Also, add a testcase.
Formerly, the list_extend() code used several local variables to remember
its state across iterations. Since an iteration could call arbitrary
Python code, it was possible for the list state to be changed. The new
code uses dynamic structure references instead of C locals. So, they
are always up-to-date.
After list_resize() is called, its size has been updated but the new
cells are filled with NULLs. These needed to be filled before arbitrary
iteration code was called; otherwise, that code could attempt to modify
a list that was in a semi-invalid state. The solution was to change
the ob->size field back to a value reflecting the actual number of valid
cells.
The list resizing scheme only downsized when more than 16 elements were
removed in a single step: del a[100:120]. As a result, the list would
never shrink when popping elements off one at a time.
This patch makes it shrink whenever more than half of the space is unused.
Also, at Tim's suggestion, renamed _new_size to new_allocated. This makes
the code easier to understand.
comma expression in listpop() that was being returned. Still essentially
unused (as it is meant to be), but now the compiler thinks it is worth
*something* by having it incremented.
of no more than 8 elements cannot fail.
listpop(): Take advantage of that its calls to list_resize() and
list_ass_slice() can't fail. This is assert'ed in a debug build now, but
in an icky way. That is, you can't say:
assert(some_call() >= 0);
because then some_call() won't occur at all in a release build. So it
has to be a big pile of #ifdefs on Py_DEBUG (yuck), or the pleasant:
status = some_call();
assert(status >= 0);
But in that case, compilers may whine in a release build, because status
appears unused then. I'm not certain the ugly trick I used here will
convince all compilers to shut up about status (status is always "used" now,
as the first (ignored) clause in a comma expression).
impossible to remember, so renamed one to something obvious. Headed
off potential signed-vs-unsigned compiler complaints I introduced by
changing the type of a vrbl to unsigned. Removed the need for the
tedious explanation about "backward pointer loops" by looping on an
int instead.
result.
list_resize(): Document the intent. Code is increasingly relying on
subtle aspects of its behavior, and they deserve to be spelled out.
list_ass_slice(): A bit more simplification, by giving it a common
error exit and initializing more values.
Be clearer in comments about what "size" means (# of elements? # of
bytes?).
While the number of elements in a list slice must fit in an int, there's
no guarantee that the number of bytes occupied by the slice will. That
malloc() and memmove() take size_t arguments is a hint about that <wink>.
So changed to use size_t where appropriate.
ihigh - ilow should always be >= 0, but we never asserted that. We do
now.
The loop decref'ing the recycled slice had a subtle insecurity: C doesn't
guarantee that a pointer one slot *before* an array will compare "less
than" to a pointer within the array (it does guarantee that a pointer
one beyond the end of the array compares as expected). This was actually
an issue in KSR's C implementation, so isn't purely theoretical. Python
probably has other "go backwards" loops with a similar glitch.
list_clear() is OK (it marches an integer backwards, not a pointer).
though I tried to be very careful. This is a slight simplification, and it
adds a new feature: a small stack-allocated "recycled" array for the cases
when we don't remove too many items.
It allows PyList_SetSlice() to never fail if:
* you are sure that the object is a list; and
* you either do not remove more than 8 items, or clear the list.
This makes a number of other places in the source code correct again -- there
are some places that delete a single item without checking for MemoryErrors
raised by PyList_SetSlice(), or that clear the whole list, and sometimes the
context doesn't allow an error to be propagated.
invariants allows the ob_item != NULL check to be replaced with an
assertion.
* Added assertions to list_init() which document and verify that the
tp_new slot establishes the invariants. This may preclude a future
bug if a custom tp_new slot is written.
to NULL during the lifetime of the object.
* listobject.c nevertheless did not conform to the other invariants,
either; fixed.
* listobject.c now uses list_clear() as the obvious internal way to clear
a list, instead of abusing list_ass_slice() for that. It makes it easier
to enforce the invariant about ob_item == NULL.
* listsort() sets allocated to -1 during sort; any mutation will set it
to a value >= 0, so it is a safe way to detect mutation. A negative
value for allocated does not cause a problem elsewhere currently.
test_sort.py has a new test for this fix.
* listsort() leak: if items were added to the list during the sort, AND if
these items had a __del__ that puts still more stuff into the list,
then this more stuff (and the PyObject** array to hold them) were
overridden at the end of listsort() and never released.
mutation during list.sort() used to rely on that listobject.c always
NULL'ed ob_item when ob_size fell to 0. That's no longer true, so the
test for list mutation during a sort is no longer reliable. Changed the
test to rely instead on that listobject.c now never NULLs-out ob_item
after (if ever) ob_item gets a non-NULL value. This new assumption is
also documented now, as a required invariant in listobject.h.
The new assumption allowed some real simplification to some of the
hairier code in listsort(), so is a Good Thing on that count.
- Neatened the braces in PyList_New().
- Made sure "indexerr" was initialized to NULL.
- Factored if blocks in PyList_Append().
- Made sure "allocated" is initialized in list_init().
scheme in situations that likely won't benefit from it. This further
improves memory utilization from Py2.3 which always over-allocates
except for PyList_New().
Situations expected to benefit from over-allocation:
list.insert(), list.pop(), list.append(), and list.extend()
Situations deemed unlikely to benefit:
list_inplace_repeat, list_ass_slice, list_ass_subscript
The most gray area was for listextend_internal() which only runs
when the argument is a list or a tuple. This could be viewed as
a one-time fixed length addition or it could be viewed as wrapping
a series of appends. I left its over-allocation turned on but
could be convinced otherwise.
worth it to in-line the call to PyIter_Next().
Saves another 15% on most list operations that acceptable a general
iterable argument (such as the list constructor).
avoids creating an intermediate tuple for iterable arguments other than
lists or tuples.
In other words, a+=b no longer requires extra memory when b is not a
list or tuple. The list and tuple cases are unchanged.
for xrange and list objects).
* list.__reversed__ now checks the length of the sequence object before
calling PyList_GET_ITEM() because the mutable could have changed length.
* all three implementations are now tranparent with respect to length and
maintain the invariant len(it) == len(list(it)) even when the underlying
sequence mutates.
* __builtin__.reversed() now frees the underlying sequence as soon
as the iterator is exhausted.
* the code paths were rearranged so that the most common paths
do not require a jump.
* Replace sprintf message with a constant message string -- this error
message ran on every invocation except straight deletions but it was
only needed when the rhs was not iterable. The message was also
out-of-date and did not reflect that iterable arguments were allowed.
* For inner loops that do not make ref count adjustments, use memmove()
for fast copying and better readability.
* For inner loops that do make ref count adjustments, speed them up by
factoring out the constant structure reference and using vitem[] instead.
* Using addition instead of substraction on array indices allows the
compiler to use a fast addressing mode. Saves about 10%.
* Using PyTuple_GET_ITEM and PyList_SET_ITEM is about 7% faster than
PySequenceFast_GET_ITEM which has to make a list check on every pass.
utilization, and speed:
* Moved the responsibility for emptying the previous list from list_fill
to list_init.
* Replaced the code in list_extend with the superior code from list_fill.
* Eliminated list_fill.
Results:
* list.extend() no longer creates an intermediate tuple except to handle
the special case of x.extend(x). The saves memory and time.
* list.extend(x) runs
5 to 10% faster when x is a list or tuple
15% faster when x is an iterable not defining __len__
twice as fast when x is an iterable defining __len__
* the code is about 15 lines shorter and no longer duplicates
functionality.
The Py2.3 approach overallocated small lists by up to 8 elements.
The last checkin would limited this to one but slowed down (by 20 to 30%)
the creation of small lists between 3 to 8 elements.
This tune-up balances the two, limiting overallocation to 3 elements
(significantly reducing space consumption from Py2.3) and running faster
than the previous checkin.
The first part of the growth pattern (0, 4, 8, 16) neatly meshes with
allocators that trigger data movement only when crossing a power of two
boundary. Also, then even numbers mesh well with common data alignments.
realloc(). This is achieved by tracking the overallocation size in a new
field and using that information to skip calls to realloc() whenever
possible.
* Simplified and tightened the amount of overallocation. For larger lists,
this overallocates by 1/8th (compared to the previous scheme which ranged
between 1/4th to 1/32nd over-allocation). For smaller lists (n<6), the
maximum overallocation is one byte (formerly it could be upto eight bytes).
This saves memory in applications with large numbers of small lists.
* Eliminated the NRESIZE macro in favor of a new, static list_resize function
that encapsulates the resizing logic. Coverting this back to macro would
give a small (under 1%) speed-up. This was too small to warrant the loss
of readability, maintainability, and de-coupling.
* Some functions using NRESIZE had grown unnecessarily complex in their
efforts to bend to the macro's calling pattern. With the new list_resize
function in place, those other functions could be simplified. That is
being saved for a separate patch.
* The ob_item==NULL check could be eliminated from the new list_resize
function. This would entail finding each piece of code that sets ob_item
to NULL and adding a new line to invalidate the overallocation tracking
field. Rather than impose a new requirement on other pieces of list code,
it was preferred to leave the NULL check in place and retain the benefits
of decoupling, maintainability and information hiding (only PyList_New()
and list_sort() need to know about the new field). This approach also
reduces the odds of breaking an extension module.
(Collaborative effort by Raymond Hettinger, Hye-Shik Chang, Tim Peters,
and Armin Rigo.)
Raymond Hettinger
Andrew M. Kuchling
Neal Norwitz
Brett Cannon
Tim Peters
Armin Rigo
Nicholas Bastin