Add SourceView which doesn't own source text as base class of Source (pytorch#65309)

Summary:
This would save the cost copying text from stack to heap in some cases (like
parsing function schema during loading phase of libtorch.so)

Pull Request resolved: pytorch#65309

Reviewed By: swolchok

Differential Revision: D31060315

Pulled By: gmagogsfm

fbshipit-source-id: 0caf7a688b40df52bb4388c5191d1a42351d6f1a

Author: gmagogsfm

test/cpp/jit/test_class_import.cpp

@@ -9,7 +9,7 @@
 namespace torch {
 namespace jit {
 
-static const auto classSrcs1 = R"JIT(
+static constexpr c10::string_view classSrcs1 = R"JIT(
 class FooNestedTest:
     def __init__(self, y):
         self.y = y
@@ -26,7 +26,7 @@ class FooTest:
         self.x = self.class_attr.y + self.class_attr2.y
 )JIT";
 
-static const auto classSrcs2 = R"JIT(
+static constexpr c10::string_view classSrcs2 = R"JIT(
 class FooTest:
     def __init__(self, x):
       self.dx = x
@@ -134,7 +134,7 @@ TEST(ClassImportTest, ClassDerive) {
   ASSERT_TRUE(newCls2->findMethod(method->name()));
 }
 
-static const auto torchbindSrc = R"JIT(
+static constexpr c10::string_view torchbindSrc = R"JIT(
 class FooBar1234(Module):
   __parameters__ = []
   f : __torch__.torch.classes._TorchScriptTesting._StackString

test/cpp/jit/test_class_parser.cpp

@@ -5,7 +5,7 @@
 
 namespace torch {
 namespace jit {
-const auto testSource = R"JIT(
+constexpr c10::string_view testSource = R"JIT(
   class FooTest:
     def __init__(self, x):
       self.x = x

test/cpp/jit/test_interface.cpp

@@ -18,12 +18,12 @@ def one(self, x: Tensor, y: Tensor) -> Tensor:
 def forward(self, x: Tensor) -> Tensor:
     return x
 )JIT"};
-static const auto parentForward = R"JIT(
+static const std::string parentForward = R"JIT(
 def forward(self, x: Tensor) -> Tensor:
     return self.subMod.forward(x)
 )JIT";
 
-static const auto moduleInterfaceSrc = R"JIT(
+static constexpr c10::string_view moduleInterfaceSrc = R"JIT(
 class OneForward(ModuleInterface):
     def one(self, x: Tensor, y: Tensor) -> Tensor:
         pass

test/cpp/jit/test_module_api.cpp

@@ -13,7 +13,7 @@
 namespace torch {
 namespace jit {
 
-static const auto moduleInterfaceSrc = R"JIT(
+static constexpr c10::string_view moduleInterfaceSrc = R"JIT(
 class OneInterface(ModuleInterface):
     def one(self, x: Tensor, y: Tensor) -> Tensor:
         pass
@@ -27,7 +27,7 @@ def forward(self, x: Tensor) -> Tensor:
     return self.attr + x
 )JIT"};
 
-static const auto parentForward = R"JIT(
+static const std::string parentForward = R"JIT(
 def forward(self, x: Tensor) -> Tensor:
     return self.subMod1.one(x, x) + self.subMod2.one(x, x)
 )JIT";

torch/csrc/jit/frontend/error_report.h

@@ -39,7 +39,7 @@ struct TORCH_API ErrorReport : public std::exception {
   friend const ErrorReport& operator<<(const ErrorReport& e, const T& t);
 
   mutable std::stringstream ss;
-  SourceRange context;
+  OwnedSourceRange context;
   mutable std::string the_message;
   std::vector<Call> error_stack;
 };

torch/csrc/jit/frontend/function_schema_parser.cpp

@@ -27,7 +27,7 @@ namespace jit {
 namespace {
 struct SchemaParser {
   SchemaParser(const std::string& str)
-      : L(std::make_shared<Source>(str)),
+      : L(std::make_shared<SourceView>(c10::string_view(str))),
         type_parser(L, /*parse_complete_tensor_types*/ false) {}
 
   either<OperatorName, FunctionSchema> parseDeclaration() {

torch/csrc/jit/frontend/lexer.h

@@ -185,7 +185,7 @@ struct TORCH_API SharedParserData {
   // find the longest match of str.substring(pos) against a token, return true
   // if successful filling in kind, start,and len
   bool match(
-      const std::string& str,
+      c10::string_view str,
       size_t pos,
       bool continuation, // are we inside a scope where newlines don't count
                          // (e.g. inside parens)
@@ -300,15 +300,15 @@ struct TORCH_API SharedParserData {
   // 1. skip whitespace
   // 2. handle comment or newline
   //
-  bool isNumber(const std::string& str, size_t start, size_t* len) {
+  bool isNumber(c10::string_view str, size_t start, size_t* len) {
     char first = str[start];
     // strtod allows numbers to start with + or - or nan or inf
     // http://en.cppreference.com/w/cpp/string/byte/strtof
     // but we want only the number part, otherwise 1+3 will turn into two
     // adjacent numbers in the lexer
     if (first == '-' || first == '+' || isalpha(first))
       return false;
-    const char* startptr = str.c_str() + start;
+    const char* startptr = str.data() + start;
     // NOLINTNEXTLINE(cppcoreguidelines-init-variables)
     char* endptr;
     torch::jit::strtod_c(startptr, &endptr);
@@ -321,7 +321,7 @@ struct TORCH_API SharedParserData {
     return *len > 0;
   }
 
-  bool isCharCount(char c, const std::string& str, size_t start, int len) {
+  bool isCharCount(char c, c10::string_view str, size_t start, int len) {
     // count checks from [start, start + len)
     return start + len <= str.size() &&
         std::count(str.begin() + start, str.begin() + start + len, c) == len;
@@ -331,7 +331,7 @@ struct TORCH_API SharedParserData {
   // strings can be enclosed with 1 or 3 single or double quotes
   // if enclosed with 3 quotes newlines are valid
   // as elsewhere, backslash and new line should be ignored
-  bool isString(const std::string& str, size_t start, size_t* len) {
+  bool isString(c10::string_view str, size_t start, size_t* len) {
     char quote = str[start];
     if (quote != '\"' && quote != '\'')
       return false;
@@ -362,9 +362,8 @@ struct TORCH_API SharedParserData {
   bool isblank(int n) {
     return isspace(n) && n != '\n';
   }
-
   // Make an exception ignoring comments for type annotation comments
-  bool isTypeComment(const std::string& str, size_t pos) {
+  bool isTypeComment(c10::string_view str, size_t pos) {
     const std::string type_string = "# type:";
     if (str.size() < pos + type_string.length()) {
       return false;
@@ -391,7 +390,7 @@ struct Token {
 };
 
 struct Lexer {
-  explicit Lexer(std::shared_ptr<Source> source)
+  explicit Lexer(std::shared_ptr<SourceView> source)
       : source(std::move(source)),
         pos(0),
         nesting(0),
@@ -532,7 +531,7 @@ struct Lexer {
     return t;
   }
 
-  std::shared_ptr<Source> source;
+  std::shared_ptr<SourceView> source;
   size_t pos;
   size_t nesting; // depth of ( [ { nesting...
   std::vector<int> indent_stack; // stack of indentation level of blocks

torch/csrc/jit/frontend/parser.cpp

@@ -46,7 +46,7 @@ Decl mergeTypesFromTypeComment(
 }
 
 struct ParserImpl {
-  explicit ParserImpl(const std::shared_ptr<Source>& source)
+  explicit ParserImpl(const std::shared_ptr<SourceView>& source)
       : L(source), shared(sharedParserData()) {}
 
   Ident parseIdent() {
@@ -801,7 +801,7 @@ struct ParserImpl {
   SharedParserData& shared;
 };
 
-Parser::Parser(const std::shared_ptr<Source>& src)
+Parser::Parser(const std::shared_ptr<SourceView>& src)
     : pImpl(new ParserImpl(src)) {}
 
 Parser::~Parser() = default;

torch/csrc/jit/frontend/parser.h

@@ -17,7 +17,7 @@ TORCH_API Decl mergeTypesFromTypeComment(
     bool is_method);
 
 struct TORCH_API Parser {
-  explicit Parser(const std::shared_ptr<Source>& src);
+  explicit Parser(const std::shared_ptr<SourceView>& src);
   TreeRef parseFunction(bool is_method);
   TreeRef parseClass();
   Decl parseTypeComment();

torch/csrc/jit/frontend/source_range.cpp

@@ -10,7 +10,7 @@ size_t SourceRangeHasher::operator()(const torch::jit::SourceRange& key) const {
       std::hash<size_t>()(key.start()) ^ std::hash<size_t>()(key.end()));
 }
 
-c10::optional<SourceRange> Source::findSourceRangeThatGenerated(
+c10::optional<SourceRange> SourceView::findSourceRangeThatGenerated(
     const SourceRange& range) {
   if (!gen_ranges_) {
     return c10::nullopt;
@@ -69,11 +69,11 @@ C10_EXPORT void SourceRange::print_with_context(
     bool highlight,
     const std::string& funcname) const {
   // This is an empty SourceRange, used as a sentinel value.
-  if (!source_) {
+  if (!source_view_) {
     return;
   }
 
-  const std::string& str = source_->text();
+  c10::string_view str = source_view_->text();
   if (size() == str.size()) {
     // this is just the entire file, not a subset, so print it out.
     // primarily used to print out python stack traces