| 1 | //===-- PythonDataObjects.cpp ---------------------------------------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "lldb/Host/Config.h" |
| 10 | |
| 11 | #if LLDB_ENABLE_PYTHON |
| 12 | |
| 13 | #include "PythonDataObjects.h" |
| 14 | #include "ScriptInterpreterPython.h" |
| 15 | |
| 16 | #include "lldb/Host/File.h" |
| 17 | #include "lldb/Host/FileSystem.h" |
| 18 | #include "lldb/Interpreter/ScriptInterpreter.h" |
| 19 | #include "lldb/Utility/LLDBLog.h" |
| 20 | #include "lldb/Utility/Log.h" |
| 21 | #include "lldb/Utility/Stream.h" |
| 22 | |
| 23 | #include "llvm/Support/Casting.h" |
| 24 | #include "llvm/Support/ConvertUTF.h" |
| 25 | #include "llvm/Support/Errno.h" |
| 26 | |
| 27 | #include <cstdio> |
| 28 | #include <variant> |
| 29 | |
| 30 | using namespace lldb_private; |
| 31 | using namespace lldb; |
| 32 | using namespace lldb_private::python; |
| 33 | using llvm::cantFail; |
| 34 | using llvm::Error; |
| 35 | using llvm::Expected; |
| 36 | using llvm::Twine; |
| 37 | |
| 38 | template <> Expected<bool> python::As<bool>(Expected<PythonObject> &&obj) { |
| 39 | if (!obj) |
| 40 | return obj.takeError(); |
| 41 | return obj.get().IsTrue(); |
| 42 | } |
| 43 | |
| 44 | template <> |
| 45 | Expected<long long> python::As<long long>(Expected<PythonObject> &&obj) { |
| 46 | if (!obj) |
| 47 | return obj.takeError(); |
| 48 | return obj->AsLongLong(); |
| 49 | } |
| 50 | |
| 51 | template <> |
| 52 | Expected<unsigned long long> |
| 53 | python::As<unsigned long long>(Expected<PythonObject> &&obj) { |
| 54 | if (!obj) |
| 55 | return obj.takeError(); |
| 56 | return obj->AsUnsignedLongLong(); |
| 57 | } |
| 58 | |
| 59 | template <> |
| 60 | Expected<std::string> python::As<std::string>(Expected<PythonObject> &&obj) { |
| 61 | if (!obj) |
| 62 | return obj.takeError(); |
| 63 | PyObject *str_obj = PyObject_Str(obj.get().get()); |
| 64 | if (!str_obj) |
| 65 | return llvm::make_error<PythonException>(); |
| 66 | auto str = Take<PythonString>(obj: str_obj); |
| 67 | auto utf8 = str.AsUTF8(); |
| 68 | if (!utf8) |
| 69 | return utf8.takeError(); |
| 70 | return std::string(utf8.get()); |
| 71 | } |
| 72 | |
| 73 | static bool python_is_finalizing() { |
| 74 | #if PY_VERSION_HEX >= 0x030d0000 |
| 75 | return Py_IsFinalizing(); |
| 76 | #else |
| 77 | return _Py_IsFinalizing(); |
| 78 | #endif |
| 79 | } |
| 80 | |
| 81 | void PythonObject::Reset() { |
| 82 | if (m_py_obj && Py_IsInitialized()) { |
| 83 | if (python_is_finalizing()) { |
| 84 | // Leak m_py_obj rather than crashing the process. |
| 85 | // https://docs.python.org/3/c-api/init.html#c.PyGILState_Ensure |
| 86 | } else { |
| 87 | PyGILState_STATE state = PyGILState_Ensure(); |
| 88 | Py_DECREF(m_py_obj); |
| 89 | PyGILState_Release(state); |
| 90 | } |
| 91 | } |
| 92 | m_py_obj = nullptr; |
| 93 | } |
| 94 | |
| 95 | Expected<long long> PythonObject::AsLongLong() const { |
| 96 | if (!m_py_obj) |
| 97 | return nullDeref(); |
| 98 | assert(!PyErr_Occurred()); |
| 99 | long long r = PyLong_AsLongLong(m_py_obj); |
| 100 | if (PyErr_Occurred()) |
| 101 | return exception(); |
| 102 | return r; |
| 103 | } |
| 104 | |
| 105 | Expected<unsigned long long> PythonObject::AsUnsignedLongLong() const { |
| 106 | if (!m_py_obj) |
| 107 | return nullDeref(); |
| 108 | assert(!PyErr_Occurred()); |
| 109 | long long r = PyLong_AsUnsignedLongLong(m_py_obj); |
| 110 | if (PyErr_Occurred()) |
| 111 | return exception(); |
| 112 | return r; |
| 113 | } |
| 114 | |
| 115 | // wraps on overflow, instead of raising an error. |
| 116 | Expected<unsigned long long> PythonObject::AsModuloUnsignedLongLong() const { |
| 117 | if (!m_py_obj) |
| 118 | return nullDeref(); |
| 119 | assert(!PyErr_Occurred()); |
| 120 | unsigned long long r = PyLong_AsUnsignedLongLongMask(m_py_obj); |
| 121 | // FIXME: We should fetch the exception message and hoist it. |
| 122 | if (PyErr_Occurred()) |
| 123 | return exception(); |
| 124 | return r; |
| 125 | } |
| 126 | |
| 127 | void StructuredPythonObject::Serialize(llvm::json::OStream &s) const { |
| 128 | s.value(V: llvm::formatv(Fmt: "Python Obj: {0:X}" , Vals: GetValue()).str()); |
| 129 | } |
| 130 | |
| 131 | // PythonObject |
| 132 | |
| 133 | void PythonObject::Dump(Stream &strm) const { |
| 134 | if (m_py_obj) { |
| 135 | FILE *file = llvm::sys::RetryAfterSignal(Fail: nullptr, F&: ::tmpfile); |
| 136 | if (file) { |
| 137 | ::PyObject_Print(m_py_obj, file, 0); |
| 138 | const long length = ftell(stream: file); |
| 139 | if (length) { |
| 140 | ::rewind(stream: file); |
| 141 | std::vector<char> file_contents(length, '\0'); |
| 142 | const size_t length_read = |
| 143 | ::fread(ptr: file_contents.data(), size: 1, n: file_contents.size(), stream: file); |
| 144 | if (length_read > 0) |
| 145 | strm.Write(src: file_contents.data(), src_len: length_read); |
| 146 | } |
| 147 | ::fclose(stream: file); |
| 148 | } |
| 149 | } else |
| 150 | strm.PutCString(cstr: "NULL" ); |
| 151 | } |
| 152 | |
| 153 | PyObjectType PythonObject::GetObjectType() const { |
| 154 | if (!IsAllocated()) |
| 155 | return PyObjectType::None; |
| 156 | |
| 157 | if (PythonModule::Check(py_obj: m_py_obj)) |
| 158 | return PyObjectType::Module; |
| 159 | if (PythonList::Check(py_obj: m_py_obj)) |
| 160 | return PyObjectType::List; |
| 161 | if (PythonTuple::Check(py_obj: m_py_obj)) |
| 162 | return PyObjectType::Tuple; |
| 163 | if (PythonDictionary::Check(py_obj: m_py_obj)) |
| 164 | return PyObjectType::Dictionary; |
| 165 | if (PythonString::Check(py_obj: m_py_obj)) |
| 166 | return PyObjectType::String; |
| 167 | if (PythonBytes::Check(py_obj: m_py_obj)) |
| 168 | return PyObjectType::Bytes; |
| 169 | if (PythonByteArray::Check(py_obj: m_py_obj)) |
| 170 | return PyObjectType::ByteArray; |
| 171 | if (PythonBoolean::Check(py_obj: m_py_obj)) |
| 172 | return PyObjectType::Boolean; |
| 173 | if (PythonInteger::Check(py_obj: m_py_obj)) |
| 174 | return PyObjectType::Integer; |
| 175 | if (PythonFile::Check(py_obj: m_py_obj)) |
| 176 | return PyObjectType::File; |
| 177 | if (PythonCallable::Check(py_obj: m_py_obj)) |
| 178 | return PyObjectType::Callable; |
| 179 | return PyObjectType::Unknown; |
| 180 | } |
| 181 | |
| 182 | PythonString PythonObject::Repr() const { |
| 183 | if (!m_py_obj) |
| 184 | return PythonString(); |
| 185 | PyObject *repr = PyObject_Repr(m_py_obj); |
| 186 | if (!repr) |
| 187 | return PythonString(); |
| 188 | return PythonString(PyRefType::Owned, repr); |
| 189 | } |
| 190 | |
| 191 | PythonString PythonObject::Str() const { |
| 192 | if (!m_py_obj) |
| 193 | return PythonString(); |
| 194 | PyObject *str = PyObject_Str(m_py_obj); |
| 195 | if (!str) |
| 196 | return PythonString(); |
| 197 | return PythonString(PyRefType::Owned, str); |
| 198 | } |
| 199 | |
| 200 | PythonObject |
| 201 | PythonObject::ResolveNameWithDictionary(llvm::StringRef name, |
| 202 | const PythonDictionary &dict) { |
| 203 | size_t dot_pos = name.find(C: '.'); |
| 204 | llvm::StringRef piece = name.substr(Start: 0, N: dot_pos); |
| 205 | PythonObject result = dict.GetItemForKey(key: PythonString(piece)); |
| 206 | if (dot_pos == llvm::StringRef::npos) { |
| 207 | // There was no dot, we're done. |
| 208 | return result; |
| 209 | } |
| 210 | |
| 211 | // There was a dot. The remaining portion of the name should be looked up in |
| 212 | // the context of the object that was found in the dictionary. |
| 213 | return result.ResolveName(name: name.substr(Start: dot_pos + 1)); |
| 214 | } |
| 215 | |
| 216 | PythonObject PythonObject::ResolveName(llvm::StringRef name) const { |
| 217 | // Resolve the name in the context of the specified object. If, for example, |
| 218 | // `this` refers to a PyModule, then this will look for `name` in this |
| 219 | // module. If `this` refers to a PyType, then it will resolve `name` as an |
| 220 | // attribute of that type. If `this` refers to an instance of an object, |
| 221 | // then it will resolve `name` as the value of the specified field. |
| 222 | // |
| 223 | // This function handles dotted names so that, for example, if `m_py_obj` |
| 224 | // refers to the `sys` module, and `name` == "path.append", then it will find |
| 225 | // the function `sys.path.append`. |
| 226 | |
| 227 | size_t dot_pos = name.find(C: '.'); |
| 228 | if (dot_pos == llvm::StringRef::npos) { |
| 229 | // No dots in the name, we should be able to find the value immediately as |
| 230 | // an attribute of `m_py_obj`. |
| 231 | return GetAttributeValue(attribute: name); |
| 232 | } |
| 233 | |
| 234 | // Look up the first piece of the name, and resolve the rest as a child of |
| 235 | // that. |
| 236 | PythonObject parent = ResolveName(name: name.substr(Start: 0, N: dot_pos)); |
| 237 | if (!parent.IsAllocated()) |
| 238 | return PythonObject(); |
| 239 | |
| 240 | // Tail recursion.. should be optimized by the compiler |
| 241 | return parent.ResolveName(name: name.substr(Start: dot_pos + 1)); |
| 242 | } |
| 243 | |
| 244 | bool PythonObject::HasAttribute(llvm::StringRef attr) const { |
| 245 | if (!IsValid()) |
| 246 | return false; |
| 247 | PythonString py_attr(attr); |
| 248 | return !!PyObject_HasAttr(m_py_obj, py_attr.get()); |
| 249 | } |
| 250 | |
| 251 | PythonObject PythonObject::GetAttributeValue(llvm::StringRef attr) const { |
| 252 | if (!IsValid()) |
| 253 | return PythonObject(); |
| 254 | |
| 255 | PythonString py_attr(attr); |
| 256 | if (!PyObject_HasAttr(m_py_obj, py_attr.get())) |
| 257 | return PythonObject(); |
| 258 | |
| 259 | return PythonObject(PyRefType::Owned, |
| 260 | PyObject_GetAttr(m_py_obj, py_attr.get())); |
| 261 | } |
| 262 | |
| 263 | StructuredData::ObjectSP PythonObject::CreateStructuredObject() const { |
| 264 | assert(PyGILState_Check()); |
| 265 | switch (GetObjectType()) { |
| 266 | case PyObjectType::Dictionary: |
| 267 | return PythonDictionary(PyRefType::Borrowed, m_py_obj) |
| 268 | .CreateStructuredDictionary(); |
| 269 | case PyObjectType::Boolean: |
| 270 | return PythonBoolean(PyRefType::Borrowed, m_py_obj) |
| 271 | .CreateStructuredBoolean(); |
| 272 | case PyObjectType::Integer: { |
| 273 | StructuredData::IntegerSP int_sp = |
| 274 | PythonInteger(PyRefType::Borrowed, m_py_obj).CreateStructuredInteger(); |
| 275 | if (std::holds_alternative<StructuredData::UnsignedIntegerSP>(v: int_sp)) |
| 276 | return std::get<StructuredData::UnsignedIntegerSP>(v&: int_sp); |
| 277 | if (std::holds_alternative<StructuredData::SignedIntegerSP>(v: int_sp)) |
| 278 | return std::get<StructuredData::SignedIntegerSP>(v&: int_sp); |
| 279 | return nullptr; |
| 280 | }; |
| 281 | case PyObjectType::List: |
| 282 | return PythonList(PyRefType::Borrowed, m_py_obj).CreateStructuredArray(); |
| 283 | case PyObjectType::String: |
| 284 | return PythonString(PyRefType::Borrowed, m_py_obj).CreateStructuredString(); |
| 285 | case PyObjectType::Bytes: |
| 286 | return PythonBytes(PyRefType::Borrowed, m_py_obj).CreateStructuredString(); |
| 287 | case PyObjectType::ByteArray: |
| 288 | return PythonByteArray(PyRefType::Borrowed, m_py_obj) |
| 289 | .CreateStructuredString(); |
| 290 | case PyObjectType::None: |
| 291 | return StructuredData::ObjectSP(); |
| 292 | default: |
| 293 | return StructuredData::ObjectSP(new StructuredPythonObject( |
| 294 | PythonObject(PyRefType::Borrowed, m_py_obj))); |
| 295 | } |
| 296 | } |
| 297 | |
| 298 | // PythonString |
| 299 | |
| 300 | PythonBytes::PythonBytes(llvm::ArrayRef<uint8_t> bytes) { SetBytes(bytes); } |
| 301 | |
| 302 | PythonBytes::PythonBytes(const uint8_t *bytes, size_t length) { |
| 303 | SetBytes(llvm::ArrayRef<uint8_t>(bytes, length)); |
| 304 | } |
| 305 | |
| 306 | bool PythonBytes::Check(PyObject *py_obj) { |
| 307 | if (!py_obj) |
| 308 | return false; |
| 309 | return PyBytes_Check(py_obj); |
| 310 | } |
| 311 | |
| 312 | llvm::ArrayRef<uint8_t> PythonBytes::GetBytes() const { |
| 313 | if (!IsValid()) |
| 314 | return llvm::ArrayRef<uint8_t>(); |
| 315 | |
| 316 | Py_ssize_t size; |
| 317 | char *c; |
| 318 | |
| 319 | PyBytes_AsStringAndSize(obj: m_py_obj, s: &c, len: &size); |
| 320 | return llvm::ArrayRef<uint8_t>(reinterpret_cast<uint8_t *>(c), size); |
| 321 | } |
| 322 | |
| 323 | size_t PythonBytes::GetSize() const { |
| 324 | if (!IsValid()) |
| 325 | return 0; |
| 326 | return PyBytes_Size(m_py_obj); |
| 327 | } |
| 328 | |
| 329 | void PythonBytes::SetBytes(llvm::ArrayRef<uint8_t> bytes) { |
| 330 | const char *data = reinterpret_cast<const char *>(bytes.data()); |
| 331 | *this = Take<PythonBytes>(obj: PyBytes_FromStringAndSize(data, bytes.size())); |
| 332 | } |
| 333 | |
| 334 | StructuredData::StringSP PythonBytes::CreateStructuredString() const { |
| 335 | StructuredData::StringSP result(new StructuredData::String); |
| 336 | Py_ssize_t size; |
| 337 | char *c; |
| 338 | PyBytes_AsStringAndSize(obj: m_py_obj, s: &c, len: &size); |
| 339 | result->SetValue(std::string(c, size)); |
| 340 | return result; |
| 341 | } |
| 342 | |
| 343 | PythonByteArray::PythonByteArray(llvm::ArrayRef<uint8_t> bytes) |
| 344 | : PythonByteArray(bytes.data(), bytes.size()) {} |
| 345 | |
| 346 | PythonByteArray::PythonByteArray(const uint8_t *bytes, size_t length) { |
| 347 | const char *str = reinterpret_cast<const char *>(bytes); |
| 348 | *this = Take<PythonByteArray>(obj: PyByteArray_FromStringAndSize(str, length)); |
| 349 | } |
| 350 | |
| 351 | bool PythonByteArray::Check(PyObject *py_obj) { |
| 352 | if (!py_obj) |
| 353 | return false; |
| 354 | return PyByteArray_Check(py_obj); |
| 355 | } |
| 356 | |
| 357 | llvm::ArrayRef<uint8_t> PythonByteArray::GetBytes() const { |
| 358 | if (!IsValid()) |
| 359 | return llvm::ArrayRef<uint8_t>(); |
| 360 | |
| 361 | char *c = PyByteArray_AsString(m_py_obj); |
| 362 | size_t size = GetSize(); |
| 363 | return llvm::ArrayRef<uint8_t>(reinterpret_cast<uint8_t *>(c), size); |
| 364 | } |
| 365 | |
| 366 | size_t PythonByteArray::GetSize() const { |
| 367 | if (!IsValid()) |
| 368 | return 0; |
| 369 | |
| 370 | return PyByteArray_Size(m_py_obj); |
| 371 | } |
| 372 | |
| 373 | StructuredData::StringSP PythonByteArray::CreateStructuredString() const { |
| 374 | StructuredData::StringSP result(new StructuredData::String); |
| 375 | llvm::ArrayRef<uint8_t> bytes = GetBytes(); |
| 376 | const char *str = reinterpret_cast<const char *>(bytes.data()); |
| 377 | result->SetValue(std::string(str, bytes.size())); |
| 378 | return result; |
| 379 | } |
| 380 | |
| 381 | // PythonString |
| 382 | |
| 383 | Expected<PythonString> PythonString::FromUTF8(llvm::StringRef string) { |
| 384 | PyObject *str = PyUnicode_FromStringAndSize(u: string.data(), size: string.size()); |
| 385 | if (!str) |
| 386 | return llvm::make_error<PythonException>(); |
| 387 | return Take<PythonString>(obj: str); |
| 388 | } |
| 389 | |
| 390 | PythonString::PythonString(llvm::StringRef string) { SetString(string); } |
| 391 | |
| 392 | bool PythonString::Check(PyObject *py_obj) { |
| 393 | if (!py_obj) |
| 394 | return false; |
| 395 | |
| 396 | if (PyUnicode_Check(py_obj)) |
| 397 | return true; |
| 398 | return false; |
| 399 | } |
| 400 | |
| 401 | llvm::StringRef PythonString::GetString() const { |
| 402 | auto s = AsUTF8(); |
| 403 | if (!s) { |
| 404 | llvm::consumeError(Err: s.takeError()); |
| 405 | return llvm::StringRef("" ); |
| 406 | } |
| 407 | return s.get(); |
| 408 | } |
| 409 | |
| 410 | Expected<llvm::StringRef> PythonString::AsUTF8() const { |
| 411 | if (!IsValid()) |
| 412 | return nullDeref(); |
| 413 | |
| 414 | Py_ssize_t size; |
| 415 | const char *data; |
| 416 | |
| 417 | data = PyUnicode_AsUTF8AndSize(unicode: m_py_obj, size: &size); |
| 418 | |
| 419 | if (!data) |
| 420 | return exception(); |
| 421 | |
| 422 | return llvm::StringRef(data, size); |
| 423 | } |
| 424 | |
| 425 | size_t PythonString::GetSize() const { |
| 426 | if (IsValid()) { |
| 427 | #if PY_MINOR_VERSION >= 3 |
| 428 | return PyUnicode_GetLength(unicode: m_py_obj); |
| 429 | #else |
| 430 | return PyUnicode_GetSize(m_py_obj); |
| 431 | #endif |
| 432 | } |
| 433 | return 0; |
| 434 | } |
| 435 | |
| 436 | void PythonString::SetString(llvm::StringRef string) { |
| 437 | auto s = FromUTF8(string); |
| 438 | if (!s) { |
| 439 | llvm::consumeError(Err: s.takeError()); |
| 440 | Reset(); |
| 441 | } else { |
| 442 | *this = std::move(s.get()); |
| 443 | } |
| 444 | } |
| 445 | |
| 446 | StructuredData::StringSP PythonString::CreateStructuredString() const { |
| 447 | StructuredData::StringSP result(new StructuredData::String); |
| 448 | result->SetValue(GetString()); |
| 449 | return result; |
| 450 | } |
| 451 | |
| 452 | // PythonInteger |
| 453 | |
| 454 | PythonInteger::PythonInteger(int64_t value) { SetInteger(value); } |
| 455 | |
| 456 | bool PythonInteger::Check(PyObject *py_obj) { |
| 457 | if (!py_obj) |
| 458 | return false; |
| 459 | |
| 460 | // Python 3 does not have PyInt_Check. There is only one type of integral |
| 461 | // value, long. |
| 462 | return PyLong_Check(py_obj); |
| 463 | } |
| 464 | |
| 465 | void PythonInteger::SetInteger(int64_t value) { |
| 466 | *this = Take<PythonInteger>(obj: PyLong_FromLongLong(value)); |
| 467 | } |
| 468 | |
| 469 | StructuredData::IntegerSP PythonInteger::CreateStructuredInteger() const { |
| 470 | StructuredData::UnsignedIntegerSP uint_sp = CreateStructuredUnsignedInteger(); |
| 471 | return uint_sp ? StructuredData::IntegerSP(uint_sp) |
| 472 | : CreateStructuredSignedInteger(); |
| 473 | } |
| 474 | |
| 475 | StructuredData::UnsignedIntegerSP |
| 476 | PythonInteger::CreateStructuredUnsignedInteger() const { |
| 477 | StructuredData::UnsignedIntegerSP result = nullptr; |
| 478 | llvm::Expected<unsigned long long> value = AsUnsignedLongLong(); |
| 479 | if (!value) |
| 480 | llvm::consumeError(Err: value.takeError()); |
| 481 | else |
| 482 | result = std::make_shared<StructuredData::UnsignedInteger>(args&: value.get()); |
| 483 | |
| 484 | return result; |
| 485 | } |
| 486 | |
| 487 | StructuredData::SignedIntegerSP |
| 488 | PythonInteger::CreateStructuredSignedInteger() const { |
| 489 | StructuredData::SignedIntegerSP result = nullptr; |
| 490 | llvm::Expected<long long> value = AsLongLong(); |
| 491 | if (!value) |
| 492 | llvm::consumeError(Err: value.takeError()); |
| 493 | else |
| 494 | result = std::make_shared<StructuredData::SignedInteger>(args&: value.get()); |
| 495 | |
| 496 | return result; |
| 497 | } |
| 498 | |
| 499 | // PythonBoolean |
| 500 | |
| 501 | PythonBoolean::PythonBoolean(bool value) { |
| 502 | SetValue(value); |
| 503 | } |
| 504 | |
| 505 | bool PythonBoolean::Check(PyObject *py_obj) { |
| 506 | return py_obj ? PyBool_Check(py_obj) : false; |
| 507 | } |
| 508 | |
| 509 | bool PythonBoolean::GetValue() const { |
| 510 | return m_py_obj ? PyObject_IsTrue(m_py_obj) : false; |
| 511 | } |
| 512 | |
| 513 | void PythonBoolean::SetValue(bool value) { |
| 514 | *this = Take<PythonBoolean>(obj: PyBool_FromLong(value)); |
| 515 | } |
| 516 | |
| 517 | StructuredData::BooleanSP PythonBoolean::CreateStructuredBoolean() const { |
| 518 | StructuredData::BooleanSP result(new StructuredData::Boolean); |
| 519 | result->SetValue(GetValue()); |
| 520 | return result; |
| 521 | } |
| 522 | |
| 523 | // PythonList |
| 524 | |
| 525 | PythonList::PythonList(PyInitialValue value) { |
| 526 | if (value == PyInitialValue::Empty) |
| 527 | *this = Take<PythonList>(obj: PyList_New(size: 0)); |
| 528 | } |
| 529 | |
| 530 | PythonList::PythonList(int list_size) { |
| 531 | *this = Take<PythonList>(obj: PyList_New(size: list_size)); |
| 532 | } |
| 533 | |
| 534 | bool PythonList::Check(PyObject *py_obj) { |
| 535 | if (!py_obj) |
| 536 | return false; |
| 537 | return PyList_Check(py_obj); |
| 538 | } |
| 539 | |
| 540 | uint32_t PythonList::GetSize() const { |
| 541 | if (IsValid()) |
| 542 | return PyList_GET_SIZE(m_py_obj); |
| 543 | return 0; |
| 544 | } |
| 545 | |
| 546 | PythonObject PythonList::GetItemAtIndex(uint32_t index) const { |
| 547 | if (IsValid()) |
| 548 | return PythonObject(PyRefType::Borrowed, PyList_GetItem(m_py_obj, index)); |
| 549 | return PythonObject(); |
| 550 | } |
| 551 | |
| 552 | void PythonList::SetItemAtIndex(uint32_t index, const PythonObject &object) { |
| 553 | if (IsAllocated() && object.IsValid()) { |
| 554 | // PyList_SetItem is documented to "steal" a reference, so we need to |
| 555 | // convert it to an owned reference by incrementing it. |
| 556 | Py_INCREF(object.get()); |
| 557 | PyList_SetItem(m_py_obj, index, object.get()); |
| 558 | } |
| 559 | } |
| 560 | |
| 561 | void PythonList::AppendItem(const PythonObject &object) { |
| 562 | if (IsAllocated() && object.IsValid()) { |
| 563 | // `PyList_Append` does *not* steal a reference, so do not call `Py_INCREF` |
| 564 | // here like we do with `PyList_SetItem`. |
| 565 | PyList_Append(m_py_obj, object.get()); |
| 566 | } |
| 567 | } |
| 568 | |
| 569 | StructuredData::ArraySP PythonList::CreateStructuredArray() const { |
| 570 | StructuredData::ArraySP result(new StructuredData::Array); |
| 571 | uint32_t count = GetSize(); |
| 572 | for (uint32_t i = 0; i < count; ++i) { |
| 573 | PythonObject obj = GetItemAtIndex(index: i); |
| 574 | result->AddItem(item: obj.CreateStructuredObject()); |
| 575 | } |
| 576 | return result; |
| 577 | } |
| 578 | |
| 579 | // PythonTuple |
| 580 | |
| 581 | PythonTuple::PythonTuple(PyInitialValue value) { |
| 582 | if (value == PyInitialValue::Empty) |
| 583 | *this = Take<PythonTuple>(obj: PyTuple_New(size: 0)); |
| 584 | } |
| 585 | |
| 586 | PythonTuple::PythonTuple(int tuple_size) { |
| 587 | *this = Take<PythonTuple>(obj: PyTuple_New(size: tuple_size)); |
| 588 | } |
| 589 | |
| 590 | PythonTuple::PythonTuple(std::initializer_list<PythonObject> objects) { |
| 591 | m_py_obj = PyTuple_New(size: objects.size()); |
| 592 | |
| 593 | uint32_t idx = 0; |
| 594 | for (auto object : objects) { |
| 595 | if (object.IsValid()) |
| 596 | SetItemAtIndex(index: idx, object); |
| 597 | idx++; |
| 598 | } |
| 599 | } |
| 600 | |
| 601 | PythonTuple::PythonTuple(std::initializer_list<PyObject *> objects) { |
| 602 | m_py_obj = PyTuple_New(size: objects.size()); |
| 603 | |
| 604 | uint32_t idx = 0; |
| 605 | for (auto py_object : objects) { |
| 606 | PythonObject object(PyRefType::Borrowed, py_object); |
| 607 | if (object.IsValid()) |
| 608 | SetItemAtIndex(index: idx, object); |
| 609 | idx++; |
| 610 | } |
| 611 | } |
| 612 | |
| 613 | bool PythonTuple::Check(PyObject *py_obj) { |
| 614 | if (!py_obj) |
| 615 | return false; |
| 616 | return PyTuple_Check(py_obj); |
| 617 | } |
| 618 | |
| 619 | uint32_t PythonTuple::GetSize() const { |
| 620 | if (IsValid()) |
| 621 | return PyTuple_GET_SIZE(m_py_obj); |
| 622 | return 0; |
| 623 | } |
| 624 | |
| 625 | PythonObject PythonTuple::GetItemAtIndex(uint32_t index) const { |
| 626 | if (IsValid()) |
| 627 | return PythonObject(PyRefType::Borrowed, PyTuple_GetItem(m_py_obj, index)); |
| 628 | return PythonObject(); |
| 629 | } |
| 630 | |
| 631 | void PythonTuple::SetItemAtIndex(uint32_t index, const PythonObject &object) { |
| 632 | if (IsAllocated() && object.IsValid()) { |
| 633 | // PyTuple_SetItem is documented to "steal" a reference, so we need to |
| 634 | // convert it to an owned reference by incrementing it. |
| 635 | Py_INCREF(object.get()); |
| 636 | PyTuple_SetItem(m_py_obj, index, object.get()); |
| 637 | } |
| 638 | } |
| 639 | |
| 640 | StructuredData::ArraySP PythonTuple::CreateStructuredArray() const { |
| 641 | StructuredData::ArraySP result(new StructuredData::Array); |
| 642 | uint32_t count = GetSize(); |
| 643 | for (uint32_t i = 0; i < count; ++i) { |
| 644 | PythonObject obj = GetItemAtIndex(index: i); |
| 645 | result->AddItem(item: obj.CreateStructuredObject()); |
| 646 | } |
| 647 | return result; |
| 648 | } |
| 649 | |
| 650 | // PythonDictionary |
| 651 | |
| 652 | PythonDictionary::PythonDictionary(PyInitialValue value) { |
| 653 | if (value == PyInitialValue::Empty) |
| 654 | *this = Take<PythonDictionary>(obj: PyDict_New()); |
| 655 | } |
| 656 | |
| 657 | bool PythonDictionary::Check(PyObject *py_obj) { |
| 658 | if (!py_obj) |
| 659 | return false; |
| 660 | |
| 661 | return PyDict_Check(py_obj); |
| 662 | } |
| 663 | |
| 664 | bool PythonDictionary::HasKey(const llvm::Twine &key) const { |
| 665 | if (!IsValid()) |
| 666 | return false; |
| 667 | |
| 668 | PythonString key_object(key.isSingleStringRef() ? key.getSingleStringRef() |
| 669 | : key.str()); |
| 670 | |
| 671 | if (int res = PyDict_Contains(mp: m_py_obj, key: key_object.get()) > 0) |
| 672 | return res; |
| 673 | |
| 674 | PyErr_Print(); |
| 675 | return false; |
| 676 | } |
| 677 | |
| 678 | uint32_t PythonDictionary::GetSize() const { |
| 679 | if (IsValid()) |
| 680 | return PyDict_Size(mp: m_py_obj); |
| 681 | return 0; |
| 682 | } |
| 683 | |
| 684 | PythonList PythonDictionary::GetKeys() const { |
| 685 | if (IsValid()) |
| 686 | return PythonList(PyRefType::Owned, PyDict_Keys(mp: m_py_obj)); |
| 687 | return PythonList(PyInitialValue::Invalid); |
| 688 | } |
| 689 | |
| 690 | PythonObject PythonDictionary::GetItemForKey(const PythonObject &key) const { |
| 691 | auto item = GetItem(key); |
| 692 | if (!item) { |
| 693 | llvm::consumeError(Err: item.takeError()); |
| 694 | return PythonObject(); |
| 695 | } |
| 696 | return std::move(item.get()); |
| 697 | } |
| 698 | |
| 699 | Expected<PythonObject> |
| 700 | PythonDictionary::GetItem(const PythonObject &key) const { |
| 701 | if (!IsValid()) |
| 702 | return nullDeref(); |
| 703 | PyObject *o = PyDict_GetItemWithError(mp: m_py_obj, key: key.get()); |
| 704 | if (PyErr_Occurred()) |
| 705 | return exception(); |
| 706 | if (!o) |
| 707 | return keyError(); |
| 708 | return Retain<PythonObject>(obj: o); |
| 709 | } |
| 710 | |
| 711 | Expected<PythonObject> PythonDictionary::GetItem(const Twine &key) const { |
| 712 | if (!IsValid()) |
| 713 | return nullDeref(); |
| 714 | PyObject *o = PyDict_GetItemString(dp: m_py_obj, key: NullTerminated(key)); |
| 715 | if (PyErr_Occurred()) |
| 716 | return exception(); |
| 717 | if (!o) |
| 718 | return keyError(); |
| 719 | return Retain<PythonObject>(obj: o); |
| 720 | } |
| 721 | |
| 722 | Error PythonDictionary::SetItem(const PythonObject &key, |
| 723 | const PythonObject &value) const { |
| 724 | if (!IsValid() || !value.IsValid()) |
| 725 | return nullDeref(); |
| 726 | int r = PyDict_SetItem(mp: m_py_obj, key: key.get(), item: value.get()); |
| 727 | if (r < 0) |
| 728 | return exception(); |
| 729 | return Error::success(); |
| 730 | } |
| 731 | |
| 732 | Error PythonDictionary::SetItem(const Twine &key, |
| 733 | const PythonObject &value) const { |
| 734 | if (!IsValid() || !value.IsValid()) |
| 735 | return nullDeref(); |
| 736 | int r = PyDict_SetItemString(dp: m_py_obj, key: NullTerminated(key), item: value.get()); |
| 737 | if (r < 0) |
| 738 | return exception(); |
| 739 | return Error::success(); |
| 740 | } |
| 741 | |
| 742 | void PythonDictionary::SetItemForKey(const PythonObject &key, |
| 743 | const PythonObject &value) { |
| 744 | Error error = SetItem(key, value); |
| 745 | if (error) |
| 746 | llvm::consumeError(Err: std::move(error)); |
| 747 | } |
| 748 | |
| 749 | StructuredData::DictionarySP |
| 750 | PythonDictionary::CreateStructuredDictionary() const { |
| 751 | StructuredData::DictionarySP result(new StructuredData::Dictionary); |
| 752 | PythonList keys(GetKeys()); |
| 753 | uint32_t num_keys = keys.GetSize(); |
| 754 | for (uint32_t i = 0; i < num_keys; ++i) { |
| 755 | PythonObject key = keys.GetItemAtIndex(index: i); |
| 756 | PythonObject value = GetItemForKey(key); |
| 757 | StructuredData::ObjectSP structured_value = value.CreateStructuredObject(); |
| 758 | result->AddItem(key: key.Str().GetString(), value_sp: structured_value); |
| 759 | } |
| 760 | return result; |
| 761 | } |
| 762 | |
| 763 | PythonModule PythonModule::BuiltinsModule() { return AddModule(module: "builtins" ); } |
| 764 | |
| 765 | PythonModule PythonModule::MainModule() { return AddModule(module: "__main__" ); } |
| 766 | |
| 767 | PythonModule PythonModule::AddModule(llvm::StringRef module) { |
| 768 | std::string str = module.str(); |
| 769 | return PythonModule(PyRefType::Borrowed, PyImport_AddModule(name: str.c_str())); |
| 770 | } |
| 771 | |
| 772 | Expected<PythonModule> PythonModule::Import(const Twine &name) { |
| 773 | PyObject *mod = PyImport_ImportModule(name: NullTerminated(name)); |
| 774 | if (!mod) |
| 775 | return exception(); |
| 776 | return Take<PythonModule>(obj: mod); |
| 777 | } |
| 778 | |
| 779 | Expected<PythonObject> PythonModule::Get(const Twine &name) { |
| 780 | if (!IsValid()) |
| 781 | return nullDeref(); |
| 782 | PyObject *dict = PyModule_GetDict(m_py_obj); |
| 783 | if (!dict) |
| 784 | return exception(); |
| 785 | PyObject *item = PyDict_GetItemString(dp: dict, key: NullTerminated(name)); |
| 786 | if (!item) |
| 787 | return exception(); |
| 788 | return Retain<PythonObject>(obj: item); |
| 789 | } |
| 790 | |
| 791 | bool PythonModule::Check(PyObject *py_obj) { |
| 792 | if (!py_obj) |
| 793 | return false; |
| 794 | |
| 795 | return PyModule_Check(py_obj); |
| 796 | } |
| 797 | |
| 798 | PythonDictionary PythonModule::GetDictionary() const { |
| 799 | if (!IsValid()) |
| 800 | return PythonDictionary(); |
| 801 | return Retain<PythonDictionary>(obj: PyModule_GetDict(m_py_obj)); |
| 802 | } |
| 803 | |
| 804 | bool PythonCallable::Check(PyObject *py_obj) { |
| 805 | if (!py_obj) |
| 806 | return false; |
| 807 | |
| 808 | return PyCallable_Check(py_obj); |
| 809 | } |
| 810 | |
| 811 | static const char get_arg_info_script[] = R"( |
| 812 | from inspect import signature, Parameter, ismethod |
| 813 | from collections import namedtuple |
| 814 | ArgInfo = namedtuple('ArgInfo', ['count', 'has_varargs']) |
| 815 | def main(f): |
| 816 | count = 0 |
| 817 | varargs = False |
| 818 | for parameter in signature(f).parameters.values(): |
| 819 | kind = parameter.kind |
| 820 | if kind in (Parameter.POSITIONAL_ONLY, |
| 821 | Parameter.POSITIONAL_OR_KEYWORD): |
| 822 | count += 1 |
| 823 | elif kind == Parameter.VAR_POSITIONAL: |
| 824 | varargs = True |
| 825 | elif kind in (Parameter.KEYWORD_ONLY, |
| 826 | Parameter.VAR_KEYWORD): |
| 827 | pass |
| 828 | else: |
| 829 | raise Exception(f'unknown parameter kind: {kind}') |
| 830 | return ArgInfo(count, varargs) |
| 831 | )" ; |
| 832 | |
| 833 | Expected<PythonCallable::ArgInfo> PythonCallable::GetArgInfo() const { |
| 834 | ArgInfo result = {}; |
| 835 | if (!IsValid()) |
| 836 | return nullDeref(); |
| 837 | |
| 838 | // no need to synchronize access to this global, we already have the GIL |
| 839 | static PythonScript get_arg_info(get_arg_info_script); |
| 840 | Expected<PythonObject> pyarginfo = get_arg_info(*this); |
| 841 | if (!pyarginfo) |
| 842 | return pyarginfo.takeError(); |
| 843 | long long count = |
| 844 | cantFail(ValOrErr: As<long long>(obj: pyarginfo.get().GetAttribute(name: "count" ))); |
| 845 | bool has_varargs = |
| 846 | cantFail(ValOrErr: As<bool>(obj: pyarginfo.get().GetAttribute(name: "has_varargs" ))); |
| 847 | result.max_positional_args = has_varargs ? ArgInfo::UNBOUNDED : count; |
| 848 | |
| 849 | return result; |
| 850 | } |
| 851 | |
| 852 | constexpr unsigned |
| 853 | PythonCallable::ArgInfo::UNBOUNDED; // FIXME delete after c++17 |
| 854 | |
| 855 | PythonObject PythonCallable::operator()() { |
| 856 | return PythonObject(PyRefType::Owned, PyObject_CallObject(callable: m_py_obj, args: nullptr)); |
| 857 | } |
| 858 | |
| 859 | PythonObject PythonCallable:: |
| 860 | operator()(std::initializer_list<PyObject *> args) { |
| 861 | PythonTuple arg_tuple(args); |
| 862 | return PythonObject(PyRefType::Owned, |
| 863 | PyObject_CallObject(callable: m_py_obj, args: arg_tuple.get())); |
| 864 | } |
| 865 | |
| 866 | PythonObject PythonCallable:: |
| 867 | operator()(std::initializer_list<PythonObject> args) { |
| 868 | PythonTuple arg_tuple(args); |
| 869 | return PythonObject(PyRefType::Owned, |
| 870 | PyObject_CallObject(callable: m_py_obj, args: arg_tuple.get())); |
| 871 | } |
| 872 | |
| 873 | bool PythonFile::Check(PyObject *py_obj) { |
| 874 | if (!py_obj) |
| 875 | return false; |
| 876 | // In Python 3, there is no `PyFile_Check`, and in fact PyFile is not even a |
| 877 | // first-class object type anymore. `PyFile_FromFd` is just a thin wrapper |
| 878 | // over `io.open()`, which returns some object derived from `io.IOBase`. As a |
| 879 | // result, the only way to detect a file in Python 3 is to check whether it |
| 880 | // inherits from `io.IOBase`. |
| 881 | auto io_module = PythonModule::Import(name: "io" ); |
| 882 | if (!io_module) { |
| 883 | llvm::consumeError(Err: io_module.takeError()); |
| 884 | return false; |
| 885 | } |
| 886 | auto iobase = io_module.get().Get(name: "IOBase" ); |
| 887 | if (!iobase) { |
| 888 | llvm::consumeError(Err: iobase.takeError()); |
| 889 | return false; |
| 890 | } |
| 891 | int r = PyObject_IsInstance(object: py_obj, typeorclass: iobase.get().get()); |
| 892 | if (r < 0) { |
| 893 | llvm::consumeError(Err: exception()); // clear the exception and log it. |
| 894 | return false; |
| 895 | } |
| 896 | return !!r; |
| 897 | } |
| 898 | |
| 899 | const char *PythonException::toCString() const { |
| 900 | if (!m_repr_bytes) |
| 901 | return "unknown exception" ; |
| 902 | return PyBytes_AS_STRING(m_repr_bytes); |
| 903 | } |
| 904 | |
| 905 | PythonException::PythonException(const char *caller) { |
| 906 | assert(PyErr_Occurred()); |
| 907 | m_exception_type = m_exception = m_traceback = m_repr_bytes = nullptr; |
| 908 | PyErr_Fetch(&m_exception_type, &m_exception, &m_traceback); |
| 909 | PyErr_NormalizeException(&m_exception_type, &m_exception, &m_traceback); |
| 910 | PyErr_Clear(); |
| 911 | if (m_exception) { |
| 912 | PyObject *repr = PyObject_Repr(m_exception); |
| 913 | if (repr) { |
| 914 | m_repr_bytes = PyUnicode_AsEncodedString(unicode: repr, encoding: "utf-8" , errors: nullptr); |
| 915 | if (!m_repr_bytes) { |
| 916 | PyErr_Clear(); |
| 917 | } |
| 918 | Py_XDECREF(repr); |
| 919 | } else { |
| 920 | PyErr_Clear(); |
| 921 | } |
| 922 | } |
| 923 | Log *log = GetLog(mask: LLDBLog::Script); |
| 924 | if (caller) |
| 925 | LLDB_LOGF(log, "%s failed with exception: %s" , caller, toCString()); |
| 926 | else |
| 927 | LLDB_LOGF(log, "python exception: %s" , toCString()); |
| 928 | } |
| 929 | void PythonException::Restore() { |
| 930 | if (m_exception_type && m_exception) { |
| 931 | PyErr_Restore(m_exception_type, m_exception, m_traceback); |
| 932 | } else { |
| 933 | PyErr_SetString(exception: PyExc_Exception, string: toCString()); |
| 934 | } |
| 935 | m_exception_type = m_exception = m_traceback = nullptr; |
| 936 | } |
| 937 | |
| 938 | PythonException::~PythonException() { |
| 939 | Py_XDECREF(m_exception_type); |
| 940 | Py_XDECREF(m_exception); |
| 941 | Py_XDECREF(m_traceback); |
| 942 | Py_XDECREF(m_repr_bytes); |
| 943 | } |
| 944 | |
| 945 | void PythonException::log(llvm::raw_ostream &OS) const { OS << toCString(); } |
| 946 | |
| 947 | std::error_code PythonException::convertToErrorCode() const { |
| 948 | return llvm::inconvertibleErrorCode(); |
| 949 | } |
| 950 | |
| 951 | bool PythonException::Matches(PyObject *exc) const { |
| 952 | return PyErr_GivenExceptionMatches(m_exception_type, exc); |
| 953 | } |
| 954 | |
| 955 | const char read_exception_script[] = R"( |
| 956 | import sys |
| 957 | from traceback import print_exception |
| 958 | if sys.version_info.major < 3: |
| 959 | from StringIO import StringIO |
| 960 | else: |
| 961 | from io import StringIO |
| 962 | def main(exc_type, exc_value, tb): |
| 963 | f = StringIO() |
| 964 | print_exception(exc_type, exc_value, tb, file=f) |
| 965 | return f.getvalue() |
| 966 | )" ; |
| 967 | |
| 968 | std::string PythonException::ReadBacktrace() const { |
| 969 | |
| 970 | if (!m_traceback) |
| 971 | return toCString(); |
| 972 | |
| 973 | // no need to synchronize access to this global, we already have the GIL |
| 974 | static PythonScript read_exception(read_exception_script); |
| 975 | |
| 976 | Expected<std::string> backtrace = As<std::string>( |
| 977 | obj: read_exception(m_exception_type, m_exception, m_traceback)); |
| 978 | |
| 979 | if (!backtrace) { |
| 980 | std::string message = |
| 981 | std::string(toCString()) + "\n" + |
| 982 | "Traceback unavailable, an error occurred while reading it:\n" ; |
| 983 | return (message + llvm::toString(E: backtrace.takeError())); |
| 984 | } |
| 985 | |
| 986 | return std::move(backtrace.get()); |
| 987 | } |
| 988 | |
| 989 | char PythonException::ID = 0; |
| 990 | |
| 991 | llvm::Expected<File::OpenOptions> |
| 992 | GetOptionsForPyObject(const PythonObject &obj) { |
| 993 | auto options = File::OpenOptions(0); |
| 994 | auto readable = As<bool>(obj: obj.CallMethod(name: "readable" )); |
| 995 | if (!readable) |
| 996 | return readable.takeError(); |
| 997 | auto writable = As<bool>(obj: obj.CallMethod(name: "writable" )); |
| 998 | if (!writable) |
| 999 | return writable.takeError(); |
| 1000 | if (readable.get() && writable.get()) |
| 1001 | options |= File::eOpenOptionReadWrite; |
| 1002 | else if (writable.get()) |
| 1003 | options |= File::eOpenOptionWriteOnly; |
| 1004 | else if (readable.get()) |
| 1005 | options |= File::eOpenOptionReadOnly; |
| 1006 | return options; |
| 1007 | } |
| 1008 | |
| 1009 | // Base class template for python files. All it knows how to do |
| 1010 | // is hold a reference to the python object and close or flush it |
| 1011 | // when the File is closed. |
| 1012 | namespace { |
| 1013 | template <typename Base> class OwnedPythonFile : public Base { |
| 1014 | public: |
| 1015 | template <typename... Args> |
| 1016 | OwnedPythonFile(const PythonFile &file, bool borrowed, Args... args) |
| 1017 | : Base(args...), m_py_obj(file), m_borrowed(borrowed) { |
| 1018 | assert(m_py_obj); |
| 1019 | } |
| 1020 | |
| 1021 | ~OwnedPythonFile() override { |
| 1022 | assert(m_py_obj); |
| 1023 | GIL takeGIL; |
| 1024 | Close(); |
| 1025 | // we need to ensure the python object is released while we still |
| 1026 | // hold the GIL |
| 1027 | m_py_obj.Reset(); |
| 1028 | } |
| 1029 | |
| 1030 | bool IsPythonSideValid() const { |
| 1031 | GIL takeGIL; |
| 1032 | auto closed = As<bool>(obj: m_py_obj.GetAttribute(name: "closed" )); |
| 1033 | if (!closed) { |
| 1034 | llvm::consumeError(Err: closed.takeError()); |
| 1035 | return false; |
| 1036 | } |
| 1037 | return !closed.get(); |
| 1038 | } |
| 1039 | |
| 1040 | bool IsValid() const override { |
| 1041 | return IsPythonSideValid() && Base::IsValid(); |
| 1042 | } |
| 1043 | |
| 1044 | Status Close() override { |
| 1045 | assert(m_py_obj); |
| 1046 | Status py_error, base_error; |
| 1047 | GIL takeGIL; |
| 1048 | if (!m_borrowed) { |
| 1049 | auto r = m_py_obj.CallMethod(name: "close" ); |
| 1050 | if (!r) |
| 1051 | py_error = Status::FromError(error: r.takeError()); |
| 1052 | } |
| 1053 | base_error = Base::Close(); |
| 1054 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1055 | if (py_error.Fail()) |
| 1056 | return py_error.Clone(); |
| 1057 | return base_error.Clone(); |
| 1058 | }; |
| 1059 | |
| 1060 | PyObject *GetPythonObject() const { |
| 1061 | assert(m_py_obj.IsValid()); |
| 1062 | return m_py_obj.get(); |
| 1063 | } |
| 1064 | |
| 1065 | static bool classof(const File *file) = delete; |
| 1066 | |
| 1067 | protected: |
| 1068 | PythonFile m_py_obj; |
| 1069 | bool m_borrowed; |
| 1070 | }; |
| 1071 | } // namespace |
| 1072 | |
| 1073 | // A SimplePythonFile is a OwnedPythonFile that just does all I/O as |
| 1074 | // a NativeFile |
| 1075 | namespace { |
| 1076 | class SimplePythonFile : public OwnedPythonFile<NativeFile> { |
| 1077 | public: |
| 1078 | SimplePythonFile(const PythonFile &file, bool borrowed, int fd, |
| 1079 | File::OpenOptions options) |
| 1080 | : OwnedPythonFile(file, borrowed, fd, options, false) {} |
| 1081 | |
| 1082 | static char ID; |
| 1083 | bool isA(const void *classID) const override { |
| 1084 | return classID == &ID || NativeFile::isA(classID); |
| 1085 | } |
| 1086 | static bool classof(const File *file) { return file->isA(classID: &ID); } |
| 1087 | }; |
| 1088 | char SimplePythonFile::ID = 0; |
| 1089 | } // namespace |
| 1090 | |
| 1091 | namespace { |
| 1092 | class PythonBuffer { |
| 1093 | public: |
| 1094 | PythonBuffer &operator=(const PythonBuffer &) = delete; |
| 1095 | PythonBuffer(const PythonBuffer &) = delete; |
| 1096 | |
| 1097 | static Expected<PythonBuffer> Create(PythonObject &obj, |
| 1098 | int flags = PyBUF_SIMPLE) { |
| 1099 | Py_buffer py_buffer = {}; |
| 1100 | PyObject_GetBuffer(obj: obj.get(), view: &py_buffer, flags); |
| 1101 | if (!py_buffer.obj) |
| 1102 | return llvm::make_error<PythonException>(); |
| 1103 | return PythonBuffer(py_buffer); |
| 1104 | } |
| 1105 | |
| 1106 | PythonBuffer(PythonBuffer &&other) { |
| 1107 | m_buffer = other.m_buffer; |
| 1108 | other.m_buffer.obj = nullptr; |
| 1109 | } |
| 1110 | |
| 1111 | ~PythonBuffer() { |
| 1112 | if (m_buffer.obj) |
| 1113 | PyBuffer_Release(view: &m_buffer); |
| 1114 | } |
| 1115 | |
| 1116 | Py_buffer &get() { return m_buffer; } |
| 1117 | |
| 1118 | private: |
| 1119 | // takes ownership of the buffer. |
| 1120 | PythonBuffer(const Py_buffer &py_buffer) : m_buffer(py_buffer) {} |
| 1121 | Py_buffer m_buffer; |
| 1122 | }; |
| 1123 | } // namespace |
| 1124 | |
| 1125 | // Shared methods between TextPythonFile and BinaryPythonFile |
| 1126 | namespace { |
| 1127 | class PythonIOFile : public OwnedPythonFile<File> { |
| 1128 | public: |
| 1129 | PythonIOFile(const PythonFile &file, bool borrowed) |
| 1130 | : OwnedPythonFile(file, borrowed) {} |
| 1131 | |
| 1132 | ~PythonIOFile() override { Close(); } |
| 1133 | |
| 1134 | bool IsValid() const override { return IsPythonSideValid(); } |
| 1135 | |
| 1136 | Status Close() override { |
| 1137 | assert(m_py_obj); |
| 1138 | GIL takeGIL; |
| 1139 | if (m_borrowed) |
| 1140 | return Flush(); |
| 1141 | auto r = m_py_obj.CallMethod(name: "close" ); |
| 1142 | if (!r) |
| 1143 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1144 | return Status::FromError(error: r.takeError()).Clone(); |
| 1145 | return Status(); |
| 1146 | } |
| 1147 | |
| 1148 | Status Flush() override { |
| 1149 | GIL takeGIL; |
| 1150 | auto r = m_py_obj.CallMethod(name: "flush" ); |
| 1151 | if (!r) |
| 1152 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1153 | return Status::FromError(error: r.takeError()).Clone(); |
| 1154 | return Status(); |
| 1155 | } |
| 1156 | |
| 1157 | Expected<File::OpenOptions> GetOptions() const override { |
| 1158 | GIL takeGIL; |
| 1159 | return GetOptionsForPyObject(obj: m_py_obj); |
| 1160 | } |
| 1161 | |
| 1162 | static char ID; |
| 1163 | bool isA(const void *classID) const override { |
| 1164 | return classID == &ID || File::isA(classID); |
| 1165 | } |
| 1166 | static bool classof(const File *file) { return file->isA(classID: &ID); } |
| 1167 | }; |
| 1168 | char PythonIOFile::ID = 0; |
| 1169 | } // namespace |
| 1170 | |
| 1171 | namespace { |
| 1172 | class BinaryPythonFile : public PythonIOFile { |
| 1173 | protected: |
| 1174 | int m_descriptor; |
| 1175 | |
| 1176 | public: |
| 1177 | BinaryPythonFile(int fd, const PythonFile &file, bool borrowed) |
| 1178 | : PythonIOFile(file, borrowed), |
| 1179 | m_descriptor(File::DescriptorIsValid(descriptor: fd) ? fd |
| 1180 | : File::kInvalidDescriptor) {} |
| 1181 | |
| 1182 | int GetDescriptor() const override { return m_descriptor; } |
| 1183 | |
| 1184 | Status Write(const void *buf, size_t &num_bytes) override { |
| 1185 | GIL takeGIL; |
| 1186 | PyObject *pybuffer_p = PyMemoryView_FromMemory( |
| 1187 | mem: const_cast<char *>((const char *)buf), size: num_bytes, PyBUF_READ); |
| 1188 | if (!pybuffer_p) |
| 1189 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1190 | return Status::FromError(error: llvm::make_error<PythonException>()).Clone(); |
| 1191 | auto pybuffer = Take<PythonObject>(obj: pybuffer_p); |
| 1192 | num_bytes = 0; |
| 1193 | auto bytes_written = As<long long>(obj: m_py_obj.CallMethod(name: "write" , t: pybuffer)); |
| 1194 | if (!bytes_written) |
| 1195 | return Status::FromError(error: bytes_written.takeError()); |
| 1196 | if (bytes_written.get() < 0) |
| 1197 | return Status::FromErrorString( |
| 1198 | str: ".write() method returned a negative number!" ); |
| 1199 | static_assert(sizeof(long long) >= sizeof(size_t), "overflow" ); |
| 1200 | num_bytes = bytes_written.get(); |
| 1201 | return Status(); |
| 1202 | } |
| 1203 | |
| 1204 | Status Read(void *buf, size_t &num_bytes) override { |
| 1205 | GIL takeGIL; |
| 1206 | static_assert(sizeof(long long) >= sizeof(size_t), "overflow" ); |
| 1207 | auto pybuffer_obj = |
| 1208 | m_py_obj.CallMethod(name: "read" , t: (unsigned long long)num_bytes); |
| 1209 | if (!pybuffer_obj) |
| 1210 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1211 | return Status::FromError(error: pybuffer_obj.takeError()).Clone(); |
| 1212 | num_bytes = 0; |
| 1213 | if (pybuffer_obj.get().IsNone()) { |
| 1214 | // EOF |
| 1215 | num_bytes = 0; |
| 1216 | return Status(); |
| 1217 | } |
| 1218 | auto pybuffer = PythonBuffer::Create(obj&: pybuffer_obj.get()); |
| 1219 | if (!pybuffer) |
| 1220 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1221 | return Status::FromError(error: pybuffer.takeError()).Clone(); |
| 1222 | memcpy(dest: buf, src: pybuffer.get().get().buf, n: pybuffer.get().get().len); |
| 1223 | num_bytes = pybuffer.get().get().len; |
| 1224 | return Status(); |
| 1225 | } |
| 1226 | }; |
| 1227 | } // namespace |
| 1228 | |
| 1229 | namespace { |
| 1230 | class TextPythonFile : public PythonIOFile { |
| 1231 | protected: |
| 1232 | int m_descriptor; |
| 1233 | |
| 1234 | public: |
| 1235 | TextPythonFile(int fd, const PythonFile &file, bool borrowed) |
| 1236 | : PythonIOFile(file, borrowed), |
| 1237 | m_descriptor(File::DescriptorIsValid(descriptor: fd) ? fd |
| 1238 | : File::kInvalidDescriptor) {} |
| 1239 | |
| 1240 | int GetDescriptor() const override { return m_descriptor; } |
| 1241 | |
| 1242 | Status Write(const void *buf, size_t &num_bytes) override { |
| 1243 | GIL takeGIL; |
| 1244 | auto pystring = |
| 1245 | PythonString::FromUTF8(string: llvm::StringRef((const char *)buf, num_bytes)); |
| 1246 | if (!pystring) |
| 1247 | return Status::FromError(error: pystring.takeError()); |
| 1248 | num_bytes = 0; |
| 1249 | auto bytes_written = |
| 1250 | As<long long>(obj: m_py_obj.CallMethod(name: "write" , t: pystring.get())); |
| 1251 | if (!bytes_written) |
| 1252 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1253 | return Status::FromError(error: bytes_written.takeError()).Clone(); |
| 1254 | if (bytes_written.get() < 0) |
| 1255 | return Status::FromErrorString( |
| 1256 | str: ".write() method returned a negative number!" ); |
| 1257 | static_assert(sizeof(long long) >= sizeof(size_t), "overflow" ); |
| 1258 | num_bytes = bytes_written.get(); |
| 1259 | return Status(); |
| 1260 | } |
| 1261 | |
| 1262 | Status Read(void *buf, size_t &num_bytes) override { |
| 1263 | GIL takeGIL; |
| 1264 | size_t num_chars = num_bytes / 6; |
| 1265 | size_t orig_num_bytes = num_bytes; |
| 1266 | num_bytes = 0; |
| 1267 | if (orig_num_bytes < 6) { |
| 1268 | return Status::FromErrorString( |
| 1269 | str: "can't read less than 6 bytes from a utf8 text stream" ); |
| 1270 | } |
| 1271 | auto pystring = As<PythonString>( |
| 1272 | obj: m_py_obj.CallMethod(name: "read" , t: (unsigned long long)num_chars)); |
| 1273 | if (!pystring) |
| 1274 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1275 | return Status::FromError(error: pystring.takeError()).Clone(); |
| 1276 | if (pystring.get().IsNone()) { |
| 1277 | // EOF |
| 1278 | return Status(); |
| 1279 | } |
| 1280 | auto stringref = pystring.get().AsUTF8(); |
| 1281 | if (!stringref) |
| 1282 | // Cloning since the wrapped exception may still reference the PyThread. |
| 1283 | return Status::FromError(error: stringref.takeError()).Clone(); |
| 1284 | num_bytes = stringref.get().size(); |
| 1285 | memcpy(dest: buf, src: stringref.get().begin(), n: num_bytes); |
| 1286 | return Status(); |
| 1287 | } |
| 1288 | }; |
| 1289 | } // namespace |
| 1290 | |
| 1291 | llvm::Expected<FileSP> PythonFile::ConvertToFile(bool borrowed) { |
| 1292 | if (!IsValid()) |
| 1293 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
| 1294 | S: "invalid PythonFile" ); |
| 1295 | |
| 1296 | int fd = PyObject_AsFileDescriptor(m_py_obj); |
| 1297 | if (fd < 0) { |
| 1298 | PyErr_Clear(); |
| 1299 | return ConvertToFileForcingUseOfScriptingIOMethods(borrowed); |
| 1300 | } |
| 1301 | auto options = GetOptionsForPyObject(obj: *this); |
| 1302 | if (!options) |
| 1303 | return options.takeError(); |
| 1304 | |
| 1305 | File::OpenOptions rw = |
| 1306 | options.get() & (File::eOpenOptionReadOnly | File::eOpenOptionWriteOnly | |
| 1307 | File::eOpenOptionReadWrite); |
| 1308 | if (rw == File::eOpenOptionWriteOnly || rw == File::eOpenOptionReadWrite) { |
| 1309 | // LLDB and python will not share I/O buffers. We should probably |
| 1310 | // flush the python buffers now. |
| 1311 | auto r = CallMethod(name: "flush" ); |
| 1312 | if (!r) |
| 1313 | return r.takeError(); |
| 1314 | } |
| 1315 | |
| 1316 | FileSP file_sp; |
| 1317 | if (borrowed) { |
| 1318 | // In this case we don't need to retain the python |
| 1319 | // object at all. |
| 1320 | file_sp = std::make_shared<NativeFile>(args&: fd, args&: options.get(), args: false); |
| 1321 | } else { |
| 1322 | file_sp = std::static_pointer_cast<File>( |
| 1323 | r: std::make_shared<SimplePythonFile>(args&: *this, args&: borrowed, args&: fd, args&: options.get())); |
| 1324 | } |
| 1325 | if (!file_sp->IsValid()) |
| 1326 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
| 1327 | S: "invalid File" ); |
| 1328 | |
| 1329 | return file_sp; |
| 1330 | } |
| 1331 | |
| 1332 | llvm::Expected<FileSP> |
| 1333 | PythonFile::ConvertToFileForcingUseOfScriptingIOMethods(bool borrowed) { |
| 1334 | |
| 1335 | assert(!PyErr_Occurred()); |
| 1336 | |
| 1337 | if (!IsValid()) |
| 1338 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
| 1339 | S: "invalid PythonFile" ); |
| 1340 | |
| 1341 | int fd = PyObject_AsFileDescriptor(m_py_obj); |
| 1342 | if (fd < 0) { |
| 1343 | PyErr_Clear(); |
| 1344 | fd = File::kInvalidDescriptor; |
| 1345 | } |
| 1346 | |
| 1347 | auto io_module = PythonModule::Import(name: "io" ); |
| 1348 | if (!io_module) |
| 1349 | return io_module.takeError(); |
| 1350 | auto textIOBase = io_module.get().Get(name: "TextIOBase" ); |
| 1351 | if (!textIOBase) |
| 1352 | return textIOBase.takeError(); |
| 1353 | auto rawIOBase = io_module.get().Get(name: "RawIOBase" ); |
| 1354 | if (!rawIOBase) |
| 1355 | return rawIOBase.takeError(); |
| 1356 | auto bufferedIOBase = io_module.get().Get(name: "BufferedIOBase" ); |
| 1357 | if (!bufferedIOBase) |
| 1358 | return bufferedIOBase.takeError(); |
| 1359 | |
| 1360 | FileSP file_sp; |
| 1361 | |
| 1362 | auto isTextIO = IsInstance(cls: textIOBase.get()); |
| 1363 | if (!isTextIO) |
| 1364 | return isTextIO.takeError(); |
| 1365 | if (isTextIO.get()) |
| 1366 | file_sp = std::static_pointer_cast<File>( |
| 1367 | r: std::make_shared<TextPythonFile>(args&: fd, args&: *this, args&: borrowed)); |
| 1368 | |
| 1369 | auto isRawIO = IsInstance(cls: rawIOBase.get()); |
| 1370 | if (!isRawIO) |
| 1371 | return isRawIO.takeError(); |
| 1372 | auto isBufferedIO = IsInstance(cls: bufferedIOBase.get()); |
| 1373 | if (!isBufferedIO) |
| 1374 | return isBufferedIO.takeError(); |
| 1375 | |
| 1376 | if (isRawIO.get() || isBufferedIO.get()) { |
| 1377 | file_sp = std::static_pointer_cast<File>( |
| 1378 | r: std::make_shared<BinaryPythonFile>(args&: fd, args&: *this, args&: borrowed)); |
| 1379 | } |
| 1380 | |
| 1381 | if (!file_sp) |
| 1382 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
| 1383 | S: "python file is neither text nor binary" ); |
| 1384 | |
| 1385 | if (!file_sp->IsValid()) |
| 1386 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
| 1387 | S: "invalid File" ); |
| 1388 | |
| 1389 | return file_sp; |
| 1390 | } |
| 1391 | |
| 1392 | Expected<PythonFile> PythonFile::FromFile(File &file, const char *mode) { |
| 1393 | if (!file.IsValid()) |
| 1394 | return llvm::createStringError(EC: llvm::inconvertibleErrorCode(), |
| 1395 | S: "invalid file" ); |
| 1396 | |
| 1397 | if (auto *simple = llvm::dyn_cast<SimplePythonFile>(Val: &file)) |
| 1398 | return Retain<PythonFile>(obj: simple->GetPythonObject()); |
| 1399 | if (auto *pythonio = llvm::dyn_cast<PythonIOFile>(Val: &file)) |
| 1400 | return Retain<PythonFile>(obj: pythonio->GetPythonObject()); |
| 1401 | |
| 1402 | if (!mode) { |
| 1403 | auto m = file.GetOpenMode(); |
| 1404 | if (!m) |
| 1405 | return m.takeError(); |
| 1406 | mode = m.get(); |
| 1407 | } |
| 1408 | |
| 1409 | PyObject *file_obj; |
| 1410 | file_obj = PyFile_FromFd(file.GetDescriptor(), nullptr, mode, -1, nullptr, |
| 1411 | "ignore" , nullptr, /*closefd=*/0); |
| 1412 | |
| 1413 | if (!file_obj) |
| 1414 | return exception(); |
| 1415 | |
| 1416 | return Take<PythonFile>(obj: file_obj); |
| 1417 | } |
| 1418 | |
| 1419 | Error PythonScript::Init() { |
| 1420 | if (function.IsValid()) |
| 1421 | return Error::success(); |
| 1422 | |
| 1423 | PythonDictionary globals(PyInitialValue::Empty); |
| 1424 | auto builtins = PythonModule::BuiltinsModule(); |
| 1425 | if (Error error = globals.SetItem(key: "__builtins__" , value: builtins)) |
| 1426 | return error; |
| 1427 | PyObject *o = |
| 1428 | PyRun_String(script, Py_file_input, globals.get(), globals.get()); |
| 1429 | if (!o) |
| 1430 | return exception(); |
| 1431 | Take<PythonObject>(obj: o); |
| 1432 | auto f = As<PythonCallable>(obj: globals.GetItem(key: "main" )); |
| 1433 | if (!f) |
| 1434 | return f.takeError(); |
| 1435 | function = std::move(f.get()); |
| 1436 | |
| 1437 | return Error::success(); |
| 1438 | } |
| 1439 | |
| 1440 | llvm::Expected<PythonObject> |
| 1441 | python::runStringOneLine(const llvm::Twine &string, |
| 1442 | const PythonDictionary &globals, |
| 1443 | const PythonDictionary &locals) { |
| 1444 | if (!globals.IsValid() || !locals.IsValid()) |
| 1445 | return nullDeref(); |
| 1446 | |
| 1447 | PyObject *code = |
| 1448 | Py_CompileString(NullTerminated(string), "<string>" , Py_eval_input); |
| 1449 | if (!code) { |
| 1450 | PyErr_Clear(); |
| 1451 | code = |
| 1452 | Py_CompileString(NullTerminated(string), "<string>" , Py_single_input); |
| 1453 | } |
| 1454 | if (!code) |
| 1455 | return exception(); |
| 1456 | auto code_ref = Take<PythonObject>(obj: code); |
| 1457 | |
| 1458 | PyObject *result = PyEval_EvalCode(code, globals.get(), locals.get()); |
| 1459 | |
| 1460 | if (!result) |
| 1461 | return exception(); |
| 1462 | |
| 1463 | return Take<PythonObject>(obj: result); |
| 1464 | } |
| 1465 | |
| 1466 | llvm::Expected<PythonObject> |
| 1467 | python::runStringMultiLine(const llvm::Twine &string, |
| 1468 | const PythonDictionary &globals, |
| 1469 | const PythonDictionary &locals) { |
| 1470 | if (!globals.IsValid() || !locals.IsValid()) |
| 1471 | return nullDeref(); |
| 1472 | PyObject *result = PyRun_String(NullTerminated(string), Py_file_input, |
| 1473 | globals.get(), locals.get()); |
| 1474 | if (!result) |
| 1475 | return exception(); |
| 1476 | return Take<PythonObject>(obj: result); |
| 1477 | } |
| 1478 | |
| 1479 | #endif |
| 1480 | |