Mailman 3 December 2003 - Python-checkins

python/dist/src/Objects listobject.c,2.172,2.173
by jhylton＠users.sourceforge.net Dec. 26, 2003

Dec. 26, 2003

Update of /cvsroot/python/python/dist/src/Objects In directory sc8-pr-cvs1:/tmp/cvs-serv32521 Modified Files: listobject.c Log Message: Revert previous two checkins to repair test failure. The special-case code that was removed could return a value indicating success but leave an exception set. test_fileinput failed in a debug build as a result. Index: listobject.c =================================================================== RCS file: /cvsroot/python/python/dist/src/Objects/listobject.c,v retrieving revision 2.172 retrieving revision 2.173 diff -C2 -d -r2.172 -r2.173 *** listobject.c 26 Dec 2003 00:09:04 -0000 2.172 --- listobject.c 26 Dec 2003 19:05:04 -0000 2.173 *************** *** 2235,2245 **** } - /* empirically determined threshold for activating an optimisation - * in list_fill() - 100 seems close to optimum for current CPUs and - * compilers, as of December '03. - * see also comment in list_fill(). - */ - #define LISTFILL_OPT_THRESHOLD 100 - /* Adapted from newer code by Tim */ static int --- 2235,2238 ---- *************** *** 2256,2276 **** n = result->ob_size; ! /* Special-case list(a_list), for speed: ! * - if the source has 0 elements, there's nothing to copy. ! * - if the source has more than a threshold number of elements ! * slice assignment is a faster way of filling the target ! * (the exact threshold is subject to empirical determination). ! * Also special case any other zero length sequence, including ! * subclasses of list, there being nothing to copy. ! */ ! if (PyList_CheckExact(v)) { ! i = PyList_GET_SIZE(v); ! if (i == 0) ! return 0; ! if (i > LISTFILL_OPT_THRESHOLD) ! return list_ass_slice(result, 0, n, v); ! } else ! if (PySequence_Check(v) && PySequence_Size(v) == 0) ! return 0; /* Empty previous contents */ --- 2249,2258 ---- n = result->ob_size; ! /* Special-case list(a_list), for speed. */ ! if (PyList_Check(v)) { ! if (v == (PyObject *)result) ! return 0; /* source is destination, we're done */ ! return list_ass_slice(result, 0, n, v); ! } /* Empty previous contents */

1 0

python/dist/src/Include dictobject.h,2.31,2.32
by jhylton＠users.sourceforge.net Dec. 26, 2003

Dec. 26, 2003

Update of /cvsroot/python/python/dist/src/Include In directory sc8-pr-cvs1:/tmp/cvs-serv17110 Modified Files: dictobject.h Log Message: Fix name problem in previous checkin: Dict not List Index: dictobject.h =================================================================== RCS file: /cvsroot/python/python/dist/src/Include/dictobject.h,v retrieving revision 2.31 retrieving revision 2.32 diff -C2 -d -r2.31 -r2.32 *** dictobject.h 26 Dec 2003 00:20:53 -0000 2.31 --- dictobject.h 26 Dec 2003 17:17:49 -0000 2.32 *************** *** 88,92 **** #define PyDict_Check(op) PyObject_TypeCheck(op, &PyDict_Type) ! #define PyList_CheckExact(op) ((op)->ob_type == &PyDict_Type) PyAPI_FUNC(PyObject *) PyDict_New(void); --- 88,92 ---- #define PyDict_Check(op) PyObject_TypeCheck(op, &PyDict_Type) ! #define PyDict_CheckExact(op) ((op)->ob_type == &PyDict_Type) PyAPI_FUNC(PyObject *) PyDict_New(void);

1 0

python/dist/src/Include dictobject.h,2.29,2.30
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Include In directory sc8-pr-cvs1:/tmp/cvs-serv23304 Modified Files: dictobject.h Log Message: reverting 2.29: the patch was Ok, but the commit msg wrong Index: dictobject.h =================================================================== RCS file: /cvsroot/python/python/dist/src/Include/dictobject.h,v retrieving revision 2.29 retrieving revision 2.30 diff -C2 -d -r2.29 -r2.30 *** dictobject.h 26 Dec 2003 00:02:23 -0000 2.29 --- dictobject.h 26 Dec 2003 00:19:28 -0000 2.30 *************** *** 88,92 **** #define PyDict_Check(op) PyObject_TypeCheck(op, &PyDict_Type) - #define PyList_CheckExact(op) ((op)->ob_type == &PyDict_Type) PyAPI_FUNC(PyObject *) PyDict_New(void); --- 88,91 ----

2 1

python/dist/src/Include dictobject.h,2.30,2.31
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Include In directory sc8-pr-cvs1:/tmp/cvs-serv23724 Modified Files: dictobject.h Log Message: At 2.2, the Py<type>_Check() family of API functions (macros) changed semantics to include subtypes. Most concrete object APIs then had a Py<type>_CheckExact() macro added to test for an object's type not including subtypes. The PyDict_CheckExact() macro wasn't created at that time, so I've added it for API completeness/symmetry - even though nobody has complained about its absence in the time since 2.2 was released. Not a backport candidate. Index: dictobject.h =================================================================== RCS file: /cvsroot/python/python/dist/src/Include/dictobject.h,v retrieving revision 2.30 retrieving revision 2.31 diff -C2 -d -r2.30 -r2.31 *** dictobject.h 26 Dec 2003 00:19:28 -0000 2.30 --- dictobject.h 26 Dec 2003 00:20:53 -0000 2.31 *************** *** 88,91 **** --- 88,92 ---- #define PyDict_Check(op) PyObject_TypeCheck(op, &PyDict_Type) + #define PyList_CheckExact(op) ((op)->ob_type == &PyDict_Type) PyAPI_FUNC(PyObject *) PyDict_New(void);

1 0

python/dist/src/Objects listobject.c,2.171,2.172
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Objects In directory sc8-pr-cvs1:/tmp/cvs-serv21959 Modified Files: listobject.c Log Message: use the correct macro to access list size Index: listobject.c =================================================================== RCS file: /cvsroot/python/python/dist/src/Objects/listobject.c,v retrieving revision 2.171 retrieving revision 2.172 diff -C2 -d -r2.171 -r2.172 *** listobject.c 25 Dec 2003 13:27:20 -0000 2.171 --- listobject.c 26 Dec 2003 00:09:04 -0000 2.172 *************** *** 2265,2269 **** */ if (PyList_CheckExact(v)) { ! i = ((PyListObject*)v)->ob_size; if (i == 0) return 0; --- 2265,2269 ---- */ if (PyList_CheckExact(v)) { ! i = PyList_GET_SIZE(v); if (i == 0) return 0;

1 0

python/dist/src/Doc/api concrete.tex,1.35,1.36
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Doc/api In directory sc8-pr-cvs1:/tmp/cvs-serv21831 Modified Files: concrete.tex Log Message: At 2.2, the Py<type>_Check() family of API functions (macros) changed semantics to include subtypes. Most concrete object APIs then had a Py<type>_CheckExact() macro added to test for an object's type not including subtypes. The PyDict_CheckExact() macro wasn't created at that time, so I've added it for API completeness/symmetry - even though nobody has complained about its absence in the time since 2.2 was released. Not a backport candidate. Index: concrete.tex =================================================================== RCS file: /cvsroot/python/python/dist/src/Doc/api/concrete.tex,v retrieving revision 1.35 retrieving revision 1.36 diff -C2 -d -r1.35 -r1.36 *** concrete.tex 25 Dec 2003 23:57:52 -0000 1.35 --- concrete.tex 26 Dec 2003 00:07:51 -0000 1.36 *************** *** 1804,1807 **** --- 1804,1813 ---- \end{cfuncdesc} + \begin{cfuncdesc}{int}{PyDict_CheckExact}{PyObject *p} + Return true if \var{p} is a dict object, but not an instance of a + subtype of the dict type. + \versionadded{2.4} + \end{cfuncdesc} + \begin{cfuncdesc}{PyObject*}{PyDict_New}{} Returns a new empty dictionary, or \NULL{} on failure.

1 0

python/dist/src/Include dictobject.h,2.28,2.29
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Include In directory sc8-pr-cvs1:/tmp/cvs-serv21089 Modified Files: dictobject.h Log Message: The semantics of PyList_Check() and PyDict_Check() changed at 2.2, along with most other concrete object checks, but the docs weren't brought into line. PyList_CheckExact() was added at 2.2 but never documented. backport candidate. Index: dictobject.h =================================================================== RCS file: /cvsroot/python/python/dist/src/Include/dictobject.h,v retrieving revision 2.28 retrieving revision 2.29 diff -C2 -d -r2.28 -r2.29 *** dictobject.h 25 Nov 2003 21:12:14 -0000 2.28 --- dictobject.h 26 Dec 2003 00:02:23 -0000 2.29 *************** *** 88,91 **** --- 88,92 ---- #define PyDict_Check(op) PyObject_TypeCheck(op, &PyDict_Type) + #define PyList_CheckExact(op) ((op)->ob_type == &PyDict_Type) PyAPI_FUNC(PyObject *) PyDict_New(void);

1 0

python/dist/src/Doc/api concrete.tex, 1.6.6.6, 1.6.6.7
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Doc/api In directory sc8-pr-cvs1:/tmp/cvs-serv20589 Modified Files: Tag: release22-maint concrete.tex Log Message: backport of concrete.tex v1.35: The semantics of PyList_Check() and PyDict_Check() changed at 2.2, along with most other concrete object checks, but the docs weren't brought into line. PyList_CheckExact() was added at 2.2 but never documented. Index: concrete.tex =================================================================== RCS file: /cvsroot/python/python/dist/src/Doc/api/concrete.tex,v retrieving revision 1.6.6.6 retrieving revision 1.6.6.7 diff -C2 -d -r1.6.6.6 -r1.6.6.7 *** concrete.tex 22 Oct 2002 20:21:06 -0000 1.6.6.6 --- concrete.tex 26 Dec 2003 00:00:29 -0000 1.6.6.7 *************** *** 1603,1607 **** \begin{cfuncdesc}{int}{PyList_Check}{PyObject *p} ! Returns true if its argument is a \ctype{PyListObject}. \end{cfuncdesc} --- 1603,1615 ---- \begin{cfuncdesc}{int}{PyList_Check}{PyObject *p} ! Returns true if \var{p} is a list object or an instance of a ! subtype of the list type. ! \versionchanged[Allowed subtypes to be accepted]{2.2} ! \end{cfuncdesc} ! ! \begin{cfuncdesc}{int}{PyList_CheckExact}{PyObject *p} ! Return true if \var{p} is a list object, but not an instance of a ! subtype of the list type. ! \versionadded{2.2} \end{cfuncdesc} *************** *** 1721,1725 **** \begin{cfuncdesc}{int}{PyDict_Check}{PyObject *p} ! Returns true if its argument is a \ctype{PyDictObject}. \end{cfuncdesc} --- 1729,1735 ---- \begin{cfuncdesc}{int}{PyDict_Check}{PyObject *p} ! Returns true if \var{p} is a dict object or an instance of a ! subtype of the dict type. ! \versionchanged[Allowed subtypes to be accepted]{2.2} \end{cfuncdesc}

1 0

python/dist/src/Doc/api concrete.tex,1.34,1.35
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Doc/api In directory sc8-pr-cvs1:/tmp/cvs-serv20083 Modified Files: concrete.tex Log Message: The semantics of PyList_Check() and PyDict_Check() changed at 2.2, along with most other concrete object checks, but the docs weren't brought into line. PyList_CheckExact() was added at 2.2 but never documented. backport candidate. Index: concrete.tex =================================================================== RCS file: /cvsroot/python/python/dist/src/Doc/api/concrete.tex,v retrieving revision 1.34 retrieving revision 1.35 diff -C2 -d -r1.34 -r1.35 *** concrete.tex 13 Dec 2003 19:48:41 -0000 1.34 --- concrete.tex 25 Dec 2003 23:57:52 -0000 1.35 *************** *** 1671,1675 **** \begin{cfuncdesc}{int}{PyList_Check}{PyObject *p} ! Returns true if its argument is a \ctype{PyListObject}. \end{cfuncdesc} --- 1671,1683 ---- \begin{cfuncdesc}{int}{PyList_Check}{PyObject *p} ! Returns true if \var{p} is a list object or an instance of a ! subtype of the list type. ! \versionchanged[Allowed subtypes to be accepted]{2.2} ! \end{cfuncdesc} ! ! \begin{cfuncdesc}{int}{PyList_CheckExact}{PyObject *p} ! Return true if \var{p} is a list object, but not an instance of a ! subtype of the list type. ! \versionadded{2.2} \end{cfuncdesc} *************** *** 1791,1795 **** \begin{cfuncdesc}{int}{PyDict_Check}{PyObject *p} ! Returns true if its argument is a \ctype{PyDictObject}. \end{cfuncdesc} --- 1799,1805 ---- \begin{cfuncdesc}{int}{PyDict_Check}{PyObject *p} ! Returns true if \var{p} is a dict object or an instance of a ! subtype of the dict type. ! \versionchanged[Allowed subtypes to be accepted]{2.2} \end{cfuncdesc}

1 0

python/dist/src/Misc ACKS, 1.254, 1.255 NEWS, 1.915, 1.916
by aimacintyre＠users.sourceforge.net Dec. 25, 2003

Dec. 25, 2003

Update of /cvsroot/python/python/dist/src/Misc In directory sc8-pr-cvs1:/tmp/cvs-serv6382 Modified Files: ACKS NEWS Log Message: Performance of list([]) in 2.3 came up in a thread on comp.lang.python, which can be reviewed via http://coding.derkeiler.com/Archive/Python/comp.lang.python/2003-12/1011.ht… Duncan Booth investigated, and discovered that an "optimisation" was in fact a pessimisation for small numbers of elements in a source list, compared to not having the optimisation, although with large numbers of elements in the source list the optimisation was quite beneficial. He posted his change to comp.lang.python (but not to SF). Further research has confirmed his assessment that the optimisation only becomes a net win when the source list has more than 100 elements. I also found that the optimisation could apply to tuples as well, but the gains only arrive with source tuples larger than about 320 elements and are nowhere near as significant as the gains with lists, (~95% gain @ 10000 elements for lists, ~20% gain @ 10000 elements for tuples) so I haven't proceeded with this. The code as it was applied the optimisation to list subclasses as well, and this also appears to be a net loss for all reasonable sized sources (~80-100% for up to 100 elements, ~20% for more than 500 elements; I tested up to 10000 elements). Duncan also suggested special casing empty lists, which I've extended to all empty sequences. On the basis that list_fill() is only ever called with a list for the result argument, testing for the source being the destination has now happens before testing source types. Index: ACKS =================================================================== RCS file: /cvsroot/python/python/dist/src/Misc/ACKS,v retrieving revision 1.254 retrieving revision 1.255 diff -C2 -d -r1.254 -r1.255 *** ACKS 10 Nov 2003 06:44:44 -0000 1.254 --- ACKS 25 Dec 2003 13:28:48 -0000 1.255 *************** *** 63,66 **** --- 63,67 ---- Matthew Boedicker David Bolen + Duncan Booth Jurjen Bos Peter Bosch Index: NEWS =================================================================== RCS file: /cvsroot/python/python/dist/src/Misc/NEWS,v retrieving revision 1.915 retrieving revision 1.916 diff -C2 -d -r1.915 -r1.916 *** NEWS 17 Dec 2003 20:43:32 -0000 1.915 --- NEWS 25 Dec 2003 13:28:48 -0000 1.916 *************** *** 119,122 **** --- 119,130 ---- working towards the beginning. See SF feature request 801847. + - in a thread on comp.lang.python, several people noted that list() + was much slower than in 2.1 and earlier versions of Python, when used + to create new lists from existing lists. Duncan Booth did some + research that uncovered an optimisation that, for lists below + about 100 elements, was actually slower than the normal case. The + special case criteria have been tightened to rectify the performance + regression. + Extension modules -----------------

1 0