Author: mark.dickinson Date: Sun Sep 27 18:05:21 2009 New Revision: 75084 Log: Issue #6713: Improve decimal int -> string conversions. Thanks Gawain Bolton for the suggestion and original patches. Modified: python/trunk/Misc/NEWS python/trunk/Objects/intobject.c Modified: python/trunk/Misc/NEWS ============================================================================== --- python/trunk/Misc/NEWS (original) +++ python/trunk/Misc/NEWS Sun Sep 27 18:05:21 2009 @@ -18,7 +18,8 @@ - Issue #6922: Fix an infinite loop when trying to decode an invalid UTF-32 stream with a non-raising error handler like "replace" or "ignore". -- Issue #6713: Improve performance of integer -> string conversions. +- Issue #6713: Improve performance of base 10 int -> string and + long -> string conversions. - Issue #1590864: Fix potential deadlock when mixing threads and fork(). Modified: python/trunk/Objects/intobject.c ============================================================================== --- python/trunk/Objects/intobject.c (original) +++ python/trunk/Objects/intobject.c Sun Sep 27 18:05:21 2009 @@ -436,12 +436,6 @@ return 0; } -static PyObject * -int_repr(PyIntObject *v) -{ - return _PyInt_Format(v, 10, 0); -} - static int int_compare(PyIntObject *v, PyIntObject *w) { @@ -1113,6 +1107,26 @@ return PyInt_FromLong(1L); } +/* Convert an integer to a decimal string. On many platforms, this + will be significantly faster than the general arbitrary-base + conversion machinery in _PyInt_Format, thanks to optimization + opportunities offered by division by a compile-time constant. */ +static PyObject * +int_to_decimal_string(PyIntObject *v) { + char buf[sizeof(long)*CHAR_BIT/3+6], *p, *bufend; + long n = v->ob_ival; + unsigned long absn; + p = bufend = buf + sizeof(buf); + absn = n < 0 ? -(unsigned long)n : n; + do { + *--p = '0' + absn % 10; + absn /= 10; + } while (absn); + if (n < 0) + *--p = '-'; + return PyString_FromStringAndSize(p, bufend - p); +} + /* Convert an integer to the given base. Returns a string. If base is 2, 8 or 16, add the proper prefix '0b', '0o' or '0x'. If newstyle is zero, then use the pre-2.6 behavior of octal having @@ -1137,6 +1151,10 @@ assert(base >= 2 && base <= 36); + /* Special case base 10, for speed */ + if (base == 10) + return int_to_decimal_string(v); + do { /* I'd use i_divmod, except it doesn't produce the results I want when n is negative. So just duplicate the salient @@ -1169,7 +1187,7 @@ *--p = 'x'; *--p = '0'; } - else if (base != 10) { + else { *--p = '#'; *--p = '0' + base%10; if (base > 10) @@ -1341,13 +1359,13 @@ 0, /* tp_getattr */ 0, /* tp_setattr */ (cmpfunc)int_compare, /* tp_compare */ - (reprfunc)int_repr, /* tp_repr */ + (reprfunc)int_to_decimal_string, /* tp_repr */ &int_as_number, /* tp_as_number */ 0, /* tp_as_sequence */ 0, /* tp_as_mapping */ (hashfunc)int_hash, /* tp_hash */ - 0, /* tp_call */ - (reprfunc)int_repr, /* tp_str */ + 0, /* tp_call */ + (reprfunc)int_to_decimal_string, /* tp_str */ PyObject_GenericGetAttr, /* tp_getattro */ 0, /* tp_setattro */ 0, /* tp_as_buffer */