Args and TokenStream use StringRefs instead of std::strings

Because TokenStream is copied around a lot, moving it to use
`StringRef` removes _a lot_ of allocations per `Opt` in the parser.
Args are not copied around much, but changing them as well makes it
obvious that they do not participate in the ownership.

The changes add up to removing ~180 allocations for "empty"
invocation of the test binary.
(`./tests/ExtraTests/NoTests --allow-running-no-tests -o /dev/null`
is down to 317 allocs)

src/catch2/internal/catch_clara.cpp

@@ -37,6 +37,19 @@ namespace {
         return optName;
     }
 
+    static size_t find_first_separator(Catch::StringRef sr) {
+        auto is_separator = []( char c ) {
+            return c == ' ' || c == ':' || c == '=';
+        };
+        size_t pos = 0;
+        while (pos < sr.size()) {
+            if (is_separator(sr[pos])) { return pos; }
+            ++pos;
+        }
+
+        return Catch::StringRef::npos;
+    }
+
 } // namespace
 
 namespace Catch {
@@ -52,23 +65,23 @@ namespace Catch {
                 }
 
                 if ( it != itEnd ) {
-                    auto const& next = *it;
+                    StringRef next = *it;
                     if ( isOptPrefix( next[0] ) ) {
-                        auto delimiterPos = next.find_first_of( " :=" );
+                        auto delimiterPos = find_first_separator(next);
-                        if ( delimiterPos != std::string::npos ) {
+                        if ( delimiterPos != StringRef::npos ) {
                             m_tokenBuffer.push_back(
                                 { TokenType::Option,
                                   next.substr( 0, delimiterPos ) } );
                             m_tokenBuffer.push_back(
                                 { TokenType::Argument,
-                                  next.substr( delimiterPos + 1 ) } );
+                                  next.substr( delimiterPos + 1, next.size() ) } );
                         } else {
                             if ( next[1] != '-' && next.size() > 2 ) {
-                                std::string opt = "- ";
+                                // Combined short args, e.g. "-ab" for "-a -b"
                                 for ( size_t i = 1; i < next.size(); ++i ) {
-                                    opt[1] = next[i];
                                     m_tokenBuffer.push_back(
-                                        { TokenType::Option, opt } );
+                                        { TokenType::Option,
+                                          next.substr( i, 1 ) } );
                                 }
                             } else {
                                 m_tokenBuffer.push_back(
@@ -128,7 +141,7 @@ namespace Catch {
             size_t ParserBase::cardinality() const { return 1; }
 
             InternalParseResult ParserBase::parse( Args const& args ) const {
-                return parse( args.exeName(), TokenStream( args ) );
+                return parse( static_cast<std::string>(args.exeName()), TokenStream( args ) );
             }
 
             ParseState::ParseState( ParseResultType type,
@@ -166,7 +179,7 @@ namespace Catch {
             auto valueRef =
                 static_cast<Detail::BoundValueRefBase*>(m_ref.get());
 
-            auto result = valueRef->setValue(remainingTokens->token);
+            auto result = valueRef->setValue(static_cast<std::string>(token.token));
             if (!result)
                 return Detail::InternalParseResult(result);
             else
@@ -192,7 +205,7 @@ namespace Catch {
             return { oss.str(), m_description };
         }
 
-        bool Opt::isMatch(std::string const& optToken) const {
+        bool Opt::isMatch(StringRef optToken) const {
             auto normalisedToken = normaliseOpt(optToken);
             for (auto const& name : m_optNames) {
                 if (normaliseOpt(name) == normalisedToken)
@@ -237,7 +250,7 @@ namespace Catch {
                             return Detail::InternalParseResult::runtimeError(
                                 "Expected argument following " +
                                 token.token);
-                        const auto result = valueRef->setValue(argToken.token);
+                        const auto result = valueRef->setValue(static_cast<std::string>(argToken.token));
                         if (!result)
                             return Detail::InternalParseResult(result);
                         if (result.value() ==
@@ -433,7 +446,7 @@ namespace Catch {
         Args::Args(int argc, char const* const* argv) :
             m_exeName(argv[0]), m_args(argv + 1, argv + argc) {}
 
-        Args::Args(std::initializer_list<std::string> args) :
+        Args::Args(std::initializer_list<StringRef> args) :
             m_exeName(*args.begin()),
             m_args(args.begin() + 1, args.end()) {}
 

src/catch2/internal/catch_clara.hpp

@@ -102,17 +102,16 @@ namespace Catch {
             enum class TokenType { Option, Argument };
             struct Token {
                 TokenType type;
-                std::string token;
+                StringRef token;
             };
 
             // Abstracts iterators into args as a stream of tokens, with option
             // arguments uniformly handled
             class TokenStream {
-                using Iterator = std::vector<std::string>::const_iterator;
+                using Iterator = std::vector<StringRef>::const_iterator;
                 Iterator it;
                 Iterator itEnd;
                 std::vector<Token> m_tokenBuffer;
-
                 void loadBuffer();
 
             public:
@@ -582,7 +581,7 @@ namespace Catch {
 
             Detail::HelpColumns getHelpColumns() const;
 
-            bool isMatch(std::string const& optToken) const;
+            bool isMatch(StringRef optToken) const;
 
             using ParserBase::parse;
 
@@ -676,18 +675,20 @@ namespace Catch {
                       Detail::TokenStream const& tokens) const override;
         };
 
-        // Transport for raw args (copied from main args, or supplied via
+        /**
-        // init list for testing)
+         * Wrapper over argc + argv, assumes that the inputs outlive it
+         */
         class Args {
             friend Detail::TokenStream;
-            std::string m_exeName;
+            StringRef m_exeName;
-            std::vector<std::string> m_args;
+            std::vector<StringRef> m_args;
 
         public:
             Args(int argc, char const* const* argv);
-            Args(std::initializer_list<std::string> args);
+            // Helper constructor for testing
+            Args(std::initializer_list<StringRef> args);
 
-            std::string const& exeName() const { return m_exeName; }
+            StringRef exeName() const { return m_exeName; }
         };
 
 

src/catch2/internal/catch_stringref.hpp

@@ -25,6 +25,8 @@ namespace Catch {
         using size_type = std::size_t;
         using const_iterator = const char*;
 
+        static constexpr size_type npos{ static_cast<size_type>( -1 ) };
+
     private:
         static constexpr char const* const s_empty = "";