cells that are also formal parameters.
Instead of emitting LOAD_FAST / STORE_DEREF pairs for each parameter,
have the argument handling code in eval_code2() do the right thing.
A side-effect of this change is that cell variables that are also
arguments are listed at the front of co_cellvars in the order they
appear in the argument list.
hashable
This patch changes the behavior of slice objects in the following
manner:
- Slice objects are now comparable with other slice objects as though
they were logically tuples of (start,stop,step). The tuple is not
created in the comparison function, but the comparison behavior is
logically equivalent.
- Slice objects are not hashable. With the above change to being
comparable, slice objects now cannot be used as keys in dictionaries.
[I've edited the patch for style. Note that this fixes the problem
that dict[i:j] seemed to work but was meaningless. --GvR]
with free variables. Thanks to Martin v. Loewis for finding two of
the problems. This fixes SF buf 405583.
There is also a C API change: PyFrame_New() is reverting to its
pre-2.1 signature. The change introduced by nested scopes was a
mistake. XXX Is this okay between beta releases?
cell_clear(), the GC helper, must decref its reference to break
cycles.
frame_dealloc() must dealloc all cell vars and free vars in addition
to locals.
eval_code2() setup code must INCREF cells it copies out of the
closure.
The STORE_DEREF opcode implementation must DECREF the object it passes
to PyCell_Set().
May or may not be related to bug 407680 (obmalloc.c - looks like it's
corrupted). This repairs the illegal vrbl names, but leaves a pile of
illegal macro names (_THIS_xxx, _SYSTEM_xxx, _SET_HOOKS, _FETCH_HOOKS).
- In _portable_ftell(), try fgetpos() before ftello() and ftell64().
I ran into a situation on a 64-bit capable Linux where the C
library's ftello() and ftell64() returned negative numbers despite
fpos_t and off_t both being 64-bit types; fgetpos() did the right
thing.
- Define a new typedef, Py_off_t, which is either fpos_t or off_t,
depending on which one is 64 bits. This removes the need for a lot
of #ifdefs later on. (XXX Should this be moved to pyport.h? That
file currently seems oblivious to large fille support, so for now
I'll leave it here where it's needed.)
set a function attribute on a method (either bound or unbound). This
reverts to Python 2.0 behavior that no attributes of the method are
writable, but provides a more informative error message.
release the interned string dictionary. This is useful for memory
use debugging because it eliminates a huge source of noise from the
reports. Only defined when INTERN_STRINGS is defined.
and the test for errors, so that an error in the default compare
doesn't go undetected. This fixes SF Bug #132933 (submitted by
effbot) -- list.sort doesn't detect comparision errors.
This fixes SF bug #132008, reported by Warren J. Hack.
The copyright for this patch (and this patch only) belongs to CNRI, as
part of the (yet to be issued) 1.6.1 release.
This is now checked into the HEAD branch. Tim will check in a test
case to check for this specific bug, and an assertion in
PyArgs_ParseTuple() to catch similar bugs in the future.
This change eliminates an extra malloc/free when a frame with free
variables is created. Any cell vars or free vars are stored in
f_localsplus after the locals and before the stack.
eval_code2() fills in the appropriate values after handling
initialization of locals.
To track the size the frame has an f_size member that tracks the total
size of f_localsplus. It used to be implicitly f_nlocals + f_stacksize.
* Removed func_hash and func_compare, so they can be treated as immutable
content-less objects (address hash and comparison)
* Added tests to that affect to test_funcattrs (also testing func_code
is writable)
* Reverse meaning of tests in test_opcodes which checked identical code
gets identical functions
Marc-André Lemburg
Guido van Rossum
Tim Peters
Jeremy Hylton
Martin v. Löwis
Fred Drake
Neil Schemenauer
Barry Warsaw
Fredrik Lundh
Moshe Zadka