static inline bool
check_length_limit(Py_ssize_t len, long limit) {
if (!check_ssize_t_32(len)) {
return false;
}
if (len > limit) {
PyErr_Format(PyExc_OverflowError, "size exceeded specified limit: %d", limit);
return false;
}
return true;
}
static bool
skip(DecodeBuffer* input, TType type) {
#define SKIPBYTES(n) \
do { \
if (!readBytes(input, &dummy_buf, (n))) { \
return false; \
} \
} while(0)
char* dummy_buf;
switch (type) {
case T_BOOL:
case T_I08:
SKIPBYTES(1);
break;
case T_I16:
SKIPBYTES(2);
break;
case T_I32:
SKIPBYTES(4);
break;
case T_I64:
case T_DOUBLE:
SKIPBYTES(8);
break;
case T_STRING: {
// TODO(dreiss): Find out if these check_ssize_t32s are really necessary.
int len = readI32(input);
if (!check_ssize_t_32(len)) {
return false;
}
SKIPBYTES(len);
break;
}
case T_LIST:
case T_SET: {
TType etype;
int len, i;
etype = readByte(input);
if (etype == -1) {
return false;
}
len = readI32(input);
if (!check_ssize_t_32(len)) {
return false;
}
for (i = 0; i < len; i++) {
if (!skip(input, etype)) {
return false;
}
}
break;
}
case T_MAP: {
TType ktype, vtype;
int len, i;
ktype = readByte(input);
if (ktype == -1) {
return false;
}
vtype = readByte(input);
if (vtype == -1) {
return false;
}
len = readI32(input);
if (!check_ssize_t_32(len)) {
return false;
}
for (i = 0; i < len; i++) {
if (!(skip(input, ktype) && skip(input, vtype))) {
return false;
}
}
break;
}
case T_STRUCT: {
while (true) {
TType type;
type = readByte(input);
if (type == -1) {
return false;
}
if (type == T_STOP)
break;
SKIPBYTES(2); // tag
if (!skip(input, type)) {
return false;
}
}
break;
}
case T_STOP:
case T_VOID:
case T_UTF16:
case T_UTF8:
case T_U64:
default:
PyErr_SetString(PyExc_TypeError, "Unexpected TType");
return false;
}
return true;
#undef SKIPBYTES
}
static bool
output_val(PyObject* output, PyObject* value, TType type, PyObject* typeargs) {
/*
* Refcounting Strategy:
*
* We assume that elements of the thrift_spec tuple are not going to be
* mutated, so we don't ref count those at all. Other than that, we try to
* keep a reference to all the user-created objects while we work with them.
* output_val assumes that a reference is already held. The *caller* is
* responsible for handling references
*/
switch (type) {
case T_BOOL: {
int v = PyObject_IsTrue(value);
if (v == -1) {
return false;
}
writeByte(output, (int8_t) v);
break;
}
case T_I08: {
int32_t val;
if (!parse_pyint(value, &val, INT8_MIN, INT8_MAX)) {
return false;
}
writeByte(output, (int8_t) val);
break;
}
case T_I16: {
int32_t val;
if (!parse_pyint(value, &val, INT16_MIN, INT16_MAX)) {
return false;
}
writeI16(output, (int16_t) val);
break;
}
case T_I32: {
int32_t val;
if (!parse_pyint(value, &val, INT32_MIN, INT32_MAX)) {
return false;
}
writeI32(output, val);
break;
}
case T_I64: {
int64_t nval = PyLong_AsLongLong(value);
if (INT_CONV_ERROR_OCCURRED(nval)) {
return false;
}
if (!CHECK_RANGE(nval, INT64_MIN, INT64_MAX)) {
PyErr_SetString(PyExc_OverflowError, "int out of range");
return false;
}
writeI64(output, nval);
break;
}
case T_DOUBLE: {
double nval = PyFloat_AsDouble(value);
if (nval == -1.0 && PyErr_Occurred()) {
return false;
}
writeDouble(output, nval);
break;
}
case T_STRING: {
Py_ssize_t len = 0;
if (is_utf8(typeargs) && PyUnicode_Check(value))
value = PyUnicode_AsUTF8String(value);
len = PyString_Size(value);
if (!check_ssize_t_32(len)) {
return false;
}
writeI32(output, (int32_t) len);
PycStringIO->cwrite(output, PyString_AsString(value), (int32_t) len);
break;
}
case T_LIST:
case T_SET: {
Py_ssize_t len;
SetListTypeArgs parsedargs;
PyObject *item;
PyObject *iterator;
if (!parse_set_list_args(&parsedargs, typeargs)) {
return false;
}
len = PyObject_Length(value);
if (!check_ssize_t_32(len)) {
return false;
}
writeByte(output, parsedargs.element_type);
writeI32(output, (int32_t) len);
iterator = PyObject_GetIter(value);
if (iterator == NULL) {
return false;
}
while ((item = PyIter_Next(iterator))) {
if (!output_val(output, item, parsedargs.element_type, parsedargs.typeargs)) {
Py_DECREF(item);
Py_DECREF(iterator);
return false;
}
Py_DECREF(item);
}
Py_DECREF(iterator);
if (PyErr_Occurred()) {
return false;
}
break;
}
case T_MAP: {
PyObject *k, *v;
Py_ssize_t pos = 0;
Py_ssize_t len;
MapTypeArgs parsedargs;
len = PyDict_Size(value);
if (!check_ssize_t_32(len)) {
return false;
}
if (!parse_map_args(&parsedargs, typeargs)) {
return false;
}
writeByte(output, parsedargs.ktag);
writeByte(output, parsedargs.vtag);
writeI32(output, len);
// TODO(bmaurer): should support any mapping, not just dicts
while (PyDict_Next(value, &pos, &k, &v)) {
// TODO(dreiss): Think hard about whether these INCREFs actually
// turn any unsafe scenarios into safe scenarios.
Py_INCREF(k);
Py_INCREF(v);
if (!output_val(output, k, parsedargs.ktag, parsedargs.ktypeargs)
|| !output_val(output, v, parsedargs.vtag, parsedargs.vtypeargs)) {
Py_DECREF(k);
Py_DECREF(v);
return false;
}
Py_DECREF(k);
Py_DECREF(v);
}
break;
}
// TODO(dreiss): Consider breaking this out as a function
// the way we did for decode_struct.
case T_STRUCT: {
StructTypeArgs parsedargs;
Py_ssize_t nspec;
Py_ssize_t i;
if (!parse_struct_args(&parsedargs, typeargs)) {
return false;
}
nspec = PyTuple_Size(parsedargs.spec);
if (nspec == -1) {
return false;
}
for (i = 0; i < nspec; i++) {
StructItemSpec parsedspec;
PyObject* spec_tuple;
PyObject* instval = NULL;
spec_tuple = PyTuple_GET_ITEM(parsedargs.spec, i);
if (spec_tuple == Py_None) {
continue;
}
if (!parse_struct_item_spec (&parsedspec, spec_tuple)) {
return false;
}
instval = PyObject_GetAttr(value, parsedspec.attrname);
if (!instval) {
return false;
}
if (instval == Py_None) {
Py_DECREF(instval);
continue;
}
writeByte(output, (int8_t) parsedspec.type);
writeI16(output, parsedspec.tag);
if (!output_val(output, instval, parsedspec.type, parsedspec.typeargs)) {
Py_DECREF(instval);
return false;
}
Py_DECREF(instval);
}
writeByte(output, (int8_t)T_STOP);
break;
}
case T_STOP:
case T_VOID:
case T_UTF16:
case T_UTF8:
case T_U64:
default:
PyErr_SetString(PyExc_TypeError, "Unexpected TType");
return false;
}
return true;
}
// Returns a new reference.
static PyObject*
decode_val(DecodeBuffer* input, TType type, PyObject* typeargs) {
switch (type) {
case T_BOOL: {
int8_t v = readByte(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
switch (v) {
case 0: Py_RETURN_FALSE;
case 1: Py_RETURN_TRUE;
// Don't laugh. This is a potentially serious issue.
default: PyErr_SetString(PyExc_TypeError, "boolean out of range"); return NULL;
}
break;
}
case T_I08: {
int8_t v = readByte(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
return PyInt_FromLong(v);
}
case T_I16: {
int16_t v = readI16(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
return PyInt_FromLong(v);
}
case T_I32: {
int32_t v = readI32(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
return PyInt_FromLong(v);
}
case T_I64: {
int64_t v = readI64(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
// TODO(dreiss): Find out if we can take this fastpath always when
// sizeof(long) == sizeof(long long).
if (CHECK_RANGE(v, LONG_MIN, LONG_MAX)) {
return PyInt_FromLong((long) v);
}
return PyLong_FromLongLong(v);
}
case T_DOUBLE: {
double v = readDouble(input);
if (v == -1.0 && PyErr_Occurred()) {
return false;
}
return PyFloat_FromDouble(v);
}
case T_STRING: {
Py_ssize_t len = readI32(input);
char* buf;
if (!readBytes(input, &buf, len)) {
return NULL;
}
return PyString_FromStringAndSize(buf, len);
}
case T_LIST:
case T_SET: {
SetListTypeArgs parsedargs;
int32_t len;
PyObject* ret = NULL;
int i;
if (!parse_set_list_args(&parsedargs, typeargs)) {
return NULL;
}
if (!checkTypeByte(input, parsedargs.element_type)) {
return NULL;
}
len = readI32(input);
if (!check_ssize_t_32(len)) {
return NULL;
}
ret = PyList_New(len);
if (!ret) {
return NULL;
}
for (i = 0; i < len; i++) {
PyObject* item = decode_val(input, parsedargs.element_type, parsedargs.typeargs);
if (!item) {
Py_DECREF(ret);
return NULL;
}
PyList_SET_ITEM(ret, i, item);
}
// TODO(dreiss): Consider biting the bullet and making two separate cases
// for list and set, avoiding this post facto conversion.
if (type == T_SET) {
PyObject* setret;
#if (PY_VERSION_HEX < 0x02050000)
// hack needed for older versions
setret = PyObject_CallFunctionObjArgs((PyObject*)&PySet_Type, ret, NULL);
#else
// official version
setret = PySet_New(ret);
#endif
Py_DECREF(ret);
return setret;
}
return ret;
}
case T_MAP: {
int32_t len;
int i;
MapTypeArgs parsedargs;
PyObject* ret = NULL;
if (!parse_map_args(&parsedargs, typeargs)) {
return NULL;
}
if (!checkTypeByte(input, parsedargs.ktag)) {
return NULL;
}
if (!checkTypeByte(input, parsedargs.vtag)) {
return NULL;
}
len = readI32(input);
if (!check_ssize_t_32(len)) {
return false;
}
ret = PyDict_New();
if (!ret) {
goto error;
}
for (i = 0; i < len; i++) {
PyObject* k = NULL;
PyObject* v = NULL;
k = decode_val(input, parsedargs.ktag, parsedargs.ktypeargs);
if (k == NULL) {
goto loop_error;
}
v = decode_val(input, parsedargs.vtag, parsedargs.vtypeargs);
if (v == NULL) {
goto loop_error;
}
if (PyDict_SetItem(ret, k, v) == -1) {
goto loop_error;
}
Py_DECREF(k);
Py_DECREF(v);
continue;
// Yuck! Destructors, anyone?
loop_error:
Py_XDECREF(k);
Py_XDECREF(v);
goto error;
}
return ret;
error:
Py_XDECREF(ret);
return NULL;
}
case T_STRUCT: {
StructTypeArgs parsedargs;
if (!parse_struct_args(&parsedargs, typeargs)) {
return NULL;
}
PyObject* ret = PyObject_CallObject(parsedargs.klass, NULL);
if (!ret) {
return NULL;
}
if (!decode_struct(input, ret, parsedargs.spec)) {
Py_DECREF(ret);
return NULL;
}
return ret;
}
case T_STOP:
case T_VOID:
case T_UTF16:
case T_UTF8:
case T_U64:
default:
PyErr_SetString(PyExc_TypeError, "Unexpected TType");
return NULL;
}
}
// Returns a new reference.
static PyObject*
decode_val(DecodeBuffer* input, TType type, PyObject* typeargs) {
switch (type) {
case T_BOOL: {
int8_t v = readByte(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
switch (v) {
case 0: Py_RETURN_FALSE;
case 1: Py_RETURN_TRUE;
// Don't laugh. This is a potentially serious issue.
default: PyErr_SetString(PyExc_TypeError, "boolean out of range"); return NULL;
}
break;
}
case T_I08: {
int8_t v = readByte(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
return PyInt_FromLong(v);
}
case T_I16: {
int16_t v = readI16(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
return PyInt_FromLong(v);
}
case T_I32: {
int32_t v = readI32(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
return PyInt_FromLong(v);
}
case T_I64: {
int64_t v = readI64(input);
if (INT_CONV_ERROR_OCCURRED(v)) {
return NULL;
}
// TODO(dreiss): Find out if we can take this fastpath always when
// sizeof(long) == sizeof(long long).
if (CHECK_RANGE(v, LONG_MIN, LONG_MAX)) {
return PyInt_FromLong((long) v);
}
return PyLong_FromLongLong(v);
}
case T_DOUBLE: {
double v = readDouble(input);
if (v == -1.0 && PyErr_Occurred()) {
return false;
}
return PyFloat_FromDouble(v);
}
case T_STRING: {
Py_ssize_t len = readI32(input);
char* buf;
if (!readBytes(input, &buf, len)) {
return NULL;
}
if (is_utf8(typeargs))
return PyUnicode_DecodeUTF8(buf, len, 0);
else
return PyString_FromStringAndSize(buf, len);
}
case T_LIST:
case T_SET: {
SetListTypeArgs parsedargs;
int32_t len;
PyObject* ret = NULL;
int i;
bool use_tuple = false;
if (!parse_set_list_args(&parsedargs, typeargs)) {
return NULL;
}
if (!checkTypeByte(input, parsedargs.element_type)) {
return NULL;
}
len = readI32(input);
if (!check_list_length(len)) {
return NULL;
}
use_tuple = type == T_LIST && parsedargs.immutable;
ret = use_tuple ? PyTuple_New(len) : PyList_New(len);
if (!ret) {
return NULL;
}
for (i = 0; i < len; i++) {
PyObject* item = decode_val(input, parsedargs.element_type, parsedargs.typeargs);
if (!item) {
Py_DECREF(ret);
return NULL;
}
if (use_tuple) {
PyTuple_SET_ITEM(ret, i, item);
} else {
PyList_SET_ITEM(ret, i, item);
}
}
// TODO(dreiss): Consider biting the bullet and making two separate cases
// for list and set, avoiding this post facto conversion.
if (type == T_SET) {
PyObject* setret;
setret = parsedargs.immutable ? PyFrozenSet_New(ret) : PySet_New(ret);
Py_DECREF(ret);
return setret;
}
return ret;
}
case T_MAP: {
int32_t len;
int i;
MapTypeArgs parsedargs;
PyObject* ret = NULL;
if (!parse_map_args(&parsedargs, typeargs)) {
return NULL;
}
if (!checkTypeByte(input, parsedargs.ktag)) {
return NULL;
}
if (!checkTypeByte(input, parsedargs.vtag)) {
return NULL;
}
len = readI32(input);
if (!check_ssize_t_32(len)) {
return false;
}
ret = PyDict_New();
if (!ret) {
goto error;
}
for (i = 0; i < len; i++) {
PyObject* k = NULL;
PyObject* v = NULL;
k = decode_val(input, parsedargs.ktag, parsedargs.ktypeargs);
if (k == NULL) {
goto loop_error;
}
v = decode_val(input, parsedargs.vtag, parsedargs.vtypeargs);
if (v == NULL) {
goto loop_error;
}
if (PyDict_SetItem(ret, k, v) == -1) {
goto loop_error;
}
Py_DECREF(k);
Py_DECREF(v);
continue;
// Yuck! Destructors, anyone?
loop_error:
Py_XDECREF(k);
Py_XDECREF(v);
goto error;
}
if (parsedargs.immutable) {
PyObject* thrift = PyImport_ImportModule("thrift.Thrift");
PyObject* cls = NULL;
PyObject* arg = NULL;
if (!thrift) {
goto error;
}
cls = PyObject_GetAttrString(thrift, "TFrozenDict");
if (!cls) {
goto error;
}
arg = PyTuple_New(1);
PyTuple_SET_ITEM(arg, 0, ret);
return PyObject_CallObject(cls, arg);
}
return ret;
error:
Py_XDECREF(ret);
return NULL;
}
case T_STRUCT: {
StructTypeArgs parsedargs;
if (!parse_struct_args(&parsedargs, typeargs)) {
return NULL;
}
return decode_struct(input, Py_None, parsedargs.klass, parsedargs.spec);
}
case T_STOP:
case T_VOID:
case T_UTF16:
case T_UTF8:
case T_U64:
default:
PyErr_SetString(PyExc_TypeError, "Unexpected TType");
return NULL;
}
}
