[issue15729] PyStructSequence_NewType enhancement

[Joannah Nanjekye]

Joannah Nanjekye  added the comment:

All enhancements specified in this issue have since been done. 
PyStructSequence_NewType() calls PyType_FromSpecWithBases() which sets the 
needed flags and even the qualname:

PyObject *
PyType_FromSpecWithBases(PyType_Spec *spec, PyObject *bases)
{
PyHeapTypeObject *res;
PyMemberDef *memb;
PyObject *modname;
PyTypeObject *type, *base;

PyType_Slot *slot;
Py_ssize_t nmembers;
char *s, *res_start;

nmembers = 0;
for (slot = spec->slots; slot->slot; slot++) {
if (slot->slot == Py_tp_members) {
nmembers = 0;
for (memb = slot->pfunc; memb->name != NULL; memb++) {
nmembers++;
}
}
}

res = (PyHeapTypeObject*)PyType_GenericAlloc(_Type, nmembers);
if (res == NULL)
return NULL;
res_start = (char*)res;

if (spec->name == NULL) {
PyErr_SetString(PyExc_SystemError,
"Type spec does not define the name field.");
goto fail;
}

/* Set the type name and qualname */
s = strrchr(spec->name, '.');
if (s == NULL)
s = (char*)spec->name;
else
s++;

type = >ht_type;
/* The flags must be initialized early, before the GC traverses us */
type->tp_flags = spec->flags | Py_TPFLAGS_HEAPTYPE;
res->ht_name = PyUnicode_FromString(s);
if (!res->ht_name)
goto fail;
res->ht_qualname = res->ht_name;
Py_INCREF(res->ht_qualname);
type->tp_name = spec->name;

/* Adjust for empty tuple bases */
if (!bases) {
base = _Type;
/* See whether Py_tp_base(s) was specified */
for (slot = spec->slots; slot->slot; slot++) {
if (slot->slot == Py_tp_base)
base = slot->pfunc;
else if (slot->slot == Py_tp_bases) {
bases = slot->pfunc;
Py_INCREF(bases);
}
}
if (!bases)
bases = PyTuple_Pack(1, base);
if (!bases)
goto fail;
}
else
Py_INCREF(bases);

/* Calculate best base, and check that all bases are type objects */
base = best_base(bases);
if (base == NULL) {
goto fail;
}
if (!PyType_HasFeature(base, Py_TPFLAGS_BASETYPE)) {
PyErr_Format(PyExc_TypeError,
 "type '%.100s' is not an acceptable base type",
 base->tp_name);
goto fail;
}

/* Initialize essential fields */
type->tp_as_async = >as_async;
type->tp_as_number = >as_number;
type->tp_as_sequence = >as_sequence;
type->tp_as_mapping = >as_mapping;
type->tp_as_buffer = >as_buffer;
/* Set tp_base and tp_bases */
type->tp_bases = bases;
bases = NULL;
Py_INCREF(base);
type->tp_base = base;

type->tp_basicsize = spec->basicsize;
type->tp_itemsize = spec->itemsize;

for (slot = spec->slots; slot->slot; slot++) {
if (slot->slot < 0
|| (size_t)slot->slot >= Py_ARRAY_LENGTH(slotoffsets)) {
PyErr_SetString(PyExc_RuntimeError, "invalid slot offset");
goto fail;
}
else if (slot->slot == Py_tp_base || slot->slot == Py_tp_bases) {
/* Processed above */
continue;
}
else if (slot->slot == Py_tp_doc) {
/* For the docstring slot, which usually points to a static string
   literal, we need to make a copy */
const char *old_doc = _PyType_DocWithoutSignature(type->tp_name, 
slot->pfunc);
size_t len = strlen(old_doc)+1;
char *tp_doc = PyObject_MALLOC(len);
if (tp_doc == NULL) {
type->tp_doc = NULL;
PyErr_NoMemory();
goto fail;
}
memcpy(tp_doc, old_doc, len);
type->tp_doc = tp_doc;
}
else if (slot->slot == Py_tp_members) {
/* Move the slots to the heap type itself */
size_t len = Py_TYPE(type)->tp_itemsize * nmembers;
memcpy(PyHeapType_GET_MEMBERS(res), slot->pfunc, len);
type->tp_members = PyHeapType_GET_MEMBERS(res);
}
else {
/* Copy other slots directly */
*(void**)(res_start + slotoffsets[slot->slot]) = slot->pfunc;
}
}
if (type->tp_dealloc == NULL) {
/* It's a heap type, so needs the heap types' dealloc.
   subtype_dealloc will call the base type's tp_dealloc, if
   necessary. */
type->tp_dealloc = subtype_dealloc;
}

if (PyType_Ready(type) < 0)
goto fail;

if (type->tp_dictoffset) {
res->ht_cached_keys = _PyDict_NewKeysForClass();
}

/* Set type.__module__ */
s = strrchr(spec->name, '.');
if (s != NULL) {
int err;
modname = PyUnicode_FromStringAndSize(
spec->name, (Py_ssize_t)(s - spec->name));
if (modname == NULL) {
goto fail;

[issue15729] PyStructSequence_NewType enhancement

[Mark Lawrence]

Changes by Mark Lawrence :


--
nosy:  -BreamoreBoy

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue15729] PyStructSequence_NewType enhancement

[elemental]

Changes by elemental :


--
nosy: +elemental

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue15729] PyStructSequence_NewType enhancement

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

See also issue20066.

--
nosy: +serhiy.storchaka

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue15729
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue15729] PyStructSequence_NewType enhancement

[Mark Lawrence]

Mark Lawrence added the comment:

I'm uncertain as to whether or not the patch can be considered stand alone, or 
whether the pep3121 keyword is also relevant here.

--
nosy: +BreamoreBoy
versions: +Python 3.5 -Python 3.4

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue15729
___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue15729] PyStructSequence_NewType enhancement

[Eric Snow]

Changes by Eric Snow ericsnowcurren...@gmail.com:


--
nosy: +eric.snow

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue15729
___
___
Python-bugs-list mailing list
Unsubscribe: 
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue15729] PyStructSequence_NewType enhancement

[Robin Schreiber]

Changes by Robin Schreiber robin.schrei...@me.com:


--
keywords: +pep3121 -patch

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue15729
___
___
Python-bugs-list mailing list
Unsubscribe: 
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue15729] PyStructSequence_NewType enhancement

[Robin Schreiber]

New submission from Robin Schreiber:

To create a HeapType from Structseq description, there is the helpful, yet 
undocumented PyStructSequence_NewType Method, which can do just that.
Until now, this method solely allocates some generic TypeObject on which it 
then performs PyStructSequence_InitType().

I have found that this is far from enough, as for one the flags of the type are 
not set appropriately and neither ht_name nor ht_qualname are set (which is as 
far as I know required of a proper Heaptype).
I have now added this missing initialization of the fields to 
PyStructSequence_NewType(). 

I have also changed the previous definition of PyStructSequence_InitType, by 
extracting a new Method _PyStructSequence_InitTypeWithFlags. This initializes 
the given type with the flags variable passed to the function.
This method extraction is needed, as we can not alter the semantics of InitType 
itself, yet need some way to initialize a SequenceType with Heaptype-flags, 
without having too much duplicate code.

--
components: Interpreter Core
files: structseq_newtype_fix.patch
keywords: patch
messages: 168595
nosy: Robin.Schreiber, loewis
priority: normal
severity: normal
status: open
title: PyStructSequence_NewType enhancement
type: enhancement
versions: Python 3.4
Added file: http://bugs.python.org/file26901/structseq_newtype_fix.patch

___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue15729
___
___
Python-bugs-list mailing list
Unsubscribe: 
http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com