* set_merge() cannot assume that the table doesn't resize during iteration.

* convert some unnecessary tests to asserts -- they were necessary in
  dictobject.c because PyDict_Next() is a public function.  The same is
  not true for set_next().

* re-arrange the order of functions to more closely match the order
  in dictobject.c.  This makes it must easier to compare the two
  and ought to simplify any issues of maintaining both.

hotshot.stats.load.

* Use set_next() for looping in issubset() and frozenset_hash().
* Re-order the presentation of cmp and hash functions.

was never called during interpreter shutdown GC, so the f_back!=NULL
assertion was correct.  Now that generators get close()d during GC,
the assertion was being triggered because the generator close() was being
called as the top-level frame.  However, nothing actually is broken by
this; it's just that the condition was unexpected in previous Python
versions.

  a frozenset conversion when the initial search attempt fails with a
  TypeError and the key is some type of set.  Add a testcase.

* Eliminate a duplicate if-stmt.

  a frozenset conversion when the initial search attempt fails with a
  TypeError and the key is some type of set.  Add a testcase.

* Eliminate a duplicate if-stmt.

unicode instance if the argument is not an instance of basestring and
calling __str__ on the argument returns a unicode instance.

  s|=s, s&=s, s-=s, or s^=s).  Add related tests.

* Improve names for several variables and functions.

* Provide alternate table access functions (next, contains, add, and discard)
  that work with an entry argument instead of just a key.  This improves
  set-vs-set operations because we already have a hash value for each key
  and can avoid unnecessary calls to PyObject_Hash().  Provides a 5% to 20%
  speed-up for quick hashing elements like strings and integers.  Provides
  much more substantial improvements for slow hashing elements like tuples
  or objects defining a custom __hash__() function.

* Have difference operations resize() when 1/5 of the elements are dummies.
  Formerly, it was 1/6.  The new ratio triggers less frequently and only
  in cases that it can resize quicker and with greater benefit.  The right
  answer is probably either 1/4, 1/5, or 1/6.  Picked the middle value for
  an even trade-off between resize time and the space/time costs of dummy
  entries.

add st_gen, st_birthtime.

Backported to 2.4.

Backported to 2.4.

* Bring in free list from dictionary code.
* Improve several comments.
* Differencing can leave many dummy entries.  If more than
  1/6 are dummies, then resize them away.
* Factor-out common code with new macro, PyAnySet_CheckExact.

* Have issubset() control its own loop instead of using set_next_internal().

  has already done the job.
* Use a macro form of PyErr_Occurred() inside the set_lookkey() function.

* Give set_lookkey_string() a fast alternate path when no dummy entries
  are present.
* Have set_swap_bodies() reset the hash field to -1 whenever either of
  bodies is not a frozenset.  Maintains the invariant of regular sets
  always having -1 in the hash field; otherwise, any mutation would make
  the hash value invalid.
* Use an entry pointer to simplify the code in frozenset_hash().

  dictobject.c.
* Have frozenset_hash() use entry->hash instead of re-computing each
  individual hash with PyObject_Hash(o);
* Finalize the dummy entry before a system exit.

Unicode support.

'quotation mark', so I decided to use it instead of 'double-quote'.

manual.  Closes #1243553.