SF bug #681003: itertools issues

* Fixed typo in exception message for times()
* Filled in missing times_traverse()
* Document reasons that imap() did not adopt a None fill-in feature
* Document that count(sys.maxint) will wrap-around on overflow
* Add overflow test to islice()
* Check that starmap()'s argument returns a tuple
* Verify that imap()'s tuple re-use is safe
* Make a similar tuple re-use (with safety check) for izip()
This commit is contained in:
Raymond Hettinger 2003-02-07 05:32:58 +00:00
parent 2b09bc4d57
commit 2012f174ea
3 changed files with 91 additions and 6 deletions

View File

@ -82,6 +82,10 @@ by functions or loops that truncate the stream.
yield cnt
cnt += 1
\end{verbatim}
Note, \function{count()} does not check for overflow and will return
negative numbers after exceeding \code{sys.maxint}. This behavior
may change in the future.
\end{funcdesc}
\begin{funcdesc}{dropwhile}{predicate, iterable}

View File

@ -1,6 +1,7 @@
import unittest
from test import test_support
from itertools import *
import sys
class TestBasicOps(unittest.TestCase):
def test_count(self):
@ -47,6 +48,7 @@ class TestBasicOps(unittest.TestCase):
import operator
self.assertEqual(list(starmap(operator.pow, zip(range(3), range(1,7)))),
[0**1, 1**2, 2**3])
self.assertRaises(TypeError, list, starmap(operator.pow, [[4,5]]))
def test_islice(self):
for args in [ # islice(args) should agree with range(args)
@ -71,6 +73,7 @@ class TestBasicOps(unittest.TestCase):
self.assertRaises(ValueError, islice, xrange(10), 1, -5, -1)
self.assertRaises(ValueError, islice, xrange(10), 1, 10, -1)
self.assertRaises(ValueError, islice, xrange(10), 1, 10, 0)
self.assertEqual(len(list(islice(count(), 1, 10, sys.maxint))), 1)
def test_takewhile(self):
data = [1, 3, 5, 20, 2, 4, 6, 8]

View File

@ -403,6 +403,7 @@ static PyObject *
islice_next(isliceobject *lz)
{
PyObject *item;
long oldnext;
while (lz->cnt < lz->next) {
item = PyIter_Next(lz->it);
@ -417,7 +418,10 @@ islice_next(isliceobject *lz)
if (item == NULL)
return NULL;
lz->cnt++;
oldnext = lz->next;
lz->next += lz->step;
if (lz->next < oldnext) /* Check for overflow */
lz->next = lz->stop;
return item;
}
@ -558,6 +562,12 @@ starmap_next(starmapobject *lz)
args = PyIter_Next(lz->it);
if (args == NULL)
return NULL;
if (!PyTuple_CheckExact(args)) {
Py_DECREF(args);
PyErr_SetString(PyExc_TypeError,
"iterator must return a tuple");
return NULL;
}
result = PyObject_Call(lz->func, args, NULL);
Py_DECREF(args);
return result;
@ -719,6 +729,31 @@ imap_traverse(imapobject *lz, visitproc visit, void *arg)
return 0;
}
/*
imap() is an iterator version of __builtins__.map() except that it does
not have the None fill-in feature. That was intentionally left out for
the following reasons:
1) Itertools are designed to be easily combined and chained together.
Having all tools stop with the shortest input is a unifying principle
that makes it easier to combine finite iterators (supplying data) with
infinite iterators like count() and repeat() (for supplying sequential
or constant arguments to a function).
2) In typical use cases for combining itertools, having one finite data
supplier run out before another is likely to be an error condition which
should not pass silently by automatically supplying None.
3) The use cases for automatic None fill-in are rare -- not many functions
do something useful when a parameter suddenly switches type and becomes
None.
4) If a need does arise, it can be met by __builtins__.map() or by
writing a generator.
5) Similar toolsets in Haskell and SML do not have automatic None fill-in.
*/
static PyObject *
imap_next(imapobject *lz)
{
@ -742,6 +777,11 @@ imap_next(imapobject *lz)
}
return argtuple;
} else {
if (argtuple->ob_refcnt > 1) {
argtuple = PyTuple_New(numargs);
if (argtuple == NULL)
return NULL;
}
for (i=0 ; i<numargs ; i++) {
val = PyIter_Next(PyTuple_GET_ITEM(lz->iters, i));
if (val == NULL)
@ -837,7 +877,7 @@ times_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
if (cnt < 0) {
PyErr_SetString(PyExc_ValueError,
"count for imap() cannot be negative.");
"count for times() cannot be negative.");
return NULL;
}
@ -860,6 +900,14 @@ times_dealloc(timesobject *lz)
lz->ob_type->tp_free(lz);
}
static int
times_traverse(timesobject *lz, visitproc visit, void *arg)
{
if (lz->obj)
return visit(lz->obj, arg);
return 0;
}
static PyObject *
times_next(timesobject *lz)
{
@ -911,7 +959,7 @@ PyTypeObject times_type = {
Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC |
Py_TPFLAGS_BASETYPE, /* tp_flags */
times_doc, /* tp_doc */
0, /* tp_traverse */
(traverseproc)times_traverse, /* tp_traverse */
0, /* tp_clear */
0, /* tp_richcompare */
0, /* tp_weaklistoffset */
@ -1191,6 +1239,7 @@ typedef struct {
PyObject_HEAD
long tuplesize;
PyObject *ittuple; /* tuple of iterators */
PyObject *result;
} izipobject;
PyTypeObject izip_type;
@ -1201,6 +1250,7 @@ izip_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
izipobject *lz;
int i;
PyObject *ittuple; /* tuple of iterators */
PyObject *result;
int tuplesize = PySequence_Length(args);
if (tuplesize < 1) {
@ -1230,14 +1280,27 @@ izip_new(PyTypeObject *type, PyObject *args, PyObject *kwds)
PyTuple_SET_ITEM(ittuple, i, it);
}
/* create a result holder */
result = PyTuple_New(tuplesize);
if (result == NULL) {
Py_DECREF(ittuple);
return NULL;
}
for (i=0 ; i < tuplesize ; i++) {
Py_INCREF(Py_None);
PyTuple_SET_ITEM(result, i, Py_None);
}
/* create izipobject structure */
lz = (izipobject *)type->tp_alloc(type, 0);
if (lz == NULL) {
Py_DECREF(ittuple);
Py_DECREF(result);
return NULL;
}
lz->ittuple = ittuple;
lz->tuplesize = tuplesize;
lz->result = result;
return (PyObject *)lz;
}
@ -1247,6 +1310,7 @@ izip_dealloc(izipobject *lz)
{
PyObject_GC_UnTrack(lz);
Py_XDECREF(lz->ittuple);
Py_XDECREF(lz->result);
lz->ob_type->tp_free(lz);
}
@ -1263,13 +1327,27 @@ izip_next(izipobject *lz)
{
int i;
long tuplesize = lz->tuplesize;
PyObject *result;
PyObject *result = lz->result;
PyObject *it;
PyObject *item;
result = PyTuple_New(tuplesize);
if (result == NULL)
return NULL;
assert(result->ob_refcnt >= 1);
if (result->ob_refcnt == 1) {
for (i=0 ; i < tuplesize ; i++) {
Py_DECREF(PyTuple_GET_ITEM(result, i));
PyTuple_SET_ITEM(result, i, NULL);
}
Py_INCREF(result);
} else {
Py_DECREF(result);
result = PyTuple_New(tuplesize);
if (result == NULL)
return NULL;
Py_INCREF(result);
lz->result = result;
}
assert(lz->result == result);
assert(result->ob_refcnt == 2);
for (i=0 ; i < tuplesize ; i++) {
it = PyTuple_GET_ITEM(lz->ittuple, i);