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/Log.h" 20 #include "lldb/Utility/Stream.h" 21 22 #include "llvm/ADT/StringSwitch.h" 23 #include "llvm/Support/Casting.h" 24 #include "llvm/Support/ConvertUTF.h" 25 #include "llvm/Support/Errno.h" 26 27 #include <stdio.h> 28 29 using namespace lldb_private; 30 using namespace lldb; 31 using namespace lldb_private::python; 32 using llvm::cantFail; 33 using llvm::Error; 34 using llvm::Expected; 35 using llvm::Twine; 36 37 template <> Expected<bool> python::As<bool>(Expected<PythonObject> &&obj) { 38 if (!obj) 39 return obj.takeError(); 40 return obj.get().IsTrue(); 41 } 42 43 template <> 44 Expected<long long> python::As<long long>(Expected<PythonObject> &&obj) { 45 if (!obj) 46 return obj.takeError(); 47 return obj->AsLongLong(); 48 } 49 50 template <> 51 Expected<unsigned long long> 52 python::As<unsigned long long>(Expected<PythonObject> &&obj) { 53 if (!obj) 54 return obj.takeError(); 55 return obj->AsUnsignedLongLong(); 56 } 57 58 template <> 59 Expected<std::string> python::As<std::string>(Expected<PythonObject> &&obj) { 60 if (!obj) 61 return obj.takeError(); 62 PyObject *str_obj = PyObject_Str(obj.get().get()); 63 if (!obj) 64 return llvm::make_error<PythonException>(); 65 auto str = Take<PythonString>(str_obj); 66 auto utf8 = str.AsUTF8(); 67 if (!utf8) 68 return utf8.takeError(); 69 return std::string(utf8.get()); 70 } 71 72 Expected<long long> PythonObject::AsLongLong() const { 73 if (!m_py_obj) 74 return nullDeref(); 75 #if PY_MAJOR_VERSION < 3 76 if (!PyLong_Check(m_py_obj)) { 77 PythonInteger i(PyRefType::Borrowed, m_py_obj); 78 return i.AsLongLong(); 79 } 80 #endif 81 assert(!PyErr_Occurred()); 82 long long r = PyLong_AsLongLong(m_py_obj); 83 if (PyErr_Occurred()) 84 return exception(); 85 return r; 86 } 87 88 Expected<long long> PythonObject::AsUnsignedLongLong() const { 89 if (!m_py_obj) 90 return nullDeref(); 91 #if PY_MAJOR_VERSION < 3 92 if (!PyLong_Check(m_py_obj)) { 93 PythonInteger i(PyRefType::Borrowed, m_py_obj); 94 return i.AsUnsignedLongLong(); 95 } 96 #endif 97 assert(!PyErr_Occurred()); 98 long long r = PyLong_AsUnsignedLongLong(m_py_obj); 99 if (PyErr_Occurred()) 100 return exception(); 101 return r; 102 } 103 104 // wraps on overflow, instead of raising an error. 105 Expected<unsigned long long> PythonObject::AsModuloUnsignedLongLong() const { 106 if (!m_py_obj) 107 return nullDeref(); 108 #if PY_MAJOR_VERSION < 3 109 if (!PyLong_Check(m_py_obj)) { 110 PythonInteger i(PyRefType::Borrowed, m_py_obj); 111 return i.AsModuloUnsignedLongLong(); 112 } 113 #endif 114 assert(!PyErr_Occurred()); 115 unsigned long long r = PyLong_AsUnsignedLongLongMask(m_py_obj); 116 if (PyErr_Occurred()) 117 return exception(); 118 return r; 119 } 120 121 void StructuredPythonObject::Serialize(llvm::json::OStream &s) const { 122 s.value(llvm::formatv("Python Obj: {0:X}", GetValue()).str()); 123 } 124 125 // PythonObject 126 127 void PythonObject::Dump(Stream &strm) const { 128 if (m_py_obj) { 129 FILE *file = llvm::sys::RetryAfterSignal(nullptr, ::tmpfile); 130 if (file) { 131 ::PyObject_Print(m_py_obj, file, 0); 132 const long length = ftell(file); 133 if (length) { 134 ::rewind(file); 135 std::vector<char> file_contents(length, '\0'); 136 const size_t length_read = 137 ::fread(file_contents.data(), 1, file_contents.size(), file); 138 if (length_read > 0) 139 strm.Write(file_contents.data(), length_read); 140 } 141 ::fclose(file); 142 } 143 } else 144 strm.PutCString("NULL"); 145 } 146 147 PyObjectType PythonObject::GetObjectType() const { 148 if (!IsAllocated()) 149 return PyObjectType::None; 150 151 if (PythonModule::Check(m_py_obj)) 152 return PyObjectType::Module; 153 if (PythonList::Check(m_py_obj)) 154 return PyObjectType::List; 155 if (PythonTuple::Check(m_py_obj)) 156 return PyObjectType::Tuple; 157 if (PythonDictionary::Check(m_py_obj)) 158 return PyObjectType::Dictionary; 159 if (PythonString::Check(m_py_obj)) 160 return PyObjectType::String; 161 #if PY_MAJOR_VERSION >= 3 162 if (PythonBytes::Check(m_py_obj)) 163 return PyObjectType::Bytes; 164 #endif 165 if (PythonByteArray::Check(m_py_obj)) 166 return PyObjectType::ByteArray; 167 if (PythonBoolean::Check(m_py_obj)) 168 return PyObjectType::Boolean; 169 if (PythonInteger::Check(m_py_obj)) 170 return PyObjectType::Integer; 171 if (PythonFile::Check(m_py_obj)) 172 return PyObjectType::File; 173 if (PythonCallable::Check(m_py_obj)) 174 return PyObjectType::Callable; 175 return PyObjectType::Unknown; 176 } 177 178 PythonString PythonObject::Repr() const { 179 if (!m_py_obj) 180 return PythonString(); 181 PyObject *repr = PyObject_Repr(m_py_obj); 182 if (!repr) 183 return PythonString(); 184 return PythonString(PyRefType::Owned, repr); 185 } 186 187 PythonString PythonObject::Str() const { 188 if (!m_py_obj) 189 return PythonString(); 190 PyObject *str = PyObject_Str(m_py_obj); 191 if (!str) 192 return PythonString(); 193 return PythonString(PyRefType::Owned, str); 194 } 195 196 PythonObject 197 PythonObject::ResolveNameWithDictionary(llvm::StringRef name, 198 const PythonDictionary &dict) { 199 size_t dot_pos = name.find('.'); 200 llvm::StringRef piece = name.substr(0, dot_pos); 201 PythonObject result = dict.GetItemForKey(PythonString(piece)); 202 if (dot_pos == llvm::StringRef::npos) { 203 // There was no dot, we're done. 204 return result; 205 } 206 207 // There was a dot. The remaining portion of the name should be looked up in 208 // the context of the object that was found in the dictionary. 209 return result.ResolveName(name.substr(dot_pos + 1)); 210 } 211 212 PythonObject PythonObject::ResolveName(llvm::StringRef name) const { 213 // Resolve the name in the context of the specified object. If, for example, 214 // `this` refers to a PyModule, then this will look for `name` in this 215 // module. If `this` refers to a PyType, then it will resolve `name` as an 216 // attribute of that type. If `this` refers to an instance of an object, 217 // then it will resolve `name` as the value of the specified field. 218 // 219 // This function handles dotted names so that, for example, if `m_py_obj` 220 // refers to the `sys` module, and `name` == "path.append", then it will find 221 // the function `sys.path.append`. 222 223 size_t dot_pos = name.find('.'); 224 if (dot_pos == llvm::StringRef::npos) { 225 // No dots in the name, we should be able to find the value immediately as 226 // an attribute of `m_py_obj`. 227 return GetAttributeValue(name); 228 } 229 230 // Look up the first piece of the name, and resolve the rest as a child of 231 // that. 232 PythonObject parent = ResolveName(name.substr(0, dot_pos)); 233 if (!parent.IsAllocated()) 234 return PythonObject(); 235 236 // Tail recursion.. should be optimized by the compiler 237 return parent.ResolveName(name.substr(dot_pos + 1)); 238 } 239 240 bool PythonObject::HasAttribute(llvm::StringRef attr) const { 241 if (!IsValid()) 242 return false; 243 PythonString py_attr(attr); 244 return !!PyObject_HasAttr(m_py_obj, py_attr.get()); 245 } 246 247 PythonObject PythonObject::GetAttributeValue(llvm::StringRef attr) const { 248 if (!IsValid()) 249 return PythonObject(); 250 251 PythonString py_attr(attr); 252 if (!PyObject_HasAttr(m_py_obj, py_attr.get())) 253 return PythonObject(); 254 255 return PythonObject(PyRefType::Owned, 256 PyObject_GetAttr(m_py_obj, py_attr.get())); 257 } 258 259 StructuredData::ObjectSP PythonObject::CreateStructuredObject() const { 260 switch (GetObjectType()) { 261 case PyObjectType::Dictionary: 262 return PythonDictionary(PyRefType::Borrowed, m_py_obj) 263 .CreateStructuredDictionary(); 264 case PyObjectType::Boolean: 265 return PythonBoolean(PyRefType::Borrowed, m_py_obj) 266 .CreateStructuredBoolean(); 267 case PyObjectType::Integer: 268 return PythonInteger(PyRefType::Borrowed, m_py_obj) 269 .CreateStructuredInteger(); 270 case PyObjectType::List: 271 return PythonList(PyRefType::Borrowed, m_py_obj).CreateStructuredArray(); 272 case PyObjectType::String: 273 return PythonString(PyRefType::Borrowed, m_py_obj).CreateStructuredString(); 274 case PyObjectType::Bytes: 275 return PythonBytes(PyRefType::Borrowed, m_py_obj).CreateStructuredString(); 276 case PyObjectType::ByteArray: 277 return PythonByteArray(PyRefType::Borrowed, m_py_obj) 278 .CreateStructuredString(); 279 case PyObjectType::None: 280 return StructuredData::ObjectSP(); 281 default: 282 return StructuredData::ObjectSP(new StructuredPythonObject(m_py_obj)); 283 } 284 } 285 286 // PythonString 287 288 PythonBytes::PythonBytes(llvm::ArrayRef<uint8_t> bytes) { SetBytes(bytes); } 289 290 PythonBytes::PythonBytes(const uint8_t *bytes, size_t length) { 291 SetBytes(llvm::ArrayRef<uint8_t>(bytes, length)); 292 } 293 294 bool PythonBytes::Check(PyObject *py_obj) { 295 if (!py_obj) 296 return false; 297 return PyBytes_Check(py_obj); 298 } 299 300 llvm::ArrayRef<uint8_t> PythonBytes::GetBytes() const { 301 if (!IsValid()) 302 return llvm::ArrayRef<uint8_t>(); 303 304 Py_ssize_t size; 305 char *c; 306 307 PyBytes_AsStringAndSize(m_py_obj, &c, &size); 308 return llvm::ArrayRef<uint8_t>(reinterpret_cast<uint8_t *>(c), size); 309 } 310 311 size_t PythonBytes::GetSize() const { 312 if (!IsValid()) 313 return 0; 314 return PyBytes_Size(m_py_obj); 315 } 316 317 void PythonBytes::SetBytes(llvm::ArrayRef<uint8_t> bytes) { 318 const char *data = reinterpret_cast<const char *>(bytes.data()); 319 *this = Take<PythonBytes>(PyBytes_FromStringAndSize(data, bytes.size())); 320 } 321 322 StructuredData::StringSP PythonBytes::CreateStructuredString() const { 323 StructuredData::StringSP result(new StructuredData::String); 324 Py_ssize_t size; 325 char *c; 326 PyBytes_AsStringAndSize(m_py_obj, &c, &size); 327 result->SetValue(std::string(c, size)); 328 return result; 329 } 330 331 PythonByteArray::PythonByteArray(llvm::ArrayRef<uint8_t> bytes) 332 : PythonByteArray(bytes.data(), bytes.size()) {} 333 334 PythonByteArray::PythonByteArray(const uint8_t *bytes, size_t length) { 335 const char *str = reinterpret_cast<const char *>(bytes); 336 *this = Take<PythonByteArray>(PyByteArray_FromStringAndSize(str, length)); 337 } 338 339 bool PythonByteArray::Check(PyObject *py_obj) { 340 if (!py_obj) 341 return false; 342 return PyByteArray_Check(py_obj); 343 } 344 345 llvm::ArrayRef<uint8_t> PythonByteArray::GetBytes() const { 346 if (!IsValid()) 347 return llvm::ArrayRef<uint8_t>(); 348 349 char *c = PyByteArray_AsString(m_py_obj); 350 size_t size = GetSize(); 351 return llvm::ArrayRef<uint8_t>(reinterpret_cast<uint8_t *>(c), size); 352 } 353 354 size_t PythonByteArray::GetSize() const { 355 if (!IsValid()) 356 return 0; 357 358 return PyByteArray_Size(m_py_obj); 359 } 360 361 StructuredData::StringSP PythonByteArray::CreateStructuredString() const { 362 StructuredData::StringSP result(new StructuredData::String); 363 llvm::ArrayRef<uint8_t> bytes = GetBytes(); 364 const char *str = reinterpret_cast<const char *>(bytes.data()); 365 result->SetValue(std::string(str, bytes.size())); 366 return result; 367 } 368 369 // PythonString 370 371 Expected<PythonString> PythonString::FromUTF8(llvm::StringRef string) { 372 #if PY_MAJOR_VERSION >= 3 373 PyObject *str = PyUnicode_FromStringAndSize(string.data(), string.size()); 374 #else 375 PyObject *str = PyString_FromStringAndSize(string.data(), string.size()); 376 #endif 377 if (!str) 378 return llvm::make_error<PythonException>(); 379 return Take<PythonString>(str); 380 } 381 382 PythonString::PythonString(llvm::StringRef string) { SetString(string); } 383 384 bool PythonString::Check(PyObject *py_obj) { 385 if (!py_obj) 386 return false; 387 388 if (PyUnicode_Check(py_obj)) 389 return true; 390 #if PY_MAJOR_VERSION < 3 391 if (PyString_Check(py_obj)) 392 return true; 393 #endif 394 return false; 395 } 396 397 void PythonString::Convert(PyRefType &type, PyObject *&py_obj) { 398 #if PY_MAJOR_VERSION < 3 399 // In Python 2, Don't store PyUnicode objects directly, because we need 400 // access to their underlying character buffers which Python 2 doesn't 401 // provide. 402 if (PyUnicode_Check(py_obj)) { 403 PyObject *s = PyUnicode_AsUTF8String(py_obj); 404 if (s == nullptr) { 405 PyErr_Clear(); 406 if (type == PyRefType::Owned) 407 Py_DECREF(py_obj); 408 return; 409 } 410 if (type == PyRefType::Owned) 411 Py_DECREF(py_obj); 412 else 413 type = PyRefType::Owned; 414 py_obj = s; 415 } 416 #endif 417 } 418 419 llvm::StringRef PythonString::GetString() const { 420 auto s = AsUTF8(); 421 if (!s) { 422 llvm::consumeError(s.takeError()); 423 return llvm::StringRef(""); 424 } 425 return s.get(); 426 } 427 428 Expected<llvm::StringRef> PythonString::AsUTF8() const { 429 if (!IsValid()) 430 return nullDeref(); 431 432 Py_ssize_t size; 433 const char *data; 434 435 #if PY_MAJOR_VERSION >= 3 436 data = PyUnicode_AsUTF8AndSize(m_py_obj, &size); 437 #else 438 char *c = NULL; 439 int r = PyString_AsStringAndSize(m_py_obj, &c, &size); 440 if (r < 0) 441 c = NULL; 442 data = c; 443 #endif 444 445 if (!data) 446 return exception(); 447 448 return llvm::StringRef(data, size); 449 } 450 451 size_t PythonString::GetSize() const { 452 if (IsValid()) { 453 #if PY_MAJOR_VERSION >= 3 454 return PyUnicode_GetSize(m_py_obj); 455 #else 456 return PyString_Size(m_py_obj); 457 #endif 458 } 459 return 0; 460 } 461 462 void PythonString::SetString(llvm::StringRef string) { 463 auto s = FromUTF8(string); 464 if (!s) { 465 llvm::consumeError(s.takeError()); 466 Reset(); 467 } else { 468 *this = std::move(s.get()); 469 } 470 } 471 472 StructuredData::StringSP PythonString::CreateStructuredString() const { 473 StructuredData::StringSP result(new StructuredData::String); 474 result->SetValue(GetString()); 475 return result; 476 } 477 478 // PythonInteger 479 480 PythonInteger::PythonInteger(int64_t value) { SetInteger(value); } 481 482 bool PythonInteger::Check(PyObject *py_obj) { 483 if (!py_obj) 484 return false; 485 486 #if PY_MAJOR_VERSION >= 3 487 // Python 3 does not have PyInt_Check. There is only one type of integral 488 // value, long. 489 return PyLong_Check(py_obj); 490 #else 491 return PyLong_Check(py_obj) || PyInt_Check(py_obj); 492 #endif 493 } 494 495 void PythonInteger::Convert(PyRefType &type, PyObject *&py_obj) { 496 #if PY_MAJOR_VERSION < 3 497 // Always store this as a PyLong, which makes interoperability between Python 498 // 2.x and Python 3.x easier. This is only necessary in 2.x, since 3.x 499 // doesn't even have a PyInt. 500 if (PyInt_Check(py_obj)) { 501 // Since we converted the original object to a different type, the new 502 // object is an owned object regardless of the ownership semantics 503 // requested by the user. 504 long long value = PyInt_AsLong(py_obj); 505 PyObject *l = nullptr; 506 if (!PyErr_Occurred()) 507 l = PyLong_FromLongLong(value); 508 if (l == nullptr) { 509 PyErr_Clear(); 510 if (type == PyRefType::Owned) 511 Py_DECREF(py_obj); 512 return; 513 } 514 if (type == PyRefType::Owned) 515 Py_DECREF(py_obj); 516 else 517 type = PyRefType::Owned; 518 py_obj = l; 519 } 520 #endif 521 } 522 523 void PythonInteger::SetInteger(int64_t value) { 524 *this = Take<PythonInteger>(PyLong_FromLongLong(value)); 525 } 526 527 StructuredData::IntegerSP PythonInteger::CreateStructuredInteger() const { 528 StructuredData::IntegerSP result(new StructuredData::Integer); 529 // FIXME this is really not ideal. Errors are silently converted to 0 530 // and overflows are silently wrapped. But we'd need larger changes 531 // to StructuredData to fix it, so that's how it is for now. 532 llvm::Expected<unsigned long long> value = AsModuloUnsignedLongLong(); 533 if (!value) { 534 llvm::consumeError(value.takeError()); 535 result->SetValue(0); 536 } else { 537 result->SetValue(value.get()); 538 } 539 return result; 540 } 541 542 // PythonBoolean 543 544 PythonBoolean::PythonBoolean(bool value) { 545 SetValue(value); 546 } 547 548 bool PythonBoolean::Check(PyObject *py_obj) { 549 return py_obj ? PyBool_Check(py_obj) : false; 550 } 551 552 bool PythonBoolean::GetValue() const { 553 return m_py_obj ? PyObject_IsTrue(m_py_obj) : false; 554 } 555 556 void PythonBoolean::SetValue(bool value) { 557 *this = Take<PythonBoolean>(PyBool_FromLong(value)); 558 } 559 560 StructuredData::BooleanSP PythonBoolean::CreateStructuredBoolean() const { 561 StructuredData::BooleanSP result(new StructuredData::Boolean); 562 result->SetValue(GetValue()); 563 return result; 564 } 565 566 // PythonList 567 568 PythonList::PythonList(PyInitialValue value) { 569 if (value == PyInitialValue::Empty) 570 *this = Take<PythonList>(PyList_New(0)); 571 } 572 573 PythonList::PythonList(int list_size) { 574 *this = Take<PythonList>(PyList_New(list_size)); 575 } 576 577 bool PythonList::Check(PyObject *py_obj) { 578 if (!py_obj) 579 return false; 580 return PyList_Check(py_obj); 581 } 582 583 uint32_t PythonList::GetSize() const { 584 if (IsValid()) 585 return PyList_GET_SIZE(m_py_obj); 586 return 0; 587 } 588 589 PythonObject PythonList::GetItemAtIndex(uint32_t index) const { 590 if (IsValid()) 591 return PythonObject(PyRefType::Borrowed, PyList_GetItem(m_py_obj, index)); 592 return PythonObject(); 593 } 594 595 void PythonList::SetItemAtIndex(uint32_t index, const PythonObject &object) { 596 if (IsAllocated() && object.IsValid()) { 597 // PyList_SetItem is documented to "steal" a reference, so we need to 598 // convert it to an owned reference by incrementing it. 599 Py_INCREF(object.get()); 600 PyList_SetItem(m_py_obj, index, object.get()); 601 } 602 } 603 604 void PythonList::AppendItem(const PythonObject &object) { 605 if (IsAllocated() && object.IsValid()) { 606 // `PyList_Append` does *not* steal a reference, so do not call `Py_INCREF` 607 // here like we do with `PyList_SetItem`. 608 PyList_Append(m_py_obj, object.get()); 609 } 610 } 611 612 StructuredData::ArraySP PythonList::CreateStructuredArray() const { 613 StructuredData::ArraySP result(new StructuredData::Array); 614 uint32_t count = GetSize(); 615 for (uint32_t i = 0; i < count; ++i) { 616 PythonObject obj = GetItemAtIndex(i); 617 result->AddItem(obj.CreateStructuredObject()); 618 } 619 return result; 620 } 621 622 // PythonTuple 623 624 PythonTuple::PythonTuple(PyInitialValue value) { 625 if (value == PyInitialValue::Empty) 626 *this = Take<PythonTuple>(PyTuple_New(0)); 627 } 628 629 PythonTuple::PythonTuple(int tuple_size) { 630 *this = Take<PythonTuple>(PyTuple_New(tuple_size)); 631 } 632 633 PythonTuple::PythonTuple(std::initializer_list<PythonObject> objects) { 634 m_py_obj = PyTuple_New(objects.size()); 635 636 uint32_t idx = 0; 637 for (auto object : objects) { 638 if (object.IsValid()) 639 SetItemAtIndex(idx, object); 640 idx++; 641 } 642 } 643 644 PythonTuple::PythonTuple(std::initializer_list<PyObject *> objects) { 645 m_py_obj = PyTuple_New(objects.size()); 646 647 uint32_t idx = 0; 648 for (auto py_object : objects) { 649 PythonObject object(PyRefType::Borrowed, py_object); 650 if (object.IsValid()) 651 SetItemAtIndex(idx, object); 652 idx++; 653 } 654 } 655 656 bool PythonTuple::Check(PyObject *py_obj) { 657 if (!py_obj) 658 return false; 659 return PyTuple_Check(py_obj); 660 } 661 662 uint32_t PythonTuple::GetSize() const { 663 if (IsValid()) 664 return PyTuple_GET_SIZE(m_py_obj); 665 return 0; 666 } 667 668 PythonObject PythonTuple::GetItemAtIndex(uint32_t index) const { 669 if (IsValid()) 670 return PythonObject(PyRefType::Borrowed, PyTuple_GetItem(m_py_obj, index)); 671 return PythonObject(); 672 } 673 674 void PythonTuple::SetItemAtIndex(uint32_t index, const PythonObject &object) { 675 if (IsAllocated() && object.IsValid()) { 676 // PyTuple_SetItem is documented to "steal" a reference, so we need to 677 // convert it to an owned reference by incrementing it. 678 Py_INCREF(object.get()); 679 PyTuple_SetItem(m_py_obj, index, object.get()); 680 } 681 } 682 683 StructuredData::ArraySP PythonTuple::CreateStructuredArray() const { 684 StructuredData::ArraySP result(new StructuredData::Array); 685 uint32_t count = GetSize(); 686 for (uint32_t i = 0; i < count; ++i) { 687 PythonObject obj = GetItemAtIndex(i); 688 result->AddItem(obj.CreateStructuredObject()); 689 } 690 return result; 691 } 692 693 // PythonDictionary 694 695 PythonDictionary::PythonDictionary(PyInitialValue value) { 696 if (value == PyInitialValue::Empty) 697 *this = Take<PythonDictionary>(PyDict_New()); 698 } 699 700 bool PythonDictionary::Check(PyObject *py_obj) { 701 if (!py_obj) 702 return false; 703 704 return PyDict_Check(py_obj); 705 } 706 707 uint32_t PythonDictionary::GetSize() const { 708 if (IsValid()) 709 return PyDict_Size(m_py_obj); 710 return 0; 711 } 712 713 PythonList PythonDictionary::GetKeys() const { 714 if (IsValid()) 715 return PythonList(PyRefType::Owned, PyDict_Keys(m_py_obj)); 716 return PythonList(PyInitialValue::Invalid); 717 } 718 719 PythonObject PythonDictionary::GetItemForKey(const PythonObject &key) const { 720 auto item = GetItem(key); 721 if (!item) { 722 llvm::consumeError(item.takeError()); 723 return PythonObject(); 724 } 725 return std::move(item.get()); 726 } 727 728 Expected<PythonObject> 729 PythonDictionary::GetItem(const PythonObject &key) const { 730 if (!IsValid()) 731 return nullDeref(); 732 #if PY_MAJOR_VERSION >= 3 733 PyObject *o = PyDict_GetItemWithError(m_py_obj, key.get()); 734 if (PyErr_Occurred()) 735 return exception(); 736 #else 737 PyObject *o = PyDict_GetItem(m_py_obj, key.get()); 738 #endif 739 if (!o) 740 return keyError(); 741 return Retain<PythonObject>(o); 742 } 743 744 Expected<PythonObject> PythonDictionary::GetItem(const Twine &key) const { 745 if (!IsValid()) 746 return nullDeref(); 747 PyObject *o = PyDict_GetItemString(m_py_obj, NullTerminated(key)); 748 if (PyErr_Occurred()) 749 return exception(); 750 if (!o) 751 return keyError(); 752 return Retain<PythonObject>(o); 753 } 754 755 Error PythonDictionary::SetItem(const PythonObject &key, 756 const PythonObject &value) const { 757 if (!IsValid() || !value.IsValid()) 758 return nullDeref(); 759 int r = PyDict_SetItem(m_py_obj, key.get(), value.get()); 760 if (r < 0) 761 return exception(); 762 return Error::success(); 763 } 764 765 Error PythonDictionary::SetItem(const Twine &key, 766 const PythonObject &value) const { 767 if (!IsValid() || !value.IsValid()) 768 return nullDeref(); 769 int r = PyDict_SetItemString(m_py_obj, NullTerminated(key), value.get()); 770 if (r < 0) 771 return exception(); 772 return Error::success(); 773 } 774 775 void PythonDictionary::SetItemForKey(const PythonObject &key, 776 const PythonObject &value) { 777 Error error = SetItem(key, value); 778 if (error) 779 llvm::consumeError(std::move(error)); 780 } 781 782 StructuredData::DictionarySP 783 PythonDictionary::CreateStructuredDictionary() const { 784 StructuredData::DictionarySP result(new StructuredData::Dictionary); 785 PythonList keys(GetKeys()); 786 uint32_t num_keys = keys.GetSize(); 787 for (uint32_t i = 0; i < num_keys; ++i) { 788 PythonObject key = keys.GetItemAtIndex(i); 789 PythonObject value = GetItemForKey(key); 790 StructuredData::ObjectSP structured_value = value.CreateStructuredObject(); 791 result->AddItem(key.Str().GetString(), structured_value); 792 } 793 return result; 794 } 795 796 PythonModule PythonModule::BuiltinsModule() { 797 #if PY_MAJOR_VERSION >= 3 798 return AddModule("builtins"); 799 #else 800 return AddModule("__builtin__"); 801 #endif 802 } 803 804 PythonModule PythonModule::MainModule() { return AddModule("__main__"); } 805 806 PythonModule PythonModule::AddModule(llvm::StringRef module) { 807 std::string str = module.str(); 808 return PythonModule(PyRefType::Borrowed, PyImport_AddModule(str.c_str())); 809 } 810 811 Expected<PythonModule> PythonModule::Import(const Twine &name) { 812 PyObject *mod = PyImport_ImportModule(NullTerminated(name)); 813 if (!mod) 814 return exception(); 815 return Take<PythonModule>(mod); 816 } 817 818 Expected<PythonObject> PythonModule::Get(const Twine &name) { 819 if (!IsValid()) 820 return nullDeref(); 821 PyObject *dict = PyModule_GetDict(m_py_obj); 822 if (!dict) 823 return exception(); 824 PyObject *item = PyDict_GetItemString(dict, NullTerminated(name)); 825 if (!item) 826 return exception(); 827 return Retain<PythonObject>(item); 828 } 829 830 bool PythonModule::Check(PyObject *py_obj) { 831 if (!py_obj) 832 return false; 833 834 return PyModule_Check(py_obj); 835 } 836 837 PythonDictionary PythonModule::GetDictionary() const { 838 if (!IsValid()) 839 return PythonDictionary(); 840 return Retain<PythonDictionary>(PyModule_GetDict(m_py_obj)); 841 } 842 843 bool PythonCallable::Check(PyObject *py_obj) { 844 if (!py_obj) 845 return false; 846 847 return PyCallable_Check(py_obj); 848 } 849 850 #if PY_MAJOR_VERSION >= 3 && PY_MINOR_VERSION >= 3 851 static const char get_arg_info_script[] = R"( 852 from inspect import signature, Parameter, ismethod 853 from collections import namedtuple 854 ArgInfo = namedtuple('ArgInfo', ['count', 'has_varargs']) 855 def main(f): 856 count = 0 857 varargs = False 858 for parameter in signature(f).parameters.values(): 859 kind = parameter.kind 860 if kind in (Parameter.POSITIONAL_ONLY, 861 Parameter.POSITIONAL_OR_KEYWORD): 862 count += 1 863 elif kind == Parameter.VAR_POSITIONAL: 864 varargs = True 865 elif kind in (Parameter.KEYWORD_ONLY, 866 Parameter.VAR_KEYWORD): 867 pass 868 else: 869 raise Exception(f'unknown parameter kind: {kind}') 870 return ArgInfo(count, varargs) 871 )"; 872 #endif 873 874 Expected<PythonCallable::ArgInfo> PythonCallable::GetArgInfo() const { 875 ArgInfo result = {}; 876 if (!IsValid()) 877 return nullDeref(); 878 879 #if PY_MAJOR_VERSION >= 3 && PY_MINOR_VERSION >= 3 880 881 // no need to synchronize access to this global, we already have the GIL 882 static PythonScript get_arg_info(get_arg_info_script); 883 Expected<PythonObject> pyarginfo = get_arg_info(*this); 884 if (!pyarginfo) 885 return pyarginfo.takeError(); 886 long long count = 887 cantFail(As<long long>(pyarginfo.get().GetAttribute("count"))); 888 bool has_varargs = 889 cantFail(As<bool>(pyarginfo.get().GetAttribute("has_varargs"))); 890 result.max_positional_args = has_varargs ? ArgInfo::UNBOUNDED : count; 891 892 #else 893 PyObject *py_func_obj; 894 bool is_bound_method = false; 895 bool is_class = false; 896 897 if (PyType_Check(m_py_obj) || PyClass_Check(m_py_obj)) { 898 auto init = GetAttribute("__init__"); 899 if (!init) 900 return init.takeError(); 901 py_func_obj = init.get().get(); 902 is_class = true; 903 } else { 904 py_func_obj = m_py_obj; 905 } 906 907 if (PyMethod_Check(py_func_obj)) { 908 py_func_obj = PyMethod_GET_FUNCTION(py_func_obj); 909 PythonObject im_self = GetAttributeValue("im_self"); 910 if (im_self.IsValid() && !im_self.IsNone()) 911 is_bound_method = true; 912 } else { 913 // see if this is a callable object with an __call__ method 914 if (!PyFunction_Check(py_func_obj)) { 915 PythonObject __call__ = GetAttributeValue("__call__"); 916 if (__call__.IsValid()) { 917 auto __callable__ = __call__.AsType<PythonCallable>(); 918 if (__callable__.IsValid()) { 919 py_func_obj = PyMethod_GET_FUNCTION(__callable__.get()); 920 PythonObject im_self = __callable__.GetAttributeValue("im_self"); 921 if (im_self.IsValid() && !im_self.IsNone()) 922 is_bound_method = true; 923 } 924 } 925 } 926 } 927 928 if (!py_func_obj) 929 return result; 930 931 PyCodeObject *code = (PyCodeObject *)PyFunction_GET_CODE(py_func_obj); 932 if (!code) 933 return result; 934 935 auto count = code->co_argcount; 936 bool has_varargs = !!(code->co_flags & CO_VARARGS); 937 result.max_positional_args = 938 has_varargs ? ArgInfo::UNBOUNDED 939 : (count - (int)is_bound_method) - (int)is_class; 940 941 #endif 942 943 return result; 944 } 945 946 constexpr unsigned 947 PythonCallable::ArgInfo::UNBOUNDED; // FIXME delete after c++17 948 949 PythonObject PythonCallable::operator()() { 950 return PythonObject(PyRefType::Owned, PyObject_CallObject(m_py_obj, nullptr)); 951 } 952 953 PythonObject PythonCallable:: 954 operator()(std::initializer_list<PyObject *> args) { 955 PythonTuple arg_tuple(args); 956 return PythonObject(PyRefType::Owned, 957 PyObject_CallObject(m_py_obj, arg_tuple.get())); 958 } 959 960 PythonObject PythonCallable:: 961 operator()(std::initializer_list<PythonObject> args) { 962 PythonTuple arg_tuple(args); 963 return PythonObject(PyRefType::Owned, 964 PyObject_CallObject(m_py_obj, arg_tuple.get())); 965 } 966 967 bool PythonFile::Check(PyObject *py_obj) { 968 if (!py_obj) 969 return false; 970 #if PY_MAJOR_VERSION < 3 971 return PyFile_Check(py_obj); 972 #else 973 // In Python 3, there is no `PyFile_Check`, and in fact PyFile is not even a 974 // first-class object type anymore. `PyFile_FromFd` is just a thin wrapper 975 // over `io.open()`, which returns some object derived from `io.IOBase`. As a 976 // result, the only way to detect a file in Python 3 is to check whether it 977 // inherits from `io.IOBase`. 978 auto io_module = PythonModule::Import("io"); 979 if (!io_module) { 980 llvm::consumeError(io_module.takeError()); 981 return false; 982 } 983 auto iobase = io_module.get().Get("IOBase"); 984 if (!iobase) { 985 llvm::consumeError(iobase.takeError()); 986 return false; 987 } 988 int r = PyObject_IsInstance(py_obj, iobase.get().get()); 989 if (r < 0) { 990 llvm::consumeError(exception()); // clear the exception and log it. 991 return false; 992 } 993 return !!r; 994 #endif 995 } 996 997 namespace { 998 class GIL { 999 public: 1000 GIL() { 1001 m_state = PyGILState_Ensure(); 1002 assert(!PyErr_Occurred()); 1003 } 1004 ~GIL() { PyGILState_Release(m_state); } 1005 1006 protected: 1007 PyGILState_STATE m_state; 1008 }; 1009 } // namespace 1010 1011 const char *PythonException::toCString() const { 1012 if (!m_repr_bytes) 1013 return "unknown exception"; 1014 return PyBytes_AS_STRING(m_repr_bytes); 1015 } 1016 1017 PythonException::PythonException(const char *caller) { 1018 assert(PyErr_Occurred()); 1019 m_exception_type = m_exception = m_traceback = m_repr_bytes = NULL; 1020 PyErr_Fetch(&m_exception_type, &m_exception, &m_traceback); 1021 PyErr_NormalizeException(&m_exception_type, &m_exception, &m_traceback); 1022 PyErr_Clear(); 1023 if (m_exception) { 1024 PyObject *repr = PyObject_Repr(m_exception); 1025 if (repr) { 1026 m_repr_bytes = PyUnicode_AsEncodedString(repr, "utf-8", nullptr); 1027 if (!m_repr_bytes) { 1028 PyErr_Clear(); 1029 } 1030 Py_XDECREF(repr); 1031 } else { 1032 PyErr_Clear(); 1033 } 1034 } 1035 Log *log = GetLogIfAllCategoriesSet(LIBLLDB_LOG_SCRIPT); 1036 if (caller) 1037 LLDB_LOGF(log, "%s failed with exception: %s", caller, toCString()); 1038 else 1039 LLDB_LOGF(log, "python exception: %s", toCString()); 1040 } 1041 void PythonException::Restore() { 1042 if (m_exception_type && m_exception) { 1043 PyErr_Restore(m_exception_type, m_exception, m_traceback); 1044 } else { 1045 PyErr_SetString(PyExc_Exception, toCString()); 1046 } 1047 m_exception_type = m_exception = m_traceback = NULL; 1048 } 1049 1050 PythonException::~PythonException() { 1051 Py_XDECREF(m_exception_type); 1052 Py_XDECREF(m_exception); 1053 Py_XDECREF(m_traceback); 1054 Py_XDECREF(m_repr_bytes); 1055 } 1056 1057 void PythonException::log(llvm::raw_ostream &OS) const { OS << toCString(); } 1058 1059 std::error_code PythonException::convertToErrorCode() const { 1060 return llvm::inconvertibleErrorCode(); 1061 } 1062 1063 bool PythonException::Matches(PyObject *exc) const { 1064 return PyErr_GivenExceptionMatches(m_exception_type, exc); 1065 } 1066 1067 const char read_exception_script[] = R"( 1068 import sys 1069 from traceback import print_exception 1070 if sys.version_info.major < 3: 1071 from StringIO import StringIO 1072 else: 1073 from io import StringIO 1074 def main(exc_type, exc_value, tb): 1075 f = StringIO() 1076 print_exception(exc_type, exc_value, tb, file=f) 1077 return f.getvalue() 1078 )"; 1079 1080 std::string PythonException::ReadBacktrace() const { 1081 1082 if (!m_traceback) 1083 return toCString(); 1084 1085 // no need to synchronize access to this global, we already have the GIL 1086 static PythonScript read_exception(read_exception_script); 1087 1088 Expected<std::string> backtrace = As<std::string>( 1089 read_exception(m_exception_type, m_exception, m_traceback)); 1090 1091 if (!backtrace) { 1092 std::string message = 1093 std::string(toCString()) + "\n" + 1094 "Traceback unavailable, an error occurred while reading it:\n"; 1095 return (message + llvm::toString(backtrace.takeError())); 1096 } 1097 1098 return std::move(backtrace.get()); 1099 } 1100 1101 char PythonException::ID = 0; 1102 1103 llvm::Expected<File::OpenOptions> 1104 GetOptionsForPyObject(const PythonObject &obj) { 1105 #if PY_MAJOR_VERSION >= 3 1106 auto options = File::OpenOptions(0); 1107 auto readable = As<bool>(obj.CallMethod("readable")); 1108 if (!readable) 1109 return readable.takeError(); 1110 auto writable = As<bool>(obj.CallMethod("writable")); 1111 if (!writable) 1112 return writable.takeError(); 1113 if (readable.get()) 1114 options |= File::eOpenOptionRead; 1115 if (writable.get()) 1116 options |= File::eOpenOptionWrite; 1117 return options; 1118 #else 1119 PythonString py_mode = obj.GetAttributeValue("mode").AsType<PythonString>(); 1120 return File::GetOptionsFromMode(py_mode.GetString()); 1121 #endif 1122 } 1123 1124 // Base class template for python files. All it knows how to do 1125 // is hold a reference to the python object and close or flush it 1126 // when the File is closed. 1127 namespace { 1128 template <typename Base> class OwnedPythonFile : public Base { 1129 public: 1130 template <typename... Args> 1131 OwnedPythonFile(const PythonFile &file, bool borrowed, Args... args) 1132 : Base(args...), m_py_obj(file), m_borrowed(borrowed) { 1133 assert(m_py_obj); 1134 } 1135 1136 ~OwnedPythonFile() override { 1137 assert(m_py_obj); 1138 GIL takeGIL; 1139 Close(); 1140 // we need to ensure the python object is released while we still 1141 // hold the GIL 1142 m_py_obj.Reset(); 1143 } 1144 1145 bool IsPythonSideValid() const { 1146 GIL takeGIL; 1147 auto closed = As<bool>(m_py_obj.GetAttribute("closed")); 1148 if (!closed) { 1149 llvm::consumeError(closed.takeError()); 1150 return false; 1151 } 1152 return !closed.get(); 1153 } 1154 1155 bool IsValid() const override { 1156 return IsPythonSideValid() && Base::IsValid(); 1157 } 1158 1159 Status Close() override { 1160 assert(m_py_obj); 1161 Status py_error, base_error; 1162 GIL takeGIL; 1163 if (!m_borrowed) { 1164 auto r = m_py_obj.CallMethod("close"); 1165 if (!r) 1166 py_error = Status(r.takeError()); 1167 } 1168 base_error = Base::Close(); 1169 if (py_error.Fail()) 1170 return py_error; 1171 return base_error; 1172 }; 1173 1174 PyObject *GetPythonObject() const { 1175 assert(m_py_obj.IsValid()); 1176 return m_py_obj.get(); 1177 } 1178 1179 static bool classof(const File *file) = delete; 1180 1181 protected: 1182 PythonFile m_py_obj; 1183 bool m_borrowed; 1184 }; 1185 } // namespace 1186 1187 // A SimplePythonFile is a OwnedPythonFile that just does all I/O as 1188 // a NativeFile 1189 namespace { 1190 class SimplePythonFile : public OwnedPythonFile<NativeFile> { 1191 public: 1192 SimplePythonFile(const PythonFile &file, bool borrowed, int fd, 1193 File::OpenOptions options) 1194 : OwnedPythonFile(file, borrowed, fd, options, false) {} 1195 1196 static char ID; 1197 bool isA(const void *classID) const override { 1198 return classID == &ID || NativeFile::isA(classID); 1199 } 1200 static bool classof(const File *file) { return file->isA(&ID); } 1201 }; 1202 char SimplePythonFile::ID = 0; 1203 } // namespace 1204 1205 #if PY_MAJOR_VERSION >= 3 1206 1207 namespace { 1208 class PythonBuffer { 1209 public: 1210 PythonBuffer &operator=(const PythonBuffer &) = delete; 1211 PythonBuffer(const PythonBuffer &) = delete; 1212 1213 static Expected<PythonBuffer> Create(PythonObject &obj, 1214 int flags = PyBUF_SIMPLE) { 1215 Py_buffer py_buffer = {}; 1216 PyObject_GetBuffer(obj.get(), &py_buffer, flags); 1217 if (!py_buffer.obj) 1218 return llvm::make_error<PythonException>(); 1219 return PythonBuffer(py_buffer); 1220 } 1221 1222 PythonBuffer(PythonBuffer &&other) { 1223 m_buffer = other.m_buffer; 1224 other.m_buffer.obj = nullptr; 1225 } 1226 1227 ~PythonBuffer() { 1228 if (m_buffer.obj) 1229 PyBuffer_Release(&m_buffer); 1230 } 1231 1232 Py_buffer &get() { return m_buffer; } 1233 1234 private: 1235 // takes ownership of the buffer. 1236 PythonBuffer(const Py_buffer &py_buffer) : m_buffer(py_buffer) {} 1237 Py_buffer m_buffer; 1238 }; 1239 } // namespace 1240 1241 // Shared methods between TextPythonFile and BinaryPythonFile 1242 namespace { 1243 class PythonIOFile : public OwnedPythonFile<File> { 1244 public: 1245 PythonIOFile(const PythonFile &file, bool borrowed) 1246 : OwnedPythonFile(file, borrowed) {} 1247 1248 ~PythonIOFile() override { Close(); } 1249 1250 bool IsValid() const override { return IsPythonSideValid(); } 1251 1252 Status Close() override { 1253 assert(m_py_obj); 1254 GIL takeGIL; 1255 if (m_borrowed) 1256 return Flush(); 1257 auto r = m_py_obj.CallMethod("close"); 1258 if (!r) 1259 return Status(r.takeError()); 1260 return Status(); 1261 } 1262 1263 Status Flush() override { 1264 GIL takeGIL; 1265 auto r = m_py_obj.CallMethod("flush"); 1266 if (!r) 1267 return Status(r.takeError()); 1268 return Status(); 1269 } 1270 1271 Expected<File::OpenOptions> GetOptions() const override { 1272 GIL takeGIL; 1273 return GetOptionsForPyObject(m_py_obj); 1274 } 1275 1276 static char ID; 1277 bool isA(const void *classID) const override { 1278 return classID == &ID || File::isA(classID); 1279 } 1280 static bool classof(const File *file) { return file->isA(&ID); } 1281 }; 1282 char PythonIOFile::ID = 0; 1283 } // namespace 1284 1285 namespace { 1286 class BinaryPythonFile : public PythonIOFile { 1287 protected: 1288 int m_descriptor; 1289 1290 public: 1291 BinaryPythonFile(int fd, const PythonFile &file, bool borrowed) 1292 : PythonIOFile(file, borrowed), 1293 m_descriptor(File::DescriptorIsValid(fd) ? fd 1294 : File::kInvalidDescriptor) {} 1295 1296 int GetDescriptor() const override { return m_descriptor; } 1297 1298 Status Write(const void *buf, size_t &num_bytes) override { 1299 GIL takeGIL; 1300 PyObject *pybuffer_p = PyMemoryView_FromMemory( 1301 const_cast<char *>((const char *)buf), num_bytes, PyBUF_READ); 1302 if (!pybuffer_p) 1303 return Status(llvm::make_error<PythonException>()); 1304 auto pybuffer = Take<PythonObject>(pybuffer_p); 1305 num_bytes = 0; 1306 auto bytes_written = As<long long>(m_py_obj.CallMethod("write", pybuffer)); 1307 if (!bytes_written) 1308 return Status(bytes_written.takeError()); 1309 if (bytes_written.get() < 0) 1310 return Status(".write() method returned a negative number!"); 1311 static_assert(sizeof(long long) >= sizeof(size_t), "overflow"); 1312 num_bytes = bytes_written.get(); 1313 return Status(); 1314 } 1315 1316 Status Read(void *buf, size_t &num_bytes) override { 1317 GIL takeGIL; 1318 static_assert(sizeof(long long) >= sizeof(size_t), "overflow"); 1319 auto pybuffer_obj = 1320 m_py_obj.CallMethod("read", (unsigned long long)num_bytes); 1321 if (!pybuffer_obj) 1322 return Status(pybuffer_obj.takeError()); 1323 num_bytes = 0; 1324 if (pybuffer_obj.get().IsNone()) { 1325 // EOF 1326 num_bytes = 0; 1327 return Status(); 1328 } 1329 auto pybuffer = PythonBuffer::Create(pybuffer_obj.get()); 1330 if (!pybuffer) 1331 return Status(pybuffer.takeError()); 1332 memcpy(buf, pybuffer.get().get().buf, pybuffer.get().get().len); 1333 num_bytes = pybuffer.get().get().len; 1334 return Status(); 1335 } 1336 }; 1337 } // namespace 1338 1339 namespace { 1340 class TextPythonFile : public PythonIOFile { 1341 protected: 1342 int m_descriptor; 1343 1344 public: 1345 TextPythonFile(int fd, const PythonFile &file, bool borrowed) 1346 : PythonIOFile(file, borrowed), 1347 m_descriptor(File::DescriptorIsValid(fd) ? fd 1348 : File::kInvalidDescriptor) {} 1349 1350 int GetDescriptor() const override { return m_descriptor; } 1351 1352 Status Write(const void *buf, size_t &num_bytes) override { 1353 GIL takeGIL; 1354 auto pystring = 1355 PythonString::FromUTF8(llvm::StringRef((const char *)buf, num_bytes)); 1356 if (!pystring) 1357 return Status(pystring.takeError()); 1358 num_bytes = 0; 1359 auto bytes_written = 1360 As<long long>(m_py_obj.CallMethod("write", pystring.get())); 1361 if (!bytes_written) 1362 return Status(bytes_written.takeError()); 1363 if (bytes_written.get() < 0) 1364 return Status(".write() method returned a negative number!"); 1365 static_assert(sizeof(long long) >= sizeof(size_t), "overflow"); 1366 num_bytes = bytes_written.get(); 1367 return Status(); 1368 } 1369 1370 Status Read(void *buf, size_t &num_bytes) override { 1371 GIL takeGIL; 1372 size_t num_chars = num_bytes / 6; 1373 size_t orig_num_bytes = num_bytes; 1374 num_bytes = 0; 1375 if (orig_num_bytes < 6) { 1376 return Status("can't read less than 6 bytes from a utf8 text stream"); 1377 } 1378 auto pystring = As<PythonString>( 1379 m_py_obj.CallMethod("read", (unsigned long long)num_chars)); 1380 if (!pystring) 1381 return Status(pystring.takeError()); 1382 if (pystring.get().IsNone()) { 1383 // EOF 1384 return Status(); 1385 } 1386 auto stringref = pystring.get().AsUTF8(); 1387 if (!stringref) 1388 return Status(stringref.takeError()); 1389 num_bytes = stringref.get().size(); 1390 memcpy(buf, stringref.get().begin(), num_bytes); 1391 return Status(); 1392 } 1393 }; 1394 } // namespace 1395 1396 #endif 1397 1398 llvm::Expected<FileSP> PythonFile::ConvertToFile(bool borrowed) { 1399 if (!IsValid()) 1400 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1401 "invalid PythonFile"); 1402 1403 int fd = PyObject_AsFileDescriptor(m_py_obj); 1404 if (fd < 0) { 1405 PyErr_Clear(); 1406 return ConvertToFileForcingUseOfScriptingIOMethods(borrowed); 1407 } 1408 auto options = GetOptionsForPyObject(*this); 1409 if (!options) 1410 return options.takeError(); 1411 1412 if (options.get() & File::eOpenOptionWrite) { 1413 // LLDB and python will not share I/O buffers. We should probably 1414 // flush the python buffers now. 1415 auto r = CallMethod("flush"); 1416 if (!r) 1417 return r.takeError(); 1418 } 1419 1420 FileSP file_sp; 1421 if (borrowed) { 1422 // In this case we we don't need to retain the python 1423 // object at all. 1424 file_sp = std::make_shared<NativeFile>(fd, options.get(), false); 1425 } else { 1426 file_sp = std::static_pointer_cast<File>( 1427 std::make_shared<SimplePythonFile>(*this, borrowed, fd, options.get())); 1428 } 1429 if (!file_sp->IsValid()) 1430 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1431 "invalid File"); 1432 1433 return file_sp; 1434 } 1435 1436 llvm::Expected<FileSP> 1437 PythonFile::ConvertToFileForcingUseOfScriptingIOMethods(bool borrowed) { 1438 1439 assert(!PyErr_Occurred()); 1440 1441 if (!IsValid()) 1442 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1443 "invalid PythonFile"); 1444 1445 #if PY_MAJOR_VERSION < 3 1446 1447 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1448 "not supported on python 2"); 1449 1450 #else 1451 1452 int fd = PyObject_AsFileDescriptor(m_py_obj); 1453 if (fd < 0) { 1454 PyErr_Clear(); 1455 fd = File::kInvalidDescriptor; 1456 } 1457 1458 auto io_module = PythonModule::Import("io"); 1459 if (!io_module) 1460 return io_module.takeError(); 1461 auto textIOBase = io_module.get().Get("TextIOBase"); 1462 if (!textIOBase) 1463 return textIOBase.takeError(); 1464 auto rawIOBase = io_module.get().Get("RawIOBase"); 1465 if (!rawIOBase) 1466 return rawIOBase.takeError(); 1467 auto bufferedIOBase = io_module.get().Get("BufferedIOBase"); 1468 if (!bufferedIOBase) 1469 return bufferedIOBase.takeError(); 1470 1471 FileSP file_sp; 1472 1473 auto isTextIO = IsInstance(textIOBase.get()); 1474 if (!isTextIO) 1475 return isTextIO.takeError(); 1476 if (isTextIO.get()) 1477 file_sp = std::static_pointer_cast<File>( 1478 std::make_shared<TextPythonFile>(fd, *this, borrowed)); 1479 1480 auto isRawIO = IsInstance(rawIOBase.get()); 1481 if (!isRawIO) 1482 return isRawIO.takeError(); 1483 auto isBufferedIO = IsInstance(bufferedIOBase.get()); 1484 if (!isBufferedIO) 1485 return isBufferedIO.takeError(); 1486 1487 if (isRawIO.get() || isBufferedIO.get()) { 1488 file_sp = std::static_pointer_cast<File>( 1489 std::make_shared<BinaryPythonFile>(fd, *this, borrowed)); 1490 } 1491 1492 if (!file_sp) 1493 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1494 "python file is neither text nor binary"); 1495 1496 if (!file_sp->IsValid()) 1497 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1498 "invalid File"); 1499 1500 return file_sp; 1501 1502 #endif 1503 } 1504 1505 Expected<PythonFile> PythonFile::FromFile(File &file, const char *mode) { 1506 if (!file.IsValid()) 1507 return llvm::createStringError(llvm::inconvertibleErrorCode(), 1508 "invalid file"); 1509 1510 if (auto *simple = llvm::dyn_cast<SimplePythonFile>(&file)) 1511 return Retain<PythonFile>(simple->GetPythonObject()); 1512 #if PY_MAJOR_VERSION >= 3 1513 if (auto *pythonio = llvm::dyn_cast<PythonIOFile>(&file)) 1514 return Retain<PythonFile>(pythonio->GetPythonObject()); 1515 #endif 1516 1517 if (!mode) { 1518 auto m = file.GetOpenMode(); 1519 if (!m) 1520 return m.takeError(); 1521 mode = m.get(); 1522 } 1523 1524 PyObject *file_obj; 1525 #if PY_MAJOR_VERSION >= 3 1526 file_obj = PyFile_FromFd(file.GetDescriptor(), nullptr, mode, -1, nullptr, 1527 "ignore", nullptr, /*closefd=*/0); 1528 #else 1529 // I'd like to pass ::fflush here if the file is writable, so that 1530 // when the python side destructs the file object it will be flushed. 1531 // However, this would be dangerous. It can cause fflush to be called 1532 // after fclose if the python program keeps a reference to the file after 1533 // the original lldb_private::File has been destructed. 1534 // 1535 // It's all well and good to ask a python program not to use a closed file 1536 // but asking a python program to make sure objects get released in a 1537 // particular order is not safe. 1538 // 1539 // The tradeoff here is that if a python 2 program wants to make sure this 1540 // file gets flushed, they'll have to do it explicitly or wait untill the 1541 // original lldb File itself gets flushed. 1542 file_obj = PyFile_FromFile(file.GetStream(), py2_const_cast(""), 1543 py2_const_cast(mode), [](FILE *) { return 0; }); 1544 #endif 1545 1546 if (!file_obj) 1547 return exception(); 1548 1549 return Take<PythonFile>(file_obj); 1550 } 1551 1552 Error PythonScript::Init() { 1553 if (function.IsValid()) 1554 return Error::success(); 1555 1556 PythonDictionary globals(PyInitialValue::Empty); 1557 auto builtins = PythonModule::BuiltinsModule(); 1558 if (Error error = globals.SetItem("__builtins__", builtins)) 1559 return error; 1560 PyObject *o = 1561 PyRun_String(script, Py_file_input, globals.get(), globals.get()); 1562 if (!o) 1563 return exception(); 1564 Take<PythonObject>(o); 1565 auto f = As<PythonCallable>(globals.GetItem("main")); 1566 if (!f) 1567 return f.takeError(); 1568 function = std::move(f.get()); 1569 1570 return Error::success(); 1571 } 1572 1573 llvm::Expected<PythonObject> 1574 python::runStringOneLine(const llvm::Twine &string, 1575 const PythonDictionary &globals, 1576 const PythonDictionary &locals) { 1577 if (!globals.IsValid() || !locals.IsValid()) 1578 return nullDeref(); 1579 1580 PyObject *code = 1581 Py_CompileString(NullTerminated(string), "<string>", Py_eval_input); 1582 if (!code) { 1583 PyErr_Clear(); 1584 code = 1585 Py_CompileString(NullTerminated(string), "<string>", Py_single_input); 1586 } 1587 if (!code) 1588 return exception(); 1589 auto code_ref = Take<PythonObject>(code); 1590 1591 #if PY_MAJOR_VERSION < 3 1592 PyObject *result = 1593 PyEval_EvalCode((PyCodeObject *)code, globals.get(), locals.get()); 1594 #else 1595 PyObject *result = PyEval_EvalCode(code, globals.get(), locals.get()); 1596 #endif 1597 1598 if (!result) 1599 return exception(); 1600 1601 return Take<PythonObject>(result); 1602 } 1603 1604 llvm::Expected<PythonObject> 1605 python::runStringMultiLine(const llvm::Twine &string, 1606 const PythonDictionary &globals, 1607 const PythonDictionary &locals) { 1608 if (!globals.IsValid() || !locals.IsValid()) 1609 return nullDeref(); 1610 PyObject *result = PyRun_String(NullTerminated(string), Py_file_input, 1611 globals.get(), locals.get()); 1612 if (!result) 1613 return exception(); 1614 return Take<PythonObject>(result); 1615 } 1616 1617 #endif 1618