Commit b08f8e50 authored by Jim Fulton's avatar Jim Fulton

Updated indentation.

parent ec66d629
......@@ -10,7 +10,7 @@
WARRANTIES OF TITLE, MERCHANTABILITY, AGAINST INFRINGEMENT, AND FITNESS
FOR A PARTICULAR PURPOSE
****************************************************************************/
****************************************************************************/
#define BUCKETTEMPLATE_C "$Id$\n"
......@@ -38,15 +38,15 @@
int _i; \
int _cmp = 1; \
for (_i = _hi >> 1; _lo < _hi; _i = (_lo + _hi) >> 1) { \
TEST_KEY_SET_OR(_cmp, (SELF)->keys[_i], (KEY)) \
ONERROR; \
if (_cmp < 0) _lo = _i + 1; \
else if (_cmp == 0) break; \
else _hi = _i; \
TEST_KEY_SET_OR(_cmp, (SELF)->keys[_i], (KEY)) \
ONERROR; \
if (_cmp < 0) _lo = _i + 1; \
else if (_cmp == 0) break; \
else _hi = _i; \
} \
(INDEX) = _i; \
(ABSENT) = _cmp; \
}
}
/*
** _bucket_get
......@@ -75,30 +75,30 @@
static PyObject *
_bucket_get(Bucket *self, PyObject *keyarg, int has_key)
{
int i, cmp;
KEY_TYPE key;
PyObject *r = NULL;
int copied = 1;
COPY_KEY_FROM_ARG(key, keyarg, copied);
UNLESS (copied) return NULL;
UNLESS (PER_USE(self)) return NULL;
BUCKET_SEARCH(i, cmp, self, key, goto Done);
if (has_key)
r = PyInt_FromLong(cmp ? 0 : has_key);
else {
if (cmp == 0) {
COPY_VALUE_TO_OBJECT(r, self->values[i]);
}
else
PyErr_SetObject(PyExc_KeyError, keyarg);
int i, cmp;
KEY_TYPE key;
PyObject *r = NULL;
int copied = 1;
COPY_KEY_FROM_ARG(key, keyarg, copied);
UNLESS (copied) return NULL;
UNLESS (PER_USE(self)) return NULL;
BUCKET_SEARCH(i, cmp, self, key, goto Done);
if (has_key)
r = PyInt_FromLong(cmp ? 0 : has_key);
else {
if (cmp == 0) {
COPY_VALUE_TO_OBJECT(r, self->values[i]);
}
else
PyErr_SetObject(PyExc_KeyError, keyarg);
}
Done:
PER_UNUSE(self);
return r;
Done:
PER_UNUSE(self);
return r;
}
......@@ -126,45 +126,45 @@ bucket_getitem(Bucket *self, PyObject *key)
static int
Bucket_grow(Bucket *self, int newsize, int noval)
{
KEY_TYPE *keys;
VALUE_TYPE *values;
if (self->size) {
if (newsize < 0)
newsize = self->size * 2;
if (newsize < 0) /* int overflow */
goto Overflow;
UNLESS (keys = BTree_Realloc(self->keys, sizeof(KEY_TYPE) * newsize))
return -1;
UNLESS (noval) {
values = BTree_Realloc(self->values, sizeof(VALUE_TYPE) * newsize);
if (values == NULL) {
free(keys);
return -1;
}
self->values = values;
}
self->keys = keys;
KEY_TYPE *keys;
VALUE_TYPE *values;
if (self->size) {
if (newsize < 0)
newsize = self->size * 2;
if (newsize < 0) /* int overflow */
goto Overflow;
UNLESS (keys = BTree_Realloc(self->keys, sizeof(KEY_TYPE) * newsize))
return -1;
UNLESS (noval) {
values = BTree_Realloc(self->values, sizeof(VALUE_TYPE) * newsize);
if (values == NULL) {
free(keys);
return -1;
}
self->values = values;
}
else {
if (newsize < 0)
newsize = MIN_BUCKET_ALLOC;
UNLESS (self->keys = BTree_Malloc(sizeof(KEY_TYPE) * newsize))
return -1;
UNLESS (noval) {
self->values = BTree_Malloc(sizeof(VALUE_TYPE) * newsize);
if (self->values == NULL) {
free(self->keys);
self->keys = NULL;
return -1;
}
}
self->keys = keys;
}
else {
if (newsize < 0)
newsize = MIN_BUCKET_ALLOC;
UNLESS (self->keys = BTree_Malloc(sizeof(KEY_TYPE) * newsize))
return -1;
UNLESS (noval) {
self->values = BTree_Malloc(sizeof(VALUE_TYPE) * newsize);
if (self->values == NULL) {
free(self->keys);
self->keys = NULL;
return -1;
}
}
self->size = newsize;
return 0;
}
self->size = newsize;
return 0;
Overflow:
Overflow:
PyErr_NoMemory();
return -1;
}
......@@ -213,54 +213,54 @@ static int
bucket_append(Bucket *self, Bucket *from, int i, int n,
int copyValues, int overallocate)
{
int newlen;
assert(self && from && self != from);
assert(i >= 0);
assert(n > 0);
assert(i+n <= from->len);
/* Make room. */
newlen = self->len + n;
if (newlen > self->size) {
int newsize = newlen;
if (overallocate) /* boost by 25% -- pretty arbitrary */
newsize += newsize >> 2;
if (Bucket_grow(self, newsize, ! copyValues) < 0)
return -1;
}
assert(newlen <= self->size);
/* Copy stuff. */
memcpy(self->keys + self->len, from->keys + i, n * sizeof(KEY_TYPE));
if (copyValues) {
assert(self->values);
assert(from->values);
memcpy(self->values + self->len, from->values + i,
n * sizeof(VALUE_TYPE));
}
self->len = newlen;
int newlen;
assert(self && from && self != from);
assert(i >= 0);
assert(n > 0);
assert(i+n <= from->len);
/* Make room. */
newlen = self->len + n;
if (newlen > self->size) {
int newsize = newlen;
if (overallocate) /* boost by 25% -- pretty arbitrary */
newsize += newsize >> 2;
if (Bucket_grow(self, newsize, ! copyValues) < 0)
return -1;
}
assert(newlen <= self->size);
/* Copy stuff. */
memcpy(self->keys + self->len, from->keys + i, n * sizeof(KEY_TYPE));
if (copyValues) {
assert(self->values);
assert(from->values);
memcpy(self->values + self->len, from->values + i,
n * sizeof(VALUE_TYPE));
}
self->len = newlen;
/* Bump refcounts. */
/* Bump refcounts. */
#ifdef KEY_TYPE_IS_PYOBJECT
{
int j;
PyObject **p = from->keys + i;
for (j = 0; j < n; ++j, ++p) {
Py_INCREF(*p);
}
{
int j;
PyObject **p = from->keys + i;
for (j = 0; j < n; ++j, ++p) {
Py_INCREF(*p);
}
}
#endif
#ifdef VALUE_TYPE_IS_PYOBJECT
if (copyValues) {
int j;
PyObject **p = from->values + i;
for (j = 0; j < n; ++j, ++p) {
Py_INCREF(*p);
}
if (copyValues) {
int j;
PyObject **p = from->values + i;
for (j = 0; j < n; ++j, ++p) {
Py_INCREF(*p);
}
}
#endif
return 0;
return 0;
}
#endif /* MULTI_INT_UNION */
......@@ -291,136 +291,136 @@ static int
_bucket_set(Bucket *self, PyObject *keyarg, PyObject *v,
int unique, int noval, int *changed)
{
int i, cmp;
KEY_TYPE key;
/* Subtle: there may or may not be a value. If there is, we need to
* check its type early, so that in case of error we can get out before
* mutating the bucket. But because value isn't used on all paths, if
* we don't initialize value then gcc gives a nuisance complaint that
* value may be used initialized (it can't be, but gcc doesn't know
* that). So we initialize it. However, VALUE_TYPE can be various types,
* including int, PyObject*, and char[6], so it's a puzzle to spell
* initialization. It so happens that {0} is a valid initializer for all
* these types.
*/
VALUE_TYPE value = {0}; /* squash nuisance warning */
int result = -1; /* until proven innocent */
int copied = 1;
COPY_KEY_FROM_ARG(key, keyarg, copied);
int i, cmp;
KEY_TYPE key;
/* Subtle: there may or may not be a value. If there is, we need to
* check its type early, so that in case of error we can get out before
* mutating the bucket. But because value isn't used on all paths, if
* we don't initialize value then gcc gives a nuisance complaint that
* value may be used initialized (it can't be, but gcc doesn't know
* that). So we initialize it. However, VALUE_TYPE can be various types,
* including int, PyObject*, and char[6], so it's a puzzle to spell
* initialization. It so happens that {0} is a valid initializer for all
* these types.
*/
VALUE_TYPE value = {0}; /* squash nuisance warning */
int result = -1; /* until proven innocent */
int copied = 1;
COPY_KEY_FROM_ARG(key, keyarg, copied);
UNLESS(copied) return -1;
/* Copy the value early (if needed), so that in case of error a
* pile of bucket mutations don't need to be undone.
*/
if (v && !noval) {
COPY_VALUE_FROM_ARG(value, v, copied);
UNLESS(copied) return -1;
}
/* Copy the value early (if needed), so that in case of error a
* pile of bucket mutations don't need to be undone.
*/
if (v && !noval) {
COPY_VALUE_FROM_ARG(value, v, copied);
UNLESS(copied) return -1;
}
UNLESS (PER_USE(self)) return -1;
UNLESS (PER_USE(self)) return -1;
BUCKET_SEARCH(i, cmp, self, key, goto Done);
if (cmp == 0) {
/* The key exists, at index i. */
BUCKET_SEARCH(i, cmp, self, key, goto Done);
if (cmp == 0) {
/* The key exists, at index i. */
if (v) {
/* The key exists at index i, and there's a new value.
* If unique, we're not supposed to replace it. If noval, or this
* is a set bucket (self->values is NULL), there's nothing to do.
*/
if (unique || noval || self->values == NULL) {
result = 0;
goto Done;
}
/* The key exists at index i, and we need to replace the value. */
if (v) {
/* The key exists at index i, and there's a new value.
* If unique, we're not supposed to replace it. If noval, or this
* is a set bucket (self->values is NULL), there's nothing to do.
*/
if (unique || noval || self->values == NULL) {
result = 0;
goto Done;
}
/* The key exists at index i, and we need to replace the value. */
#ifdef VALUE_SAME
/* short-circuit if no change */
if (VALUE_SAME(self->values[i], value)) {
result = 0;
goto Done;
}
/* short-circuit if no change */
if (VALUE_SAME(self->values[i], value)) {
result = 0;
goto Done;
}
#endif
if (changed)
*changed = 1;
DECREF_VALUE(self->values[i]);
COPY_VALUE(self->values[i], value);
INCREF_VALUE(self->values[i]);
if (PER_CHANGED(self) >= 0)
result = 0;
goto Done;
}
/* The key exists at index i, and should be deleted. */
DECREF_KEY(self->keys[i]);
self->len--;
if (i < self->len)
memmove(self->keys + i, self->keys + i+1,
sizeof(KEY_TYPE)*(self->len - i));
if (self->values) {
DECREF_VALUE(self->values[i]);
if (i < self->len)
memmove(self->values + i, self->values + i+1,
sizeof(VALUE_TYPE)*(self->len - i));
}
if (changed)
*changed = 1;
DECREF_VALUE(self->values[i]);
COPY_VALUE(self->values[i], value);
INCREF_VALUE(self->values[i]);
if (PER_CHANGED(self) >= 0)
result = 0;
goto Done;
}
if (! self->len) {
self->size = 0;
free(self->keys);
self->keys = NULL;
if (self->values) {
free(self->values);
self->values = NULL;
}
}
/* The key exists at index i, and should be deleted. */
DECREF_KEY(self->keys[i]);
self->len--;
if (i < self->len)
memmove(self->keys + i, self->keys + i+1,
sizeof(KEY_TYPE)*(self->len - i));
if (changed)
*changed = 1;
if (PER_CHANGED(self) >= 0)
result = 1;
goto Done;
if (self->values) {
DECREF_VALUE(self->values[i]);
if (i < self->len)
memmove(self->values + i, self->values + i+1,
sizeof(VALUE_TYPE)*(self->len - i));
}
/* The key doesn't exist, and belongs at index i. */
if (!v) {
/* Can't delete a non-existent key. */
PyErr_SetObject(PyExc_KeyError, keyarg);
goto Done;
if (! self->len) {
self->size = 0;
free(self->keys);
self->keys = NULL;
if (self->values) {
free(self->values);
self->values = NULL;
}
}
/* The key doesn't exist and should be inserted at index i. */
if (self->len == self->size && Bucket_grow(self, -1, noval) < 0)
goto Done;
if (changed)
*changed = 1;
if (PER_CHANGED(self) >= 0)
result = 1;
goto Done;
}
if (self->len > i) {
memmove(self->keys + i + 1, self->keys + i,
sizeof(KEY_TYPE) * (self->len - i));
if (self->values) {
memmove(self->values + i + 1, self->values + i,
sizeof(VALUE_TYPE) * (self->len - i));
}
}
/* The key doesn't exist, and belongs at index i. */
if (!v) {
/* Can't delete a non-existent key. */
PyErr_SetObject(PyExc_KeyError, keyarg);
goto Done;
}
COPY_KEY(self->keys[i], key);
INCREF_KEY(self->keys[i]);
/* The key doesn't exist and should be inserted at index i. */
if (self->len == self->size && Bucket_grow(self, -1, noval) < 0)
goto Done;
if (! noval) {
COPY_VALUE(self->values[i], value);
INCREF_VALUE(self->values[i]);
if (self->len > i) {
memmove(self->keys + i + 1, self->keys + i,
sizeof(KEY_TYPE) * (self->len - i));
if (self->values) {
memmove(self->values + i + 1, self->values + i,
sizeof(VALUE_TYPE) * (self->len - i));
}
}
self->len++;
if (changed)
*changed = 1;
if (PER_CHANGED(self) >= 0)
result = 1;
COPY_KEY(self->keys[i], key);
INCREF_KEY(self->keys[i]);
if (! noval) {
COPY_VALUE(self->values[i], value);
INCREF_VALUE(self->values[i]);
}
Done:
PER_UNUSE(self);
return result;
self->len++;
if (changed)
*changed = 1;
if (PER_CHANGED(self) >= 0)
result = 1;
Done:
PER_UNUSE(self);
return result;
}
/*
......@@ -438,9 +438,9 @@ Done:
static int
bucket_setitem(Bucket *self, PyObject *key, PyObject *v)
{
if (_bucket_set(self, key, v, 0, 0, 0) < 0)
return -1;
return 0;
if (_bucket_set(self, key, v, 0, 0, 0) < 0)
return -1;
return 0;
}
/**
......@@ -450,71 +450,71 @@ bucket_setitem(Bucket *self, PyObject *key, PyObject *v)
static int
update_from_seq(PyObject *map, PyObject *seq)
{
PyObject *iter, *o, *k, *v;
int err = -1;
/* One path creates a new seq object. The other path has an
INCREF of the seq argument. So seq must always be DECREFed on
the way out.
*/
/* Use items() if it's not a sequence. Alas, PySequence_Check()
* returns true for a PeristentMapping or PersistentDict, and we
* want to use items() in those cases too.
*/
if (!PySequence_Check(seq) || /* or it "looks like a dict" */
PyObject_HasAttrString(seq, "iteritems")) {
PyObject *items;
items = PyObject_GetAttrString(seq, "items");
if (items == NULL)
PyObject *iter, *o, *k, *v;
int err = -1;
/* One path creates a new seq object. The other path has an
INCREF of the seq argument. So seq must always be DECREFed on
the way out.
*/
/* Use items() if it's not a sequence. Alas, PySequence_Check()
* returns true for a PeristentMapping or PersistentDict, and we
* want to use items() in those cases too.
*/
if (!PySequence_Check(seq) || /* or it "looks like a dict" */
PyObject_HasAttrString(seq, "iteritems")) {
PyObject *items;
items = PyObject_GetAttrString(seq, "items");
if (items == NULL)
return -1;
seq = PyObject_CallObject(items, NULL);
Py_DECREF(items);
if (seq == NULL)
seq = PyObject_CallObject(items, NULL);
Py_DECREF(items);
if (seq == NULL)
return -1;
}
else
Py_INCREF(seq);
iter = PyObject_GetIter(seq);
if (iter == NULL)
goto err;
while (1) {
o = PyIter_Next(iter);
if (o == NULL) {
}
else
Py_INCREF(seq);
iter = PyObject_GetIter(seq);
if (iter == NULL)
goto err;
while (1) {
o = PyIter_Next(iter);
if (o == NULL) {
if (PyErr_Occurred())
goto err;
goto err;
else
break;
}
if (!PyTuple_Check(o) || PyTuple_GET_SIZE(o) != 2) {
break;
}
if (!PyTuple_Check(o) || PyTuple_GET_SIZE(o) != 2) {
Py_DECREF(o);
PyErr_SetString(PyExc_TypeError,
"Sequence must contain 2-item tuples");
"Sequence must contain 2-item tuples");
goto err;
}
k = PyTuple_GET_ITEM(o, 0);
v = PyTuple_GET_ITEM(o, 1);
if (PyObject_SetItem(map, k, v) < 0) {
}
k = PyTuple_GET_ITEM(o, 0);
v = PyTuple_GET_ITEM(o, 1);
if (PyObject_SetItem(map, k, v) < 0) {
Py_DECREF(o);
goto err;
}
Py_DECREF(o);
}
Py_DECREF(o);
}
err = 0;
err = 0;
err:
Py_DECREF(iter);
Py_DECREF(seq);
return err;
Py_DECREF(iter);
Py_DECREF(seq);
return err;
}
static PyObject *
Mapping_update(PyObject *self, PyObject *seq)
{
if (update_from_seq(self, seq) < 0)
return NULL;
Py_INCREF(Py_None);
return Py_None;
if (update_from_seq(self, seq) < 0)
return NULL;
Py_INCREF(Py_None);
return Py_None;
}
/*
......@@ -532,42 +532,42 @@ Mapping_update(PyObject *self, PyObject *seq)
static int
bucket_split(Bucket *self, int index, Bucket *next)
{
int next_size;
int next_size;
ASSERT(self->len > 1, "split of empty bucket", -1);
ASSERT(self->len > 1, "split of empty bucket", -1);
if (index < 0 || index >= self->len)
index = self->len / 2;
if (index < 0 || index >= self->len)
index = self->len / 2;
next_size = self->len - index;
next_size = self->len - index;
next->keys = BTree_Malloc(sizeof(KEY_TYPE) * next_size);
if (!next->keys)
return -1;
memcpy(next->keys, self->keys + index, sizeof(KEY_TYPE) * next_size);
if (self->values) {
next->values = BTree_Malloc(sizeof(VALUE_TYPE) * next_size);
if (!next->values) {
free(next->keys);
next->keys = NULL;
return -1;
}
memcpy(next->values, self->values + index,
sizeof(VALUE_TYPE) * next_size);
next->keys = BTree_Malloc(sizeof(KEY_TYPE) * next_size);
if (!next->keys)
return -1;
memcpy(next->keys, self->keys + index, sizeof(KEY_TYPE) * next_size);
if (self->values) {
next->values = BTree_Malloc(sizeof(VALUE_TYPE) * next_size);
if (!next->values) {
free(next->keys);
next->keys = NULL;
return -1;
}
next->size = next_size;
next->len = next_size;
self->len = index;
memcpy(next->values, self->values + index,
sizeof(VALUE_TYPE) * next_size);
}
next->size = next_size;
next->len = next_size;
self->len = index;
next->next = self->next;
next->next = self->next;
Py_INCREF(next);
self->next = next;
Py_INCREF(next);
self->next = next;
if (PER_CHANGED(self) < 0)
return -1;
if (PER_CHANGED(self) < 0)
return -1;
return 0;
return 0;
}
/* Set self->next to self->next->next, i.e. unlink self's successor from
......@@ -580,106 +580,106 @@ bucket_split(Bucket *self, int index, Bucket *next)
static int
Bucket_deleteNextBucket(Bucket *self)
{
int result = -1; /* until proven innocent */
Bucket *successor;
PER_USE_OR_RETURN(self, -1);
successor = self->next;
if (successor) {
Bucket *next;
/* Before: self -> successor -> next
* After: self --------------> next
*/
UNLESS (PER_USE(successor)) goto Done;
next = successor->next;
PER_UNUSE(successor);
Py_XINCREF(next); /* it may be NULL, of course */
self->next = next;
Py_DECREF(successor);
if (PER_CHANGED(self) < 0)
int result = -1; /* until proven innocent */
Bucket *successor;
PER_USE_OR_RETURN(self, -1);
successor = self->next;
if (successor) {
Bucket *next;
/* Before: self -> successor -> next
* After: self --------------> next
*/
UNLESS (PER_USE(successor)) goto Done;
next = successor->next;
PER_UNUSE(successor);
Py_XINCREF(next); /* it may be NULL, of course */
self->next = next;
Py_DECREF(successor);
if (PER_CHANGED(self) < 0)
goto Done;
}
result = 0;
}
result = 0;
Done:
PER_UNUSE(self);
return result;
Done:
PER_UNUSE(self);
return result;
}
/*
Bucket_findRangeEnd -- Find the index of a range endpoint
(possibly) contained in a bucket.
Arguments: self The bucket
keyarg The key to match against
low Boolean; true for low end of range, false for high
exclude_equal Boolean; if true, don't accept an exact match,
and if there is one then move right if low and
left if !low.
offset The output offset
If low true, *offset <- index of the smallest item >= key,
if low false the index of the largest item <= key. In either case, if there
is no such index, *offset is left alone and 0 is returned.
Return:
0 No suitable index exists; *offset has not been changed
1 The correct index was stored into *offset
-1 Error
Example: Suppose the keys are [2, 4], and exclude_equal is false. Searching
for 2 sets *offset to 0 and returns 1 regardless of low. Searching for 4
sets *offset to 1 and returns 1 regardless of low.
Searching for 1:
If low true, sets *offset to 0, returns 1.
If low false, returns 0.
Searching for 3:
If low true, sets *offset to 1, returns 1.
If low false, sets *offset to 0, returns 1.
Searching for 5:
If low true, returns 0.
If low false, sets *offset to 1, returns 1.
The 1, 3 and 5 examples are the same when exclude_equal is true.
*/
Bucket_findRangeEnd -- Find the index of a range endpoint
(possibly) contained in a bucket.
Arguments: self The bucket
keyarg The key to match against
low Boolean; true for low end of range, false for high
exclude_equal Boolean; if true, don't accept an exact match,
and if there is one then move right if low and
left if !low.
offset The output offset
If low true, *offset <- index of the smallest item >= key,
if low false the index of the largest item <= key. In either case, if there
is no such index, *offset is left alone and 0 is returned.
Return:
0 No suitable index exists; *offset has not been changed
1 The correct index was stored into *offset
-1 Error
Example: Suppose the keys are [2, 4], and exclude_equal is false. Searching
for 2 sets *offset to 0 and returns 1 regardless of low. Searching for 4
sets *offset to 1 and returns 1 regardless of low.
Searching for 1:
If low true, sets *offset to 0, returns 1.
If low false, returns 0.
Searching for 3:
If low true, sets *offset to 1, returns 1.
If low false, sets *offset to 0, returns 1.
Searching for 5:
If low true, returns 0.
If low false, sets *offset to 1, returns 1.
The 1, 3 and 5 examples are the same when exclude_equal is true.
*/
static int
Bucket_findRangeEnd(Bucket *self, PyObject *keyarg, int low, int exclude_equal,
int *offset)
int *offset)
{
int i, cmp;
int result = -1; /* until proven innocent */
KEY_TYPE key;
int copied = 1;
int i, cmp;
int result = -1; /* until proven innocent */
KEY_TYPE key;
int copied = 1;
COPY_KEY_FROM_ARG(key, keyarg, copied);
UNLESS (copied) return -1;
COPY_KEY_FROM_ARG(key, keyarg, copied);
UNLESS (copied) return -1;
UNLESS (PER_USE(self)) return -1;
UNLESS (PER_USE(self)) return -1;
BUCKET_SEARCH(i, cmp, self, key, goto Done);
if (cmp == 0) {
/* exact match at index i */
if (exclude_equal) {
BUCKET_SEARCH(i, cmp, self, key, goto Done);
if (cmp == 0) {
/* exact match at index i */
if (exclude_equal) {
/* but we don't want an exact match */
if (low)
++i;
else
--i;
}
}
/* Else keys[i-1] < key < keys[i], picturing infinities at OOB indices,
* and i has the smallest item > key, which is correct for low.
*/
else if (! low)
/* i-1 has the largest item < key (unless i-1 is 0OB) */
if (low)
++i;
else
--i;
result = 0 <= i && i < self->len;
if (result)
*offset = i;
Done:
}
}
/* Else keys[i-1] < key < keys[i], picturing infinities at OOB indices,
* and i has the smallest item > key, which is correct for low.
*/
else if (! low)
/* i-1 has the largest item < key (unless i-1 is 0OB) */
--i;
result = 0 <= i && i < self->len;
if (result)
*offset = i;
Done:
PER_UNUSE(self);
return result;
}
......@@ -717,8 +717,8 @@ Bucket_maxminKey(Bucket *self, PyObject *args, int min)
empty:
PyErr_SetString(PyExc_ValueError,
empty_bucket ? "empty bucket" :
"no key satisfies the conditions");
empty_bucket ? "empty bucket" :
"no key satisfies the conditions");
PER_UNUSE(self);
return NULL;
}
......@@ -737,67 +737,67 @@ Bucket_maxKey(Bucket *self, PyObject *args)
static int
Bucket_rangeSearch(Bucket *self, PyObject *args, PyObject *kw,
int *low, int *high)
int *low, int *high)
{
PyObject *min = Py_None;
PyObject *max = Py_None;
int excludemin = 0;
int excludemax = 0;
int rc;
if (args) {
if (! PyArg_ParseTupleAndKeywords(args, kw, "|OOii", search_keywords,
&min,
&max,
&excludemin,
&excludemax))
PyObject *min = Py_None;
PyObject *max = Py_None;
int excludemin = 0;
int excludemax = 0;
int rc;
if (args) {
if (! PyArg_ParseTupleAndKeywords(args, kw, "|OOii", search_keywords,
&min,
&max,
&excludemin,
&excludemax))
return -1;
}
}
UNLESS (self->len) goto empty;
UNLESS (self->len) goto empty;
/* Find the low range */
if (min != Py_None) {
rc = Bucket_findRangeEnd(self, min, 1, excludemin, low);
if (rc < 0)
return -1;
if (rc == 0)
goto empty;
}
else {
*low = 0;
if (excludemin) {
if (self->len < 2)
goto empty;
++*low;
}
/* Find the low range */
if (min != Py_None) {
rc = Bucket_findRangeEnd(self, min, 1, excludemin, low);
if (rc < 0)
return -1;
if (rc == 0)
goto empty;
}
else {
*low = 0;
if (excludemin) {
if (self->len < 2)
goto empty;
++*low;
}
}
/* Find the high range */
if (max != Py_None) {
rc = Bucket_findRangeEnd(self, max, 0, excludemax, high);
if (rc < 0)
return -1;
if (rc == 0)
goto empty;
}
else {
*high = self->len - 1;
if (excludemax) {
/* Find the high range */
if (max != Py_None) {
rc = Bucket_findRangeEnd(self, max, 0, excludemax, high);
if (rc < 0)
return -1;
if (rc == 0)
goto empty;
}
else {
*high = self->len - 1;
if (excludemax) {
if (self->len < 2)
goto empty;
--*high;
}
}
}
/* If min < max to begin with, it's quite possible that low > high now. */
if (*low <= *high)
return 0;
/* If min < max to begin with, it's quite possible that low > high now. */
if (*low <= *high)
return 0;
empty:
*low = 0;
*high = -1;
return 0;
*low = 0;
*high = -1;
return 0;
}
/*
......@@ -819,16 +819,16 @@ bucket_keys(Bucket *self, PyObject *args, PyObject *kw)
PER_USE_OR_RETURN(self, NULL);
if (Bucket_rangeSearch(self, args, kw, &low, &high) < 0)
goto err;
goto err;
r = PyList_New(high-low+1);
if (r == NULL)
goto err;
goto err;
for (i=low; i <= high; i++) {
COPY_KEY_TO_OBJECT(key, self->keys[i]);
if (PyList_SetItem(r, i-low , key) < 0)
goto err;
COPY_KEY_TO_OBJECT(key, self->keys[i]);
if (PyList_SetItem(r, i-low , key) < 0)
goto err;
}
PER_UNUSE(self);
......@@ -992,84 +992,84 @@ bucket_byValue(Bucket *self, PyObject *omin)
static int
_bucket_clear(Bucket *self)
{
const int len = self->len;
/* Don't declare i at this level. If neither keys nor values are
* PyObject*, i won't be referenced, and you'll get a nuisance compiler
* wng for declaring it here.
*/
self->len = self->size = 0;
if (self->next) {
Py_DECREF(self->next);
self->next = NULL;
}
const int len = self->len;
/* Don't declare i at this level. If neither keys nor values are
* PyObject*, i won't be referenced, and you'll get a nuisance compiler
* wng for declaring it here.
*/
self->len = self->size = 0;
if (self->next) {
Py_DECREF(self->next);
self->next = NULL;
}
/* Silence compiler warning about unused variable len for the case
when neither key nor value is an object, i.e. II. */
(void)len;
/* Silence compiler warning about unused variable len for the case
when neither key nor value is an object, i.e. II. */
(void)len;
if (self->keys) {
if (self->keys) {
#ifdef KEY_TYPE_IS_PYOBJECT
int i;
for (i = 0; i < len; ++i)
DECREF_KEY(self->keys[i]);
int i;
for (i = 0; i < len; ++i)
DECREF_KEY(self->keys[i]);
#endif
free(self->keys);
self->keys = NULL;
}
free(self->keys);
self->keys = NULL;
}
if (self->values) {
if (self->values) {
#ifdef VALUE_TYPE_IS_PYOBJECT
int i;
for (i = 0; i < len; ++i)
DECREF_VALUE(self->values[i]);
int i;
for (i = 0; i < len; ++i)
DECREF_VALUE(self->values[i]);
#endif
free(self->values);
self->values = NULL;
}
return 0;
free(self->values);
self->values = NULL;
}
return 0;
}
#ifdef PERSISTENT
static PyObject *
bucket__p_deactivate(Bucket *self, PyObject *args, PyObject *keywords)
{
int ghostify = 1;
PyObject *force = NULL;
int ghostify = 1;
PyObject *force = NULL;
if (args && PyTuple_GET_SIZE(args) > 0) {
PyErr_SetString(PyExc_TypeError,
"_p_deactivate takes not positional arguments");
return NULL;
}
if (keywords) {
int size = PyDict_Size(keywords);
force = PyDict_GetItemString(keywords, "force");
if (force)
if (args && PyTuple_GET_SIZE(args) > 0) {
PyErr_SetString(PyExc_TypeError,
"_p_deactivate takes no positional arguments");
return NULL;
}
if (keywords) {
int size = PyDict_Size(keywords);
force = PyDict_GetItemString(keywords, "force");
if (force)
size--;
if (size) {
if (size) {
PyErr_SetString(PyExc_TypeError,
"_p_deactivate only accepts keyword arg force");
"_p_deactivate only accepts keyword arg force");
return NULL;
}
}
}
if (self->jar && self->oid) {
ghostify = self->state == cPersistent_UPTODATE_STATE;
if (!ghostify && force) {
if (self->jar && self->oid) {
ghostify = self->state == cPersistent_UPTODATE_STATE;
if (!ghostify && force) {
if (PyObject_IsTrue(force))
ghostify = 1;
ghostify = 1;
if (PyErr_Occurred())
return NULL;
}
if (ghostify) {
return NULL;
}
if (ghostify) {
if (_bucket_clear(self) < 0)
return NULL;
return NULL;
PER_GHOSTIFY(self);
}
}
Py_INCREF(Py_None);
return Py_None;
}
Py_INCREF(Py_None);
return Py_None;
}
#endif
......@@ -1079,16 +1079,16 @@ bucket_clear(Bucket *self, PyObject *args)
PER_USE_OR_RETURN(self, NULL);
if (self->len) {
if (_bucket_clear(self) < 0)
return NULL;
if (PER_CHANGED(self) < 0)
goto err;
if (_bucket_clear(self) < 0)
return NULL;
if (PER_CHANGED(self) < 0)
goto err;
}
PER_UNUSE(self);
Py_INCREF(Py_None);
return Py_None;
err:
err:
PER_UNUSE(self);
return NULL;
}
......@@ -1119,178 +1119,178 @@ err:
static PyObject *
bucket_getstate(Bucket *self)
{
PyObject *o = NULL, *items = NULL, *state;
int i, len, l;
PyObject *o = NULL, *items = NULL, *state;
int i, len, l;
PER_USE_OR_RETURN(self, NULL);
PER_USE_OR_RETURN(self, NULL);
len = self->len;
len = self->len;
if (self->values) { /* Bucket */
items = PyTuple_New(len * 2);
if (items == NULL)
if (self->values) { /* Bucket */
items = PyTuple_New(len * 2);
if (items == NULL)
goto err;
for (i = 0, l = 0; i < len; i++) {
for (i = 0, l = 0; i < len; i++) {
COPY_KEY_TO_OBJECT(o, self->keys[i]);
if (o == NULL)
goto err;
goto err;
PyTuple_SET_ITEM(items, l, o);
l++;
COPY_VALUE_TO_OBJECT(o, self->values[i]);
if (o == NULL)
goto err;
goto err;
PyTuple_SET_ITEM(items, l, o);
l++;
}
} else { /* Set */
items = PyTuple_New(len);
if (items == NULL)
}
} else { /* Set */
items = PyTuple_New(len);
if (items == NULL)
goto err;
for (i = 0; i < len; i++) {
for (i = 0; i < len; i++) {
COPY_KEY_TO_OBJECT(o, self->keys[i]);
if (o == NULL)
goto err;
goto err;
PyTuple_SET_ITEM(items, i, o);
}
}
}
if (self->next)
state = Py_BuildValue("OO", items, self->next);
else
state = Py_BuildValue("(O)", items);
Py_DECREF(items);
if (self->next)
state = Py_BuildValue("OO", items, self->next);
else
state = Py_BuildValue("(O)", items);
Py_DECREF(items);
PER_UNUSE(self);
return state;
PER_UNUSE(self);
return state;
err:
PER_UNUSE(self);
Py_XDECREF(items);
return NULL;
PER_UNUSE(self);
Py_XDECREF(items);
return NULL;
}
static int
_bucket_setstate(Bucket *self, PyObject *state)
{
PyObject *k, *v, *items;
Bucket *next = NULL;
int i, l, len, copied=1;
KEY_TYPE *keys;
VALUE_TYPE *values;
if (!PyArg_ParseTuple(state, "O|O:__setstate__", &items, &next))
return -1;
if (!PyTuple_Check(items)) {
PyErr_SetString(PyExc_TypeError,
"tuple required for first state element");
return -1;
}
PyObject *k, *v, *items;
Bucket *next = NULL;
int i, l, len, copied=1;
KEY_TYPE *keys;
VALUE_TYPE *values;
if (!PyArg_ParseTuple(state, "O|O:__setstate__", &items, &next))
return -1;
if (!PyTuple_Check(items)) {
PyErr_SetString(PyExc_TypeError,
"tuple required for first state element");
return -1;
}
len = PyTuple_Size(items);
if (len < 0)
return -1;
len /= 2;
len = PyTuple_Size(items);
if (len < 0)
return -1;
len /= 2;
for (i = self->len; --i >= 0; ) {
DECREF_KEY(self->keys[i]);
DECREF_VALUE(self->values[i]);
}
self->len = 0;
for (i = self->len; --i >= 0; ) {
DECREF_KEY(self->keys[i]);
DECREF_VALUE(self->values[i]);
}
self->len = 0;
if (self->next) {
Py_DECREF(self->next);
self->next = NULL;
}
if (self->next) {
Py_DECREF(self->next);
self->next = NULL;
}
if (len > self->size) {
keys = BTree_Realloc(self->keys, sizeof(KEY_TYPE)*len);
if (keys == NULL)
if (len > self->size) {
keys = BTree_Realloc(self->keys, sizeof(KEY_TYPE)*len);
if (keys == NULL)
return -1;
values = BTree_Realloc(self->values, sizeof(VALUE_TYPE)*len);
if (values == NULL)
values = BTree_Realloc(self->values, sizeof(VALUE_TYPE)*len);
if (values == NULL)
return -1;
self->keys = keys;
self->values = values;
self->size = len;
}
self->keys = keys;
self->values = values;
self->size = len;
}
for (i=0, l=0; i < len; i++) {
k = PyTuple_GET_ITEM(items, l);
l++;
v = PyTuple_GET_ITEM(items, l);
l++;
for (i=0, l=0; i < len; i++) {
k = PyTuple_GET_ITEM(items, l);
l++;
v = PyTuple_GET_ITEM(items, l);
l++;
COPY_KEY_FROM_ARG(self->keys[i], k, copied);
if (!copied)
COPY_KEY_FROM_ARG(self->keys[i], k, copied);
if (!copied)
return -1;
COPY_VALUE_FROM_ARG(self->values[i], v, copied);
if (!copied)
COPY_VALUE_FROM_ARG(self->values[i], v, copied);
if (!copied)
return -1;
INCREF_KEY(self->keys[i]);
INCREF_VALUE(self->values[i]);
}
INCREF_KEY(self->keys[i]);
INCREF_VALUE(self->values[i]);
}
self->len = len;
self->len = len;
if (next) {
self->next = next;
Py_INCREF(next);
}
if (next) {
self->next = next;
Py_INCREF(next);
}
return 0;
return 0;
}
static PyObject *
bucket_setstate(Bucket *self, PyObject *state)
{
int r;
int r;
PER_PREVENT_DEACTIVATION(self);
r = _bucket_setstate(self, state);
PER_UNUSE(self);
PER_PREVENT_DEACTIVATION(self);
r = _bucket_setstate(self, state);
PER_UNUSE(self);
if (r < 0)
return NULL;
Py_INCREF(Py_None);
return Py_None;
if (r < 0)
return NULL;
Py_INCREF(Py_None);
return Py_None;
}
static PyObject *
bucket_has_key(Bucket *self, PyObject *key)
{
return _bucket_get(self, key, 1);
return _bucket_get(self, key, 1);
}
static PyObject *
bucket_setdefault(Bucket *self, PyObject *args)
{
PyObject *key;
PyObject *failobj; /* default */
PyObject *value; /* return value */
int dummy_changed; /* in order to call _bucket_set */
if (! PyArg_UnpackTuple(args, "setdefault", 2, 2, &key, &failobj))
return NULL;
PyObject *key;
PyObject *failobj; /* default */
PyObject *value; /* return value */
int dummy_changed; /* in order to call _bucket_set */
value = _bucket_get(self, key, 0);
if (value != NULL)
return value;
/* The key isn't in the bucket. If that's not due to a KeyError exception,
* pass back the unexpected exception.
*/
if (! PyErr_ExceptionMatches(PyExc_KeyError))
return NULL;
PyErr_Clear();
if (! PyArg_UnpackTuple(args, "setdefault", 2, 2, &key, &failobj))
return NULL;
/* Associate `key` with `failobj` in the bucket, and return `failobj`. */
value = failobj;
if (_bucket_set(self, key, failobj, 0, 0, &dummy_changed) < 0)
value = NULL;
Py_XINCREF(value);
value = _bucket_get(self, key, 0);
if (value != NULL)
return value;
/* The key isn't in the bucket. If that's not due to a KeyError exception,
* pass back the unexpected exception.
*/
if (! PyErr_ExceptionMatches(PyExc_KeyError))
return NULL;
PyErr_Clear();
/* Associate `key` with `failobj` in the bucket, and return `failobj`. */
value = failobj;
if (_bucket_set(self, key, failobj, 0, 0, &dummy_changed) < 0)
value = NULL;
Py_XINCREF(value);
return value;
}
......@@ -1301,44 +1301,44 @@ Bucket_length(Bucket *self);
static PyObject *
bucket_pop(Bucket *self, PyObject *args)
{
PyObject *key;
PyObject *failobj = NULL; /* default */
PyObject *value; /* return value */
int dummy_changed; /* in order to call _bucket_set */
if (! PyArg_UnpackTuple(args, "pop", 1, 2, &key, &failobj))
return NULL;
value = _bucket_get(self, key, 0);
if (value != NULL) {
/* Delete key and associated value. */
if (_bucket_set(self, key, NULL, 0, 0, &dummy_changed) < 0) {
Py_DECREF(value);
return NULL;
}
return value;
}
PyObject *key;
PyObject *failobj = NULL; /* default */
PyObject *value; /* return value */
int dummy_changed; /* in order to call _bucket_set */
/* The key isn't in the bucket. If that's not due to a KeyError exception,
* pass back the unexpected exception.
*/
if (! PyErr_ExceptionMatches(PyExc_KeyError))
return NULL;
if (! PyArg_UnpackTuple(args, "pop", 1, 2, &key, &failobj))
return NULL;
if (failobj != NULL) {
/* Clear the KeyError and return the explicit default. */
PyErr_Clear();
Py_INCREF(failobj);
return failobj;
value = _bucket_get(self, key, 0);
if (value != NULL) {
/* Delete key and associated value. */
if (_bucket_set(self, key, NULL, 0, 0, &dummy_changed) < 0) {
Py_DECREF(value);
return NULL;
}
return value;
}
/* No default given. The only difference in this case is the error
* message, which depends on whether the bucket is empty.
*/
if (Bucket_length(self) == 0)
PyErr_SetString(PyExc_KeyError, "pop(): Bucket is empty");
/* The key isn't in the bucket. If that's not due to a KeyError exception,
* pass back the unexpected exception.
*/
if (! PyErr_ExceptionMatches(PyExc_KeyError))
return NULL;
}
if (failobj != NULL) {
/* Clear the KeyError and return the explicit default. */
PyErr_Clear();
Py_INCREF(failobj);
return failobj;
}
/* No default given. The only difference in this case is the error
* message, which depends on whether the bucket is empty.
*/
if (Bucket_length(self) == 0)
PyErr_SetString(PyExc_KeyError, "pop(): Bucket is empty");
return NULL;
}
/* Search bucket self for key. This is the sq_contains slot of the
* PySequenceMethods.
......@@ -1351,14 +1351,14 @@ bucket_pop(Bucket *self, PyObject *args)
static int
bucket_contains(Bucket *self, PyObject *key)
{
PyObject *asobj = _bucket_get(self, key, 1);
int result = -1;
PyObject *asobj = _bucket_get(self, key, 1);
int result = -1;
if (asobj != NULL) {
result = PyInt_AsLong(asobj) ? 1 : 0;
Py_DECREF(asobj);
}
return result;
if (asobj != NULL) {
result = PyInt_AsLong(asobj) ? 1 : 0;
Py_DECREF(asobj);
}
return result;
}
/*
......@@ -1368,18 +1368,18 @@ bucket_contains(Bucket *self, PyObject *key)
static PyObject *
bucket_getm(Bucket *self, PyObject *args)
{
PyObject *key, *d=Py_None, *r;
if (!PyArg_ParseTuple(args, "O|O:get", &key, &d))
return NULL;
r = _bucket_get(self, key, 0);
if (r)
return r;
if (!PyErr_ExceptionMatches(PyExc_KeyError))
return NULL;
PyErr_Clear();
Py_INCREF(d);
return d;
PyObject *key, *d=Py_None, *r;
if (!PyArg_ParseTuple(args, "O|O:get", &key, &d))
return NULL;
r = _bucket_get(self, key, 0);
if (r)
return r;
if (!PyErr_ExceptionMatches(PyExc_KeyError))
return NULL;
PyErr_Clear();
Py_INCREF(d);
return d;
}
/**************************************************************************/
......@@ -1394,52 +1394,52 @@ bucket_getm(Bucket *self, PyObject *args)
static PyObject *
buildBucketIter(Bucket *self, PyObject *args, PyObject *kw, char kind)
{
BTreeItems *items;
int lowoffset, highoffset;
BTreeIter *result = NULL;
BTreeItems *items;
int lowoffset, highoffset;
BTreeIter *result = NULL;
PER_USE_OR_RETURN(self, NULL);
if (Bucket_rangeSearch(self, args, kw, &lowoffset, &highoffset) < 0)
goto Done;
PER_USE_OR_RETURN(self, NULL);
if (Bucket_rangeSearch(self, args, kw, &lowoffset, &highoffset) < 0)
goto Done;
items = (BTreeItems *)newBTreeItems(kind, self, lowoffset,
self, highoffset);
if (items == NULL) goto Done;
items = (BTreeItems *)newBTreeItems(kind, self, lowoffset,
self, highoffset);
if (items == NULL) goto Done;
result = BTreeIter_new(items); /* win or lose, we're done */
Py_DECREF(items);
result = BTreeIter_new(items); /* win or lose, we're done */
Py_DECREF(items);
Done:
PER_UNUSE(self);
return (PyObject *)result;
Done:
PER_UNUSE(self);
return (PyObject *)result;
}
/* The implementation of iter(Bucket_or_Set); the Bucket tp_iter slot. */
static PyObject *
Bucket_getiter(Bucket *self)
{
return buildBucketIter(self, NULL, NULL, 'k');
return buildBucketIter(self, NULL, NULL, 'k');
}
/* The implementation of Bucket.iterkeys(). */
static PyObject *
Bucket_iterkeys(Bucket *self, PyObject *args, PyObject *kw)
{
return buildBucketIter(self, args, kw, 'k');
return buildBucketIter(self, args, kw, 'k');
}
/* The implementation of Bucket.itervalues(). */
static PyObject *
Bucket_itervalues(Bucket *self, PyObject *args, PyObject *kw)
{
return buildBucketIter(self, args, kw, 'v');
return buildBucketIter(self, args, kw, 'v');
}
/* The implementation of Bucket.iteritems(). */
static PyObject *
Bucket_iteritems(Bucket *self, PyObject *args, PyObject *kw)
{
return buildBucketIter(self, args, kw, 'i');
return buildBucketIter(self, args, kw, 'i');
}
/* End of iterator support. */
......@@ -1451,61 +1451,61 @@ static PyObject *bucket_merge(Bucket *s1, Bucket *s2, Bucket *s3);
static PyObject *
_bucket__p_resolveConflict(PyObject *ob_type, PyObject *s[3])
{
PyObject *result = NULL; /* guilty until proved innocent */
Bucket *b[3] = {NULL, NULL, NULL};
PyObject *meth = NULL;
PyObject *a = NULL;
int i;
PyObject *result = NULL; /* guilty until proved innocent */
Bucket *b[3] = {NULL, NULL, NULL};
PyObject *meth = NULL;
PyObject *a = NULL;
int i;
for (i = 0; i < 3; i++) {
PyObject *r;
for (i = 0; i < 3; i++) {
PyObject *r;
b[i] = (Bucket*)PyObject_CallObject((PyObject *)ob_type, NULL);
if (b[i] == NULL)
b[i] = (Bucket*)PyObject_CallObject((PyObject *)ob_type, NULL);
if (b[i] == NULL)
goto Done;
if (s[i] == Py_None) /* None is equivalent to empty, for BTrees */
if (s[i] == Py_None) /* None is equivalent to empty, for BTrees */
continue;
meth = PyObject_GetAttr((PyObject *)b[i], __setstate___str);
if (meth == NULL)
meth = PyObject_GetAttr((PyObject *)b[i], __setstate___str);
if (meth == NULL)
goto Done;
a = PyTuple_New(1);
if (a == NULL)
a = PyTuple_New(1);
if (a == NULL)
goto Done;
PyTuple_SET_ITEM(a, 0, s[i]);
Py_INCREF(s[i]);
r = PyObject_CallObject(meth, a); /* b[i].__setstate__(s[i]) */
if (r == NULL)
PyTuple_SET_ITEM(a, 0, s[i]);
Py_INCREF(s[i]);
r = PyObject_CallObject(meth, a); /* b[i].__setstate__(s[i]) */
if (r == NULL)
goto Done;
Py_DECREF(r);
Py_DECREF(a);
Py_DECREF(meth);
a = meth = NULL;
}
Py_DECREF(r);
Py_DECREF(a);
Py_DECREF(meth);
a = meth = NULL;
}
if (b[0]->next != b[1]->next || b[0]->next != b[2]->next)
merge_error(-1, -1, -1, 0);
else
result = bucket_merge(b[0], b[1], b[2]);
if (b[0]->next != b[1]->next || b[0]->next != b[2]->next)
merge_error(-1, -1, -1, 0);
else
result = bucket_merge(b[0], b[1], b[2]);
Done:
Py_XDECREF(meth);
Py_XDECREF(a);
Py_XDECREF(b[0]);
Py_XDECREF(b[1]);
Py_XDECREF(b[2]);
Done:
Py_XDECREF(meth);
Py_XDECREF(a);
Py_XDECREF(b[0]);
Py_XDECREF(b[1]);
Py_XDECREF(b[2]);
return result;
return result;
}
static PyObject *
bucket__p_resolveConflict(Bucket *self, PyObject *args)
{
PyObject *s[3];
PyObject *s[3];
if (!PyArg_ParseTuple(args, "OOO", &s[0], &s[1], &s[2]))
return NULL;
if (!PyArg_ParseTuple(args, "OOO", &s[0], &s[1], &s[2]))
return NULL;
return _bucket__p_resolveConflict((PyObject *)self->ob_type, s);
return _bucket__p_resolveConflict((PyObject *)self->ob_type, s);
}
#endif
......@@ -1518,157 +1518,157 @@ bucket__p_resolveConflict(Bucket *self, PyObject *args)
*/
static struct PyMemberDef Bucket_members[] = {
{"_next", T_OBJECT, offsetof(Bucket, next)},
{NULL}
{"_next", T_OBJECT, offsetof(Bucket, next)},
{NULL}
};
static struct PyMethodDef Bucket_methods[] = {
{"__getstate__", (PyCFunction) bucket_getstate, METH_NOARGS,
"__getstate__() -- Return the picklable state of the object"},
{"__getstate__", (PyCFunction) bucket_getstate, METH_NOARGS,
"__getstate__() -- Return the picklable state of the object"},
{"__setstate__", (PyCFunction) bucket_setstate, METH_O,
"__setstate__() -- Set the state of the object"},
{"__setstate__", (PyCFunction) bucket_setstate, METH_O,
"__setstate__() -- Set the state of the object"},
{"keys", (PyCFunction) bucket_keys, METH_KEYWORDS,
"keys([min, max]) -- Return the keys"},
{"keys", (PyCFunction) bucket_keys, METH_KEYWORDS,
"keys([min, max]) -- Return the keys"},
{"has_key", (PyCFunction) bucket_has_key, METH_O,
"has_key(key) -- Test whether the bucket contains the given key"},
{"has_key", (PyCFunction) bucket_has_key, METH_O,
"has_key(key) -- Test whether the bucket contains the given key"},
{"clear", (PyCFunction) bucket_clear, METH_VARARGS,
"clear() -- Remove all of the items from the bucket"},
{"clear", (PyCFunction) bucket_clear, METH_VARARGS,
"clear() -- Remove all of the items from the bucket"},
{"update", (PyCFunction) Mapping_update, METH_O,
"update(collection) -- Add the items from the given collection"},
{"update", (PyCFunction) Mapping_update, METH_O,
"update(collection) -- Add the items from the given collection"},
{"maxKey", (PyCFunction) Bucket_maxKey, METH_VARARGS,
"maxKey([key]) -- Find the maximum key\n\n"
"If an argument is given, find the maximum <= the argument"},
{"maxKey", (PyCFunction) Bucket_maxKey, METH_VARARGS,
"maxKey([key]) -- Find the maximum key\n\n"
"If an argument is given, find the maximum <= the argument"},
{"minKey", (PyCFunction) Bucket_minKey, METH_VARARGS,
"minKey([key]) -- Find the minimum key\n\n"
"If an argument is given, find the minimum >= the argument"},
{"minKey", (PyCFunction) Bucket_minKey, METH_VARARGS,
"minKey([key]) -- Find the minimum key\n\n"
"If an argument is given, find the minimum >= the argument"},
{"values", (PyCFunction) bucket_values, METH_KEYWORDS,
"values([min, max]) -- Return the values"},
{"values", (PyCFunction) bucket_values, METH_KEYWORDS,
"values([min, max]) -- Return the values"},
{"items", (PyCFunction) bucket_items, METH_KEYWORDS,
"items([min, max])) -- Return the items"},
{"items", (PyCFunction) bucket_items, METH_KEYWORDS,
"items([min, max])) -- Return the items"},
{"byValue", (PyCFunction) bucket_byValue, METH_O,
"byValue(min) -- "
"Return value-keys with values >= min and reverse sorted by values"},
{"byValue", (PyCFunction) bucket_byValue, METH_O,
"byValue(min) -- "
"Return value-keys with values >= min and reverse sorted by values"},
{"get", (PyCFunction) bucket_getm, METH_VARARGS,
"get(key[,default]) -- Look up a value\n\n"
"Return the default (or None) if the key is not found."},
{"get", (PyCFunction) bucket_getm, METH_VARARGS,
"get(key[,default]) -- Look up a value\n\n"
"Return the default (or None) if the key is not found."},
{"setdefault", (PyCFunction) bucket_setdefault, METH_VARARGS,
"D.setdefault(k, d) -> D.get(k, d), also set D[k]=d if k not in D.\n\n"
"Return the value like get() except that if key is missing, d is both\n"
"returned and inserted into the bucket as the value of k."},
{"setdefault", (PyCFunction) bucket_setdefault, METH_VARARGS,
"D.setdefault(k, d) -> D.get(k, d), also set D[k]=d if k not in D.\n\n"
"Return the value like get() except that if key is missing, d is both\n"
"returned and inserted into the bucket as the value of k."},
{"pop", (PyCFunction) bucket_pop, METH_VARARGS,
"D.pop(k[, d]) -> v, remove key and return the corresponding value.\n\n"
"If key is not found, d is returned if given, otherwise KeyError\n"
"is raised."},
{"pop", (PyCFunction) bucket_pop, METH_VARARGS,
"D.pop(k[, d]) -> v, remove key and return the corresponding value.\n\n"
"If key is not found, d is returned if given, otherwise KeyError\n"
"is raised."},
{"iterkeys", (PyCFunction) Bucket_iterkeys, METH_KEYWORDS,
"B.iterkeys([min[,max]]) -> an iterator over the keys of B"},
{"iterkeys", (PyCFunction) Bucket_iterkeys, METH_KEYWORDS,
"B.iterkeys([min[,max]]) -> an iterator over the keys of B"},
{"itervalues", (PyCFunction) Bucket_itervalues, METH_KEYWORDS,
"B.itervalues([min[,max]]) -> an iterator over the values of B"},
{"itervalues", (PyCFunction) Bucket_itervalues, METH_KEYWORDS,
"B.itervalues([min[,max]]) -> an iterator over the values of B"},
{"iteritems", (PyCFunction) Bucket_iteritems, METH_KEYWORDS,
"B.iteritems([min[,max]]) -> an iterator over the (key, value) items of B"},
{"iteritems", (PyCFunction) Bucket_iteritems, METH_KEYWORDS,
"B.iteritems([min[,max]]) -> an iterator over the (key, value) items of B"},
#ifdef EXTRA_BUCKET_METHODS
EXTRA_BUCKET_METHODS
EXTRA_BUCKET_METHODS
#endif
#ifdef PERSISTENT
{"_p_resolveConflict", (PyCFunction) bucket__p_resolveConflict,
METH_VARARGS,
"_p_resolveConflict() -- Reinitialize from a newly created copy"},
{"_p_resolveConflict", (PyCFunction) bucket__p_resolveConflict,
METH_VARARGS,
"_p_resolveConflict() -- Reinitialize from a newly created copy"},
{"_p_deactivate", (PyCFunction) bucket__p_deactivate, METH_KEYWORDS,
"_p_deactivate() -- Reinitialize from a newly created copy"},
{"_p_deactivate", (PyCFunction) bucket__p_deactivate, METH_KEYWORDS,
"_p_deactivate() -- Reinitialize from a newly created copy"},
#endif
{NULL, NULL}
{NULL, NULL}
};
static int
Bucket_init(PyObject *self, PyObject *args, PyObject *kwds)
{
PyObject *v = NULL;
PyObject *v = NULL;
if (!PyArg_ParseTuple(args, "|O:" MOD_NAME_PREFIX "Bucket", &v))
return -1;
if (!PyArg_ParseTuple(args, "|O:" MOD_NAME_PREFIX "Bucket", &v))
return -1;
if (v)
return update_from_seq(self, v);
else
return 0;
if (v)
return update_from_seq(self, v);
else
return 0;
}
static void
bucket_dealloc(Bucket *self)
{
if (self->state != cPersistent_GHOST_STATE)
_bucket_clear(self);
if (self->state != cPersistent_GHOST_STATE)
_bucket_clear(self);
cPersistenceCAPI->pertype->tp_dealloc((PyObject *)self);
cPersistenceCAPI->pertype->tp_dealloc((PyObject *)self);
}
static int
bucket_traverse(Bucket *self, visitproc visit, void *arg)
{
int err = 0;
int i, len;
int err = 0;
int i, len;
#define VISIT(SLOT) \
if (SLOT) { \
err = visit((PyObject *)(SLOT), arg); \
if (err) \
goto Done; \
}
/* Call our base type's traverse function. Because buckets are
* subclasses of Peristent, there must be one.
*/
err = cPersistenceCAPI->pertype->tp_traverse((PyObject *)self, visit, arg);
if (err)
goto Done;
/* If this is registered with the persistence system, cleaning up cycles
* is the database's problem. It would be horrid to unghostify buckets
* here just to chase pointers every time gc runs.
*/
if (self->state == cPersistent_GHOST_STATE)
goto Done;
if (SLOT) { \
err = visit((PyObject *)(SLOT), arg); \
if (err) \
goto Done; \
}
len = self->len;
(void)i; /* if neither keys nor values are PyObject*, "i" is otherwise
unreferenced and we get a nuisance compiler wng */
/* Call our base type's traverse function. Because buckets are
* subclasses of Peristent, there must be one.
*/
err = cPersistenceCAPI->pertype->tp_traverse((PyObject *)self, visit, arg);
if (err)
goto Done;
/* If this is registered with the persistence system, cleaning up cycles
* is the database's problem. It would be horrid to unghostify buckets
* here just to chase pointers every time gc runs.
*/
if (self->state == cPersistent_GHOST_STATE)
goto Done;
len = self->len;
(void)i; /* if neither keys nor values are PyObject*, "i" is otherwise
unreferenced and we get a nuisance compiler wng */
#ifdef KEY_TYPE_IS_PYOBJECT
/* Keys are Python objects so need to be traversed. */
for (i = 0; i < len; i++)
VISIT(self->keys[i]);
/* Keys are Python objects so need to be traversed. */
for (i = 0; i < len; i++)
VISIT(self->keys[i]);
#endif
#ifdef VALUE_TYPE_IS_PYOBJECT
if (self->values != NULL) {
/* self->values exists (this is a mapping bucket, not a set bucket),
* and are Python objects, so need to be traversed. */
for (i = 0; i < len; i++)
VISIT(self->values[i]);
}
if (self->values != NULL) {
/* self->values exists (this is a mapping bucket, not a set bucket),
* and are Python objects, so need to be traversed. */
for (i = 0; i < len; i++)
VISIT(self->values[i]);
}
#endif
VISIT(self->next);
VISIT(self->next);
Done:
return err;
Done:
return err;
#undef VISIT
}
......@@ -1676,20 +1676,20 @@ Done:
static int
bucket_tp_clear(Bucket *self)
{
if (self->state != cPersistent_GHOST_STATE)
_bucket_clear(self);
return 0;
if (self->state != cPersistent_GHOST_STATE)
_bucket_clear(self);
return 0;
}
/* Code to access Bucket objects as mappings */
static int
Bucket_length( Bucket *self)
{
int r;
UNLESS (PER_USE(self)) return -1;
r = self->len;
PER_UNUSE(self);
return r;
int r;
UNLESS (PER_USE(self)) return -1;
r = self->len;
PER_UNUSE(self);
return r;
}
static PyMappingMethods Bucket_as_mapping = {
......@@ -1699,100 +1699,100 @@ static PyMappingMethods Bucket_as_mapping = {
};
static PySequenceMethods Bucket_as_sequence = {
(lenfunc)0, /* sq_length */
(binaryfunc)0, /* sq_concat */
(ssizeargfunc)0, /* sq_repeat */
(ssizeargfunc)0, /* sq_item */
(ssizessizeargfunc)0, /* sq_slice */
(ssizeobjargproc)0, /* sq_ass_item */
(ssizessizeobjargproc)0, /* sq_ass_slice */
(objobjproc)bucket_contains, /* sq_contains */
0, /* sq_inplace_concat */
0, /* sq_inplace_repeat */
(lenfunc)0, /* sq_length */
(binaryfunc)0, /* sq_concat */
(ssizeargfunc)0, /* sq_repeat */
(ssizeargfunc)0, /* sq_item */
(ssizessizeargfunc)0, /* sq_slice */
(ssizeobjargproc)0, /* sq_ass_item */
(ssizessizeobjargproc)0, /* sq_ass_slice */
(objobjproc)bucket_contains, /* sq_contains */
0, /* sq_inplace_concat */
0, /* sq_inplace_repeat */
};
static PyObject *
bucket_repr(Bucket *self)
{
PyObject *i, *r;
char repr[10000];
int rv;
i = bucket_items(self, NULL, NULL);
if (!i)
return NULL;
r = PyObject_Repr(i);
Py_DECREF(i);
if (!r) {
return NULL;
}
rv = PyOS_snprintf(repr, sizeof(repr),
"%s(%s)", self->ob_type->tp_name,
PyString_AS_STRING(r));
if (rv > 0 && rv < sizeof(repr)) {
Py_DECREF(r);
return PyString_FromStringAndSize(repr, strlen(repr));
}
else {
/* The static buffer wasn't big enough */
int size;
PyObject *s;
/* 3 for the parens and the null byte */
size = strlen(self->ob_type->tp_name) + PyString_GET_SIZE(r) + 3;
s = PyString_FromStringAndSize(NULL, size);
if (!s) {
PyObject *i, *r;
char repr[10000];
int rv;
i = bucket_items(self, NULL, NULL);
if (!i)
return NULL;
r = PyObject_Repr(i);
Py_DECREF(i);
if (!r) {
return NULL;
}
rv = PyOS_snprintf(repr, sizeof(repr),
"%s(%s)", self->ob_type->tp_name,
PyString_AS_STRING(r));
if (rv > 0 && rv < sizeof(repr)) {
Py_DECREF(r);
return PyString_FromStringAndSize(repr, strlen(repr));
}
else {
/* The static buffer wasn't big enough */
int size;
PyObject *s;
/* 3 for the parens and the null byte */
size = strlen(self->ob_type->tp_name) + PyString_GET_SIZE(r) + 3;
s = PyString_FromStringAndSize(NULL, size);
if (!s) {
Py_DECREF(r);
return r;
}
PyOS_snprintf(PyString_AS_STRING(s), size,
"%s(%s)", self->ob_type->tp_name, PyString_AS_STRING(r));
Py_DECREF(r);
return s;
}
PyOS_snprintf(PyString_AS_STRING(s), size,
"%s(%s)", self->ob_type->tp_name, PyString_AS_STRING(r));
Py_DECREF(r);
return s;
}
}
static PyTypeObject BucketType = {
PyObject_HEAD_INIT(NULL) /* PyPersist_Type */
0, /* ob_size */
MODULE_NAME MOD_NAME_PREFIX "Bucket",/* tp_name */
sizeof(Bucket), /* tp_basicsize */
0, /* tp_itemsize */
(destructor)bucket_dealloc, /* tp_dealloc */
0, /* tp_print */
0, /* tp_getattr */
0, /* tp_setattr */
0, /* tp_compare */
(reprfunc)bucket_repr, /* tp_repr */
0, /* tp_as_number */
&Bucket_as_sequence, /* tp_as_sequence */
&Bucket_as_mapping, /* tp_as_mapping */
0, /* tp_hash */
0, /* tp_call */
0, /* tp_str */
0, /* tp_getattro */
0, /* tp_setattro */
0, /* tp_as_buffer */
Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC |
Py_TPFLAGS_BASETYPE, /* tp_flags */
0, /* tp_doc */
(traverseproc)bucket_traverse, /* tp_traverse */
(inquiry)bucket_tp_clear, /* tp_clear */
0, /* tp_richcompare */
0, /* tp_weaklistoffset */
(getiterfunc)Bucket_getiter, /* tp_iter */
0, /* tp_iternext */
Bucket_methods, /* tp_methods */
Bucket_members, /* tp_members */
0, /* tp_getset */
0, /* tp_base */
0, /* tp_dict */
0, /* tp_descr_get */
0, /* tp_descr_set */
0, /* tp_dictoffset */
Bucket_init, /* tp_init */
0, /* tp_alloc */
0, /*PyType_GenericNew,*/ /* tp_new */
PyObject_HEAD_INIT(NULL) /* PyPersist_Type */
0, /* ob_size */
MODULE_NAME MOD_NAME_PREFIX "Bucket",/* tp_name */
sizeof(Bucket), /* tp_basicsize */
0, /* tp_itemsize */
(destructor)bucket_dealloc, /* tp_dealloc */
0, /* tp_print */
0, /* tp_getattr */
0, /* tp_setattr */
0, /* tp_compare */
(reprfunc)bucket_repr, /* tp_repr */
0, /* tp_as_number */
&Bucket_as_sequence, /* tp_as_sequence */
&Bucket_as_mapping, /* tp_as_mapping */
0, /* tp_hash */
0, /* tp_call */
0, /* tp_str */
0, /* tp_getattro */
0, /* tp_setattro */
0, /* tp_as_buffer */
Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_GC |
Py_TPFLAGS_BASETYPE, /* tp_flags */
0, /* tp_doc */
(traverseproc)bucket_traverse, /* tp_traverse */
(inquiry)bucket_tp_clear, /* tp_clear */
0, /* tp_richcompare */
0, /* tp_weaklistoffset */
(getiterfunc)Bucket_getiter, /* tp_iter */
0, /* tp_iternext */
Bucket_methods, /* tp_methods */
Bucket_members, /* tp_members */
0, /* tp_getset */
0, /* tp_base */
0, /* tp_dict */
0, /* tp_descr_get */
0, /* tp_descr_set */
0, /* tp_dictoffset */
Bucket_init, /* tp_init */
0, /* tp_alloc */
0, /*PyType_GenericNew,*/ /* tp_new */
};
static int
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment