diff options
author | mkaay <mkaay@mkaay.de> | 2011-02-07 00:32:21 +0100 |
---|---|---|
committer | mkaay <mkaay@mkaay.de> | 2011-02-07 00:32:21 +0100 |
commit | 69f2d3660e44bd138d24b6cf97b234bc2efc1c78 (patch) | |
tree | af276b8ed0b5cac5a62e6949a00d607afcc94e5d /module/lib/thrift/protocol | |
parent | closed #231 (diff) | |
download | pyload-69f2d3660e44bd138d24b6cf97b234bc2efc1c78.tar.xz |
added Thrift backend
Diffstat (limited to 'module/lib/thrift/protocol')
-rw-r--r-- | module/lib/thrift/protocol/TBinaryProtocol.py | 259 | ||||
-rw-r--r-- | module/lib/thrift/protocol/TCompactProtocol.py | 368 | ||||
-rw-r--r-- | module/lib/thrift/protocol/TProtocol.py | 205 | ||||
-rw-r--r-- | module/lib/thrift/protocol/__init__.py | 20 | ||||
-rw-r--r-- | module/lib/thrift/protocol/fastbinary.c | 1203 |
5 files changed, 2055 insertions, 0 deletions
diff --git a/module/lib/thrift/protocol/TBinaryProtocol.py b/module/lib/thrift/protocol/TBinaryProtocol.py new file mode 100644 index 000000000..50c6aa896 --- /dev/null +++ b/module/lib/thrift/protocol/TBinaryProtocol.py @@ -0,0 +1,259 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +from TProtocol import * +from struct import pack, unpack + +class TBinaryProtocol(TProtocolBase): + + """Binary implementation of the Thrift protocol driver.""" + + # NastyHaxx. Python 2.4+ on 32-bit machines forces hex constants to be + # positive, converting this into a long. If we hardcode the int value + # instead it'll stay in 32 bit-land. + + # VERSION_MASK = 0xffff0000 + VERSION_MASK = -65536 + + # VERSION_1 = 0x80010000 + VERSION_1 = -2147418112 + + TYPE_MASK = 0x000000ff + + def __init__(self, trans, strictRead=False, strictWrite=True): + TProtocolBase.__init__(self, trans) + self.strictRead = strictRead + self.strictWrite = strictWrite + + def writeMessageBegin(self, name, type, seqid): + if self.strictWrite: + self.writeI32(TBinaryProtocol.VERSION_1 | type) + self.writeString(name) + self.writeI32(seqid) + else: + self.writeString(name) + self.writeByte(type) + self.writeI32(seqid) + + def writeMessageEnd(self): + pass + + def writeStructBegin(self, name): + pass + + def writeStructEnd(self): + pass + + def writeFieldBegin(self, name, type, id): + self.writeByte(type) + self.writeI16(id) + + def writeFieldEnd(self): + pass + + def writeFieldStop(self): + self.writeByte(TType.STOP); + + def writeMapBegin(self, ktype, vtype, size): + self.writeByte(ktype) + self.writeByte(vtype) + self.writeI32(size) + + def writeMapEnd(self): + pass + + def writeListBegin(self, etype, size): + self.writeByte(etype) + self.writeI32(size) + + def writeListEnd(self): + pass + + def writeSetBegin(self, etype, size): + self.writeByte(etype) + self.writeI32(size) + + def writeSetEnd(self): + pass + + def writeBool(self, bool): + if bool: + self.writeByte(1) + else: + self.writeByte(0) + + def writeByte(self, byte): + buff = pack("!b", byte) + self.trans.write(buff) + + def writeI16(self, i16): + buff = pack("!h", i16) + self.trans.write(buff) + + def writeI32(self, i32): + buff = pack("!i", i32) + self.trans.write(buff) + + def writeI64(self, i64): + buff = pack("!q", i64) + self.trans.write(buff) + + def writeDouble(self, dub): + buff = pack("!d", dub) + self.trans.write(buff) + + def writeString(self, str): + self.writeI32(len(str)) + self.trans.write(str) + + def readMessageBegin(self): + sz = self.readI32() + if sz < 0: + version = sz & TBinaryProtocol.VERSION_MASK + if version != TBinaryProtocol.VERSION_1: + raise TProtocolException(type=TProtocolException.BAD_VERSION, message='Bad version in readMessageBegin: %d' % (sz)) + type = sz & TBinaryProtocol.TYPE_MASK + name = self.readString() + seqid = self.readI32() + else: + if self.strictRead: + raise TProtocolException(type=TProtocolException.BAD_VERSION, message='No protocol version header') + name = self.trans.readAll(sz) + type = self.readByte() + seqid = self.readI32() + return (name, type, seqid) + + def readMessageEnd(self): + pass + + def readStructBegin(self): + pass + + def readStructEnd(self): + pass + + def readFieldBegin(self): + type = self.readByte() + if type == TType.STOP: + return (None, type, 0) + id = self.readI16() + return (None, type, id) + + def readFieldEnd(self): + pass + + def readMapBegin(self): + ktype = self.readByte() + vtype = self.readByte() + size = self.readI32() + return (ktype, vtype, size) + + def readMapEnd(self): + pass + + def readListBegin(self): + etype = self.readByte() + size = self.readI32() + return (etype, size) + + def readListEnd(self): + pass + + def readSetBegin(self): + etype = self.readByte() + size = self.readI32() + return (etype, size) + + def readSetEnd(self): + pass + + def readBool(self): + byte = self.readByte() + if byte == 0: + return False + return True + + def readByte(self): + buff = self.trans.readAll(1) + val, = unpack('!b', buff) + return val + + def readI16(self): + buff = self.trans.readAll(2) + val, = unpack('!h', buff) + return val + + def readI32(self): + buff = self.trans.readAll(4) + val, = unpack('!i', buff) + return val + + def readI64(self): + buff = self.trans.readAll(8) + val, = unpack('!q', buff) + return val + + def readDouble(self): + buff = self.trans.readAll(8) + val, = unpack('!d', buff) + return val + + def readString(self): + len = self.readI32() + str = self.trans.readAll(len) + return str + + +class TBinaryProtocolFactory: + def __init__(self, strictRead=False, strictWrite=True): + self.strictRead = strictRead + self.strictWrite = strictWrite + + def getProtocol(self, trans): + prot = TBinaryProtocol(trans, self.strictRead, self.strictWrite) + return prot + + +class TBinaryProtocolAccelerated(TBinaryProtocol): + + """C-Accelerated version of TBinaryProtocol. + + This class does not override any of TBinaryProtocol's methods, + but the generated code recognizes it directly and will call into + our C module to do the encoding, bypassing this object entirely. + We inherit from TBinaryProtocol so that the normal TBinaryProtocol + encoding can happen if the fastbinary module doesn't work for some + reason. (TODO(dreiss): Make this happen sanely in more cases.) + + In order to take advantage of the C module, just use + TBinaryProtocolAccelerated instead of TBinaryProtocol. + + NOTE: This code was contributed by an external developer. + The internal Thrift team has reviewed and tested it, + but we cannot guarantee that it is production-ready. + Please feel free to report bugs and/or success stories + to the public mailing list. + """ + + pass + + +class TBinaryProtocolAcceleratedFactory: + def getProtocol(self, trans): + return TBinaryProtocolAccelerated(trans) diff --git a/module/lib/thrift/protocol/TCompactProtocol.py b/module/lib/thrift/protocol/TCompactProtocol.py new file mode 100644 index 000000000..fbc156a8f --- /dev/null +++ b/module/lib/thrift/protocol/TCompactProtocol.py @@ -0,0 +1,368 @@ +from TProtocol import * +from struct import pack, unpack + +__all__ = ['TCompactProtocol', 'TCompactProtocolFactory'] + +CLEAR = 0 +FIELD_WRITE = 1 +VALUE_WRITE = 2 +CONTAINER_WRITE = 3 +BOOL_WRITE = 4 +FIELD_READ = 5 +CONTAINER_READ = 6 +VALUE_READ = 7 +BOOL_READ = 8 + +def make_helper(v_from, container): + def helper(func): + def nested(self, *args, **kwargs): + assert self.state in (v_from, container), (self.state, v_from, container) + return func(self, *args, **kwargs) + return nested + return helper +writer = make_helper(VALUE_WRITE, CONTAINER_WRITE) +reader = make_helper(VALUE_READ, CONTAINER_READ) + +def makeZigZag(n, bits): + return (n << 1) ^ (n >> (bits - 1)) + +def fromZigZag(n): + return (n >> 1) ^ -(n & 1) + +def writeVarint(trans, n): + out = [] + while True: + if n & ~0x7f == 0: + out.append(n) + break + else: + out.append((n & 0xff) | 0x80) + n = n >> 7 + trans.write(''.join(map(chr, out))) + +def readVarint(trans): + result = 0 + shift = 0 + while True: + x = trans.readAll(1) + byte = ord(x) + result |= (byte & 0x7f) << shift + if byte >> 7 == 0: + return result + shift += 7 + +class CompactType: + TRUE = 1 + FALSE = 2 + BYTE = 0x03 + I16 = 0x04 + I32 = 0x05 + I64 = 0x06 + DOUBLE = 0x07 + BINARY = 0x08 + LIST = 0x09 + SET = 0x0A + MAP = 0x0B + STRUCT = 0x0C + +CTYPES = {TType.BOOL: CompactType.TRUE, # used for collection + TType.BYTE: CompactType.BYTE, + TType.I16: CompactType.I16, + TType.I32: CompactType.I32, + TType.I64: CompactType.I64, + TType.DOUBLE: CompactType.DOUBLE, + TType.STRING: CompactType.BINARY, + TType.STRUCT: CompactType.STRUCT, + TType.LIST: CompactType.LIST, + TType.SET: CompactType.SET, + TType.MAP: CompactType.MAP, + } + +TTYPES = {} +for k, v in CTYPES.items(): + TTYPES[v] = k +TTYPES[CompactType.FALSE] = TType.BOOL +del k +del v + +class TCompactProtocol(TProtocolBase): + "Compact implementation of the Thrift protocol driver." + + PROTOCOL_ID = 0x82 + VERSION = 1 + VERSION_MASK = 0x1f + TYPE_MASK = 0xe0 + TYPE_SHIFT_AMOUNT = 5 + + def __init__(self, trans): + TProtocolBase.__init__(self, trans) + self.state = CLEAR + self.__last_fid = 0 + self.__bool_fid = None + self.__bool_value = None + self.__structs = [] + self.__containers = [] + + def __writeVarint(self, n): + writeVarint(self.trans, n) + + def writeMessageBegin(self, name, type, seqid): + assert self.state == CLEAR + self.__writeUByte(self.PROTOCOL_ID) + self.__writeUByte(self.VERSION | (type << self.TYPE_SHIFT_AMOUNT)) + self.__writeVarint(seqid) + self.__writeString(name) + self.state = VALUE_WRITE + + def writeMessageEnd(self): + assert self.state == VALUE_WRITE + self.state = CLEAR + + def writeStructBegin(self, name): + assert self.state in (CLEAR, CONTAINER_WRITE, VALUE_WRITE), self.state + self.__structs.append((self.state, self.__last_fid)) + self.state = FIELD_WRITE + self.__last_fid = 0 + + def writeStructEnd(self): + assert self.state == FIELD_WRITE + self.state, self.__last_fid = self.__structs.pop() + + def writeFieldStop(self): + self.__writeByte(0) + + def __writeFieldHeader(self, type, fid): + delta = fid - self.__last_fid + if 0 < delta <= 15: + self.__writeUByte(delta << 4 | type) + else: + self.__writeByte(type) + self.__writeI16(fid) + self.__last_fid = fid + + def writeFieldBegin(self, name, type, fid): + assert self.state == FIELD_WRITE, self.state + if type == TType.BOOL: + self.state = BOOL_WRITE + self.__bool_fid = fid + else: + self.state = VALUE_WRITE + self.__writeFieldHeader(CTYPES[type], fid) + + def writeFieldEnd(self): + assert self.state in (VALUE_WRITE, BOOL_WRITE), self.state + self.state = FIELD_WRITE + + def __writeUByte(self, byte): + self.trans.write(pack('!B', byte)) + + def __writeByte(self, byte): + self.trans.write(pack('!b', byte)) + + def __writeI16(self, i16): + self.__writeVarint(makeZigZag(i16, 16)) + + def __writeSize(self, i32): + self.__writeVarint(i32) + + def writeCollectionBegin(self, etype, size): + assert self.state in (VALUE_WRITE, CONTAINER_WRITE), self.state + if size <= 14: + self.__writeUByte(size << 4 | CTYPES[etype]) + else: + self.__writeUByte(0xf0 | CTYPES[etype]) + self.__writeSize(size) + self.__containers.append(self.state) + self.state = CONTAINER_WRITE + writeSetBegin = writeCollectionBegin + writeListBegin = writeCollectionBegin + + def writeMapBegin(self, ktype, vtype, size): + assert self.state in (VALUE_WRITE, CONTAINER_WRITE), self.state + if size == 0: + self.__writeByte(0) + else: + self.__writeSize(size) + self.__writeUByte(CTYPES[ktype] << 4 | CTYPES[vtype]) + self.__containers.append(self.state) + self.state = CONTAINER_WRITE + + def writeCollectionEnd(self): + assert self.state == CONTAINER_WRITE, self.state + self.state = self.__containers.pop() + writeMapEnd = writeCollectionEnd + writeSetEnd = writeCollectionEnd + writeListEnd = writeCollectionEnd + + def writeBool(self, bool): + if self.state == BOOL_WRITE: + self.__writeFieldHeader(types[bool], self.__bool_fid) + elif self.state == CONTAINER_WRITE: + self.__writeByte(int(bool)) + else: + raise AssertetionError, "Invalid state in compact protocol" + + writeByte = writer(__writeByte) + writeI16 = writer(__writeI16) + + @writer + def writeI32(self, i32): + self.__writeVarint(makeZigZag(i32, 32)) + + @writer + def writeI64(self, i64): + self.__writeVarint(makeZigZag(i64, 64)) + + @writer + def writeDouble(self, dub): + self.trans.write(pack('!d', dub)) + + def __writeString(self, s): + self.__writeSize(len(s)) + self.trans.write(s) + writeString = writer(__writeString) + + def readFieldBegin(self): + assert self.state == FIELD_READ, self.state + type = self.__readUByte() + if type & 0x0f == TType.STOP: + return (None, 0, 0) + delta = type >> 4 + if delta == 0: + fid = self.__readI16() + else: + fid = self.__last_fid + delta + self.__last_fid = fid + type = type & 0x0f + if type == CompactType.TRUE: + self.state = BOOL_READ + self.__bool_value = True + elif type == CompactType.FALSE: + self.state = BOOL_READ + self.__bool_value = False + else: + self.state = VALUE_READ + return (None, self.__getTType(type), fid) + + def readFieldEnd(self): + assert self.state in (VALUE_READ, BOOL_READ), self.state + self.state = FIELD_READ + + def __readUByte(self): + result, = unpack('!B', self.trans.readAll(1)) + return result + + def __readByte(self): + result, = unpack('!b', self.trans.readAll(1)) + return result + + def __readVarint(self): + return readVarint(self.trans) + + def __readZigZag(self): + return fromZigZag(self.__readVarint()) + + def __readSize(self): + result = self.__readVarint() + if result < 0: + raise TException("Length < 0") + return result + + def readMessageBegin(self): + assert self.state == CLEAR + proto_id = self.__readUByte() + if proto_id != self.PROTOCOL_ID: + raise TProtocolException(TProtocolException.BAD_VERSION, + 'Bad protocol id in the message: %d' % proto_id) + ver_type = self.__readUByte() + type = (ver_type & self.TYPE_MASK) >> self.TYPE_SHIFT_AMOUNT + version = ver_type & self.VERSION_MASK + if version != self.VERSION: + raise TProtocolException(TProtocolException.BAD_VERSION, + 'Bad version: %d (expect %d)' % (version, self.VERSION)) + seqid = self.__readVarint() + name = self.__readString() + return (name, type, seqid) + + def readMessageEnd(self): + assert self.state == VALUE_READ + assert len(self.__structs) == 0 + self.state = CLEAR + + def readStructBegin(self): + assert self.state in (CLEAR, CONTAINER_READ, VALUE_READ), self.state + self.__structs.append((self.state, self.__last_fid)) + self.state = FIELD_READ + self.__last_fid = 0 + + def readStructEnd(self): + assert self.state == FIELD_READ + self.state, self.__last_fid = self.__structs.pop() + + def readCollectionBegin(self): + assert self.state in (VALUE_READ, CONTAINER_READ), self.state + size_type = self.__readUByte() + size = size_type >> 4 + type = self.__getTType(size_type) + if size == 15: + size = self.__readSize() + self.__containers.append(self.state) + self.state = CONTAINER_READ + return type, size + readSetBegin = readCollectionBegin + readListBegin = readCollectionBegin + + def readMapBegin(self): + assert self.state in (VALUE_READ, CONTAINER_READ), self.state + size = self.__readSize() + types = 0 + if size > 0: + types = self.__readUByte() + vtype = self.__getTType(types) + ktype = self.__getTType(types >> 4) + self.__containers.append(self.state) + self.state = CONTAINER_READ + return (ktype, vtype, size) + + def readCollectionEnd(self): + assert self.state == CONTAINER_READ, self.state + self.state = self.__containers.pop() + readSetEnd = readCollectionEnd + readListEnd = readCollectionEnd + readMapEnd = readCollectionEnd + + def readBool(self): + if self.state == BOOL_READ: + return self.__bool_value + elif self.state == CONTAINER_READ: + return bool(self.__readByte()) + else: + raise AssertionError, "Invalid state in compact protocol: %d" % self.state + + readByte = reader(__readByte) + __readI16 = __readZigZag + readI16 = reader(__readZigZag) + readI32 = reader(__readZigZag) + readI64 = reader(__readZigZag) + + @reader + def readDouble(self): + buff = self.trans.readAll(8) + val, = unpack('!d', buff) + return val + + def __readString(self): + len = self.__readSize() + return self.trans.readAll(len) + readString = reader(__readString) + + def __getTType(self, byte): + return TTYPES[byte & 0x0f] + + +class TCompactProtocolFactory: + def __init__(self): + pass + + def getProtocol(self, trans): + return TCompactProtocol(trans) diff --git a/module/lib/thrift/protocol/TProtocol.py b/module/lib/thrift/protocol/TProtocol.py new file mode 100644 index 000000000..be3cb1403 --- /dev/null +++ b/module/lib/thrift/protocol/TProtocol.py @@ -0,0 +1,205 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +from thrift.Thrift import * + +class TProtocolException(TException): + + """Custom Protocol Exception class""" + + UNKNOWN = 0 + INVALID_DATA = 1 + NEGATIVE_SIZE = 2 + SIZE_LIMIT = 3 + BAD_VERSION = 4 + + def __init__(self, type=UNKNOWN, message=None): + TException.__init__(self, message) + self.type = type + +class TProtocolBase: + + """Base class for Thrift protocol driver.""" + + def __init__(self, trans): + self.trans = trans + + def writeMessageBegin(self, name, type, seqid): + pass + + def writeMessageEnd(self): + pass + + def writeStructBegin(self, name): + pass + + def writeStructEnd(self): + pass + + def writeFieldBegin(self, name, type, id): + pass + + def writeFieldEnd(self): + pass + + def writeFieldStop(self): + pass + + def writeMapBegin(self, ktype, vtype, size): + pass + + def writeMapEnd(self): + pass + + def writeListBegin(self, etype, size): + pass + + def writeListEnd(self): + pass + + def writeSetBegin(self, etype, size): + pass + + def writeSetEnd(self): + pass + + def writeBool(self, bool): + pass + + def writeByte(self, byte): + pass + + def writeI16(self, i16): + pass + + def writeI32(self, i32): + pass + + def writeI64(self, i64): + pass + + def writeDouble(self, dub): + pass + + def writeString(self, str): + pass + + def readMessageBegin(self): + pass + + def readMessageEnd(self): + pass + + def readStructBegin(self): + pass + + def readStructEnd(self): + pass + + def readFieldBegin(self): + pass + + def readFieldEnd(self): + pass + + def readMapBegin(self): + pass + + def readMapEnd(self): + pass + + def readListBegin(self): + pass + + def readListEnd(self): + pass + + def readSetBegin(self): + pass + + def readSetEnd(self): + pass + + def readBool(self): + pass + + def readByte(self): + pass + + def readI16(self): + pass + + def readI32(self): + pass + + def readI64(self): + pass + + def readDouble(self): + pass + + def readString(self): + pass + + def skip(self, type): + if type == TType.STOP: + return + elif type == TType.BOOL: + self.readBool() + elif type == TType.BYTE: + self.readByte() + elif type == TType.I16: + self.readI16() + elif type == TType.I32: + self.readI32() + elif type == TType.I64: + self.readI64() + elif type == TType.DOUBLE: + self.readDouble() + elif type == TType.STRING: + self.readString() + elif type == TType.STRUCT: + name = self.readStructBegin() + while True: + (name, type, id) = self.readFieldBegin() + if type == TType.STOP: + break + self.skip(type) + self.readFieldEnd() + self.readStructEnd() + elif type == TType.MAP: + (ktype, vtype, size) = self.readMapBegin() + for i in range(size): + self.skip(ktype) + self.skip(vtype) + self.readMapEnd() + elif type == TType.SET: + (etype, size) = self.readSetBegin() + for i in range(size): + self.skip(etype) + self.readSetEnd() + elif type == TType.LIST: + (etype, size) = self.readListBegin() + for i in range(size): + self.skip(etype) + self.readListEnd() + +class TProtocolFactory: + def getProtocol(self, trans): + pass diff --git a/module/lib/thrift/protocol/__init__.py b/module/lib/thrift/protocol/__init__.py new file mode 100644 index 000000000..01bfe18e5 --- /dev/null +++ b/module/lib/thrift/protocol/__init__.py @@ -0,0 +1,20 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +# + +__all__ = ['TProtocol', 'TBinaryProtocol', 'fastbinary'] diff --git a/module/lib/thrift/protocol/fastbinary.c b/module/lib/thrift/protocol/fastbinary.c new file mode 100644 index 000000000..67b215a83 --- /dev/null +++ b/module/lib/thrift/protocol/fastbinary.c @@ -0,0 +1,1203 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +#include <Python.h> +#include "cStringIO.h" +#include <stdbool.h> +#include <stdint.h> +#include <netinet/in.h> + +/* Fix endianness issues on Solaris */ +#if defined (__SVR4) && defined (__sun) + #if defined(__i386) && !defined(__i386__) + #define __i386__ + #endif + + #ifndef BIG_ENDIAN + #define BIG_ENDIAN (4321) + #endif + #ifndef LITTLE_ENDIAN + #define LITTLE_ENDIAN (1234) + #endif + + /* I386 is LE, even on Solaris */ + #if !defined(BYTE_ORDER) && defined(__i386__) + #define BYTE_ORDER LITTLE_ENDIAN + #endif +#endif + +// TODO(dreiss): defval appears to be unused. Look into removing it. +// TODO(dreiss): Make parse_spec_args recursive, and cache the output +// permanently in the object. (Malloc and orphan.) +// TODO(dreiss): Why do we need cStringIO for reading, why not just char*? +// Can cStringIO let us work with a BufferedTransport? +// TODO(dreiss): Don't ignore the rv from cwrite (maybe). + +/* ====== BEGIN UTILITIES ====== */ + +#define INIT_OUTBUF_SIZE 128 + +// Stolen out of TProtocol.h. +// It would be a huge pain to have both get this from one place. +typedef enum TType { + T_STOP = 0, + T_VOID = 1, + T_BOOL = 2, + T_BYTE = 3, + T_I08 = 3, + T_I16 = 6, + T_I32 = 8, + T_U64 = 9, + T_I64 = 10, + T_DOUBLE = 4, + T_STRING = 11, + T_UTF7 = 11, + T_STRUCT = 12, + T_MAP = 13, + T_SET = 14, + T_LIST = 15, + T_UTF8 = 16, + T_UTF16 = 17 +} TType; + +#ifndef __BYTE_ORDER +# if defined(BYTE_ORDER) && defined(LITTLE_ENDIAN) && defined(BIG_ENDIAN) +# define __BYTE_ORDER BYTE_ORDER +# define __LITTLE_ENDIAN LITTLE_ENDIAN +# define __BIG_ENDIAN BIG_ENDIAN +# else +# error "Cannot determine endianness" +# endif +#endif + +// Same comment as the enum. Sorry. +#if __BYTE_ORDER == __BIG_ENDIAN +# define ntohll(n) (n) +# define htonll(n) (n) +#elif __BYTE_ORDER == __LITTLE_ENDIAN +# if defined(__GNUC__) && defined(__GLIBC__) +# include <byteswap.h> +# define ntohll(n) bswap_64(n) +# define htonll(n) bswap_64(n) +# else /* GNUC & GLIBC */ +# define ntohll(n) ( (((unsigned long long)ntohl(n)) << 32) + ntohl(n >> 32) ) +# define htonll(n) ( (((unsigned long long)htonl(n)) << 32) + htonl(n >> 32) ) +# endif /* GNUC & GLIBC */ +#else /* __BYTE_ORDER */ +# error "Can't define htonll or ntohll!" +#endif + +// Doing a benchmark shows that interning actually makes a difference, amazingly. +#define INTERN_STRING(value) _intern_ ## value + +#define INT_CONV_ERROR_OCCURRED(v) ( ((v) == -1) && PyErr_Occurred() ) +#define CHECK_RANGE(v, min, max) ( ((v) <= (max)) && ((v) >= (min)) ) + +// Py_ssize_t was not defined before Python 2.5 +#if (PY_VERSION_HEX < 0x02050000) +typedef int Py_ssize_t; +#endif + +/** + * A cache of the spec_args for a set or list, + * so we don't have to keep calling PyTuple_GET_ITEM. + */ +typedef struct { + TType element_type; + PyObject* typeargs; +} SetListTypeArgs; + +/** + * A cache of the spec_args for a map, + * so we don't have to keep calling PyTuple_GET_ITEM. + */ +typedef struct { + TType ktag; + TType vtag; + PyObject* ktypeargs; + PyObject* vtypeargs; +} MapTypeArgs; + +/** + * A cache of the spec_args for a struct, + * so we don't have to keep calling PyTuple_GET_ITEM. + */ +typedef struct { + PyObject* klass; + PyObject* spec; +} StructTypeArgs; + +/** + * A cache of the item spec from a struct specification, + * so we don't have to keep calling PyTuple_GET_ITEM. + */ +typedef struct { + int tag; + TType type; + PyObject* attrname; + PyObject* typeargs; + PyObject* defval; +} StructItemSpec; + +/** + * A cache of the two key attributes of a CReadableTransport, + * so we don't have to keep calling PyObject_GetAttr. + */ +typedef struct { + PyObject* stringiobuf; + PyObject* refill_callable; +} DecodeBuffer; + +/** Pointer to interned string to speed up attribute lookup. */ +static PyObject* INTERN_STRING(cstringio_buf); +/** Pointer to interned string to speed up attribute lookup. */ +static PyObject* INTERN_STRING(cstringio_refill); + +static inline bool +check_ssize_t_32(Py_ssize_t len) { + // error from getting the int + if (INT_CONV_ERROR_OCCURRED(len)) { + return false; + } + if (!CHECK_RANGE(len, 0, INT32_MAX)) { + PyErr_SetString(PyExc_OverflowError, "string size out of range"); + return false; + } + return true; +} + +static inline bool +parse_pyint(PyObject* o, int32_t* ret, int32_t min, int32_t max) { + long val = PyInt_AsLong(o); + + if (INT_CONV_ERROR_OCCURRED(val)) { + return false; + } + if (!CHECK_RANGE(val, min, max)) { + PyErr_SetString(PyExc_OverflowError, "int out of range"); + return false; + } + + *ret = (int32_t) val; + return true; +} + + +/* --- FUNCTIONS TO PARSE STRUCT SPECIFICATOINS --- */ + +static bool +parse_set_list_args(SetListTypeArgs* dest, PyObject* typeargs) { + if (PyTuple_Size(typeargs) != 2) { + PyErr_SetString(PyExc_TypeError, "expecting tuple of size 2 for list/set type args"); + return false; + } + + dest->element_type = PyInt_AsLong(PyTuple_GET_ITEM(typeargs, 0)); + if (INT_CONV_ERROR_OCCURRED(dest->element_type)) { + return false; + } + + dest->typeargs = PyTuple_GET_ITEM(typeargs, 1); + + return true; +} + +static bool +parse_map_args(MapTypeArgs* dest, PyObject* typeargs) { + if (PyTuple_Size(typeargs) != 4) { + PyErr_SetString(PyExc_TypeError, "expecting 4 arguments for typeargs to map"); + return false; + } + + dest->ktag = PyInt_AsLong(PyTuple_GET_ITEM(typeargs, 0)); + if (INT_CONV_ERROR_OCCURRED(dest->ktag)) { + return false; + } + + dest->vtag = PyInt_AsLong(PyTuple_GET_ITEM(typeargs, 2)); + if (INT_CONV_ERROR_OCCURRED(dest->vtag)) { + return false; + } + + dest->ktypeargs = PyTuple_GET_ITEM(typeargs, 1); + dest->vtypeargs = PyTuple_GET_ITEM(typeargs, 3); + + return true; +} + +static bool +parse_struct_args(StructTypeArgs* dest, PyObject* typeargs) { + if (PyTuple_Size(typeargs) != 2) { + PyErr_SetString(PyExc_TypeError, "expecting tuple of size 2 for struct args"); + return false; + } + + dest->klass = PyTuple_GET_ITEM(typeargs, 0); + dest->spec = PyTuple_GET_ITEM(typeargs, 1); + + return true; +} + +static int +parse_struct_item_spec(StructItemSpec* dest, PyObject* spec_tuple) { + + // i'd like to use ParseArgs here, but it seems to be a bottleneck. + if (PyTuple_Size(spec_tuple) != 5) { + PyErr_SetString(PyExc_TypeError, "expecting 5 arguments for spec tuple"); + return false; + } + + dest->tag = PyInt_AsLong(PyTuple_GET_ITEM(spec_tuple, 0)); + if (INT_CONV_ERROR_OCCURRED(dest->tag)) { + return false; + } + + dest->type = PyInt_AsLong(PyTuple_GET_ITEM(spec_tuple, 1)); + if (INT_CONV_ERROR_OCCURRED(dest->type)) { + return false; + } + + dest->attrname = PyTuple_GET_ITEM(spec_tuple, 2); + dest->typeargs = PyTuple_GET_ITEM(spec_tuple, 3); + dest->defval = PyTuple_GET_ITEM(spec_tuple, 4); + return true; +} + +/* ====== END UTILITIES ====== */ + + +/* ====== BEGIN WRITING FUNCTIONS ====== */ + +/* --- LOW-LEVEL WRITING FUNCTIONS --- */ + +static void writeByte(PyObject* outbuf, int8_t val) { + int8_t net = val; + PycStringIO->cwrite(outbuf, (char*)&net, sizeof(int8_t)); +} + +static void writeI16(PyObject* outbuf, int16_t val) { + int16_t net = (int16_t)htons(val); + PycStringIO->cwrite(outbuf, (char*)&net, sizeof(int16_t)); +} + +static void writeI32(PyObject* outbuf, int32_t val) { + int32_t net = (int32_t)htonl(val); + PycStringIO->cwrite(outbuf, (char*)&net, sizeof(int32_t)); +} + +static void writeI64(PyObject* outbuf, int64_t val) { + int64_t net = (int64_t)htonll(val); + PycStringIO->cwrite(outbuf, (char*)&net, sizeof(int64_t)); +} + +static void writeDouble(PyObject* outbuf, double dub) { + // Unfortunately, bitwise_cast doesn't work in C. Bad C! + union { + double f; + int64_t t; + } transfer; + transfer.f = dub; + writeI64(outbuf, transfer.t); +} + + +/* --- MAIN RECURSIVE OUTPUT FUCNTION -- */ + +static int +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 = 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; +} + + +/* --- TOP-LEVEL WRAPPER FOR OUTPUT -- */ + +static PyObject * +encode_binary(PyObject *self, PyObject *args) { + PyObject* enc_obj; + PyObject* type_args; + PyObject* buf; + PyObject* ret = NULL; + + if (!PyArg_ParseTuple(args, "OO", &enc_obj, &type_args)) { + return NULL; + } + + buf = PycStringIO->NewOutput(INIT_OUTBUF_SIZE); + if (output_val(buf, enc_obj, T_STRUCT, type_args)) { + ret = PycStringIO->cgetvalue(buf); + } + + Py_DECREF(buf); + return ret; +} + +/* ====== END WRITING FUNCTIONS ====== */ + + +/* ====== BEGIN READING FUNCTIONS ====== */ + +/* --- LOW-LEVEL READING FUNCTIONS --- */ + +static void +free_decodebuf(DecodeBuffer* d) { + Py_XDECREF(d->stringiobuf); + Py_XDECREF(d->refill_callable); +} + +static bool +decode_buffer_from_obj(DecodeBuffer* dest, PyObject* obj) { + dest->stringiobuf = PyObject_GetAttr(obj, INTERN_STRING(cstringio_buf)); + if (!dest->stringiobuf) { + return false; + } + + if (!PycStringIO_InputCheck(dest->stringiobuf)) { + free_decodebuf(dest); + PyErr_SetString(PyExc_TypeError, "expecting stringio input"); + return false; + } + + dest->refill_callable = PyObject_GetAttr(obj, INTERN_STRING(cstringio_refill)); + + if(!dest->refill_callable) { + free_decodebuf(dest); + return false; + } + + if (!PyCallable_Check(dest->refill_callable)) { + free_decodebuf(dest); + PyErr_SetString(PyExc_TypeError, "expecting callable"); + return false; + } + + return true; +} + +static bool readBytes(DecodeBuffer* input, char** output, int len) { + int read; + + // TODO(dreiss): Don't fear the malloc. Think about taking a copy of + // the partial read instead of forcing the transport + // to prepend it to its buffer. + + read = PycStringIO->cread(input->stringiobuf, output, len); + + if (read == len) { + return true; + } else if (read == -1) { + return false; + } else { + PyObject* newiobuf; + + // using building functions as this is a rare codepath + newiobuf = PyObject_CallFunction( + input->refill_callable, "s#i", *output, read, len, NULL); + if (newiobuf == NULL) { + return false; + } + + // must do this *AFTER* the call so that we don't deref the io buffer + Py_CLEAR(input->stringiobuf); + input->stringiobuf = newiobuf; + + read = PycStringIO->cread(input->stringiobuf, output, len); + + if (read == len) { + return true; + } else if (read == -1) { + return false; + } else { + // TODO(dreiss): This could be a valid code path for big binary blobs. + PyErr_SetString(PyExc_TypeError, + "refill claimed to have refilled the buffer, but didn't!!"); + return false; + } + } +} + +static int8_t readByte(DecodeBuffer* input) { + char* buf; + if (!readBytes(input, &buf, sizeof(int8_t))) { + return -1; + } + + return *(int8_t*) buf; +} + +static int16_t readI16(DecodeBuffer* input) { + char* buf; + if (!readBytes(input, &buf, sizeof(int16_t))) { + return -1; + } + + return (int16_t) ntohs(*(int16_t*) buf); +} + +static int32_t readI32(DecodeBuffer* input) { + char* buf; + if (!readBytes(input, &buf, sizeof(int32_t))) { + return -1; + } + return (int32_t) ntohl(*(int32_t*) buf); +} + + +static int64_t readI64(DecodeBuffer* input) { + char* buf; + if (!readBytes(input, &buf, sizeof(int64_t))) { + return -1; + } + + return (int64_t) ntohll(*(int64_t*) buf); +} + +static double readDouble(DecodeBuffer* input) { + union { + int64_t f; + double t; + } transfer; + + transfer.f = readI64(input); + if (transfer.f == -1) { + return -1; + } + return transfer.t; +} + +static bool +checkTypeByte(DecodeBuffer* input, TType expected) { + TType got = readByte(input); + if (INT_CONV_ERROR_OCCURRED(got)) { + return false; + } + + if (expected != got) { + PyErr_SetString(PyExc_TypeError, "got wrong ttype while reading field"); + 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 +} + + +/* --- HELPER FUNCTION FOR DECODE_VAL --- */ + +static PyObject* +decode_val(DecodeBuffer* input, TType type, PyObject* typeargs); + +static bool +decode_struct(DecodeBuffer* input, PyObject* output, PyObject* spec_seq) { + int spec_seq_len = PyTuple_Size(spec_seq); + if (spec_seq_len == -1) { + return false; + } + + while (true) { + TType type; + int16_t tag; + PyObject* item_spec; + PyObject* fieldval = NULL; + StructItemSpec parsedspec; + + type = readByte(input); + if (type == -1) { + return false; + } + if (type == T_STOP) { + break; + } + tag = readI16(input); + if (INT_CONV_ERROR_OCCURRED(tag)) { + return false; + } + if (tag >= 0 && tag < spec_seq_len) { + item_spec = PyTuple_GET_ITEM(spec_seq, tag); + } else { + item_spec = Py_None; + } + + if (item_spec == Py_None) { + if (!skip(input, type)) { + return false; + } else { + continue; + } + } + + if (!parse_struct_item_spec(&parsedspec, item_spec)) { + return false; + } + if (parsedspec.type != type) { + if (!skip(input, type)) { + PyErr_SetString(PyExc_TypeError, "struct field had wrong type while reading and can't be skipped"); + return false; + } else { + continue; + } + } + + fieldval = decode_val(input, parsedspec.type, parsedspec.typeargs); + if (fieldval == NULL) { + return false; + } + + if (PyObject_SetAttr(output, parsedspec.attrname, fieldval) == -1) { + Py_DECREF(fieldval); + return false; + } + Py_DECREF(fieldval); + } + return true; +} + + +/* --- MAIN RECURSIVE INPUT FUCNTION --- */ + +// 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; + } +} + + +/* --- TOP-LEVEL WRAPPER FOR INPUT -- */ + +static PyObject* +decode_binary(PyObject *self, PyObject *args) { + PyObject* output_obj = NULL; + PyObject* transport = NULL; + PyObject* typeargs = NULL; + StructTypeArgs parsedargs; + DecodeBuffer input = {}; + + if (!PyArg_ParseTuple(args, "OOO", &output_obj, &transport, &typeargs)) { + return NULL; + } + + if (!parse_struct_args(&parsedargs, typeargs)) { + return NULL; + } + + if (!decode_buffer_from_obj(&input, transport)) { + return NULL; + } + + if (!decode_struct(&input, output_obj, parsedargs.spec)) { + free_decodebuf(&input); + return NULL; + } + + free_decodebuf(&input); + + Py_RETURN_NONE; +} + +/* ====== END READING FUNCTIONS ====== */ + + +/* -- PYTHON MODULE SETUP STUFF --- */ + +static PyMethodDef ThriftFastBinaryMethods[] = { + + {"encode_binary", encode_binary, METH_VARARGS, ""}, + {"decode_binary", decode_binary, METH_VARARGS, ""}, + + {NULL, NULL, 0, NULL} /* Sentinel */ +}; + +PyMODINIT_FUNC +initfastbinary(void) { +#define INIT_INTERN_STRING(value) \ + do { \ + INTERN_STRING(value) = PyString_InternFromString(#value); \ + if(!INTERN_STRING(value)) return; \ + } while(0) + + INIT_INTERN_STRING(cstringio_buf); + INIT_INTERN_STRING(cstringio_refill); +#undef INIT_INTERN_STRING + + PycString_IMPORT; + if (PycStringIO == NULL) return; + + (void) Py_InitModule("thrift.protocol.fastbinary", ThriftFastBinaryMethods); +} |