Add character encoding examples and documentation, e.g. parsing UTF-8

cwbaker · Jun 2, 2023 · 6821f78 · 6821f78
2 parents 1f85c9d + e346baf
commit 6821f78
Show file tree

Hide file tree

Showing 6 changed files with 94 additions and 59 deletions.
diff --git a/README.md b/README.md
@@ -183,9 +183,9 @@ Parser<const char*, XmlUserData> parser( xml_parser_state_machine );
 
 Reference a parse table as an `extern` variable for offline generated parse tables.  See [lalr_calculator_example.cpp](#lalr_calculator_example.cpp) for an example of compiling a grammar to parse tables at runtime.
 
-Create a `Parser` object with the parse table as the sole argument to the constructor.  The `Parser` class template requires an iterator type template argument and optionally allows for user data, character type, character traits, and allocator to be overridden.
+Create a `Parser` object with the parse table as the sole argument to the constructor.  The `Parser` class template requires an iterator type template argument and optionally allows for user data; and character type, traits, and allocator to be overridden.  In the above example the iterator type is `const char*`, user data is the custom `XmlUserData` type, and the character parameters default to those implied by the iterator.
 
-In the above example the iterator type is `const char*` and the user data is the custom `XmlUserData` type.
+Change the `Iterator` template parameter to read input from different sources and convert character encodings, e.g. from UTF-8 in a file to UTF-32 in memory.  See [lalr_json_example.cpp](lalr/lalr_examples/lalr_json_example.cpp) for an example of reading a UTF-8 encoded file to UTF-32, `char32_t` in memory.  To parse UTF-8 input to UTF-8 encoding in memory it is usually sufficient to use a iterator templated to `unsigned char` or `uint8_t`, see [lalr_xml_example.cpp](lalr/lalr_examples/lalr_xml_example.cpp) for an example of doing so in practice.
 
 **3. Bind lexer action handlers**
 

diff --git a/src/lalr/Parser.ipp b/src/lalr/Parser.ipp
@@ -494,7 +494,7 @@ void Parser<Iterator, UserData, Char, Traits, Allocator>::debug_shift( const Par
     if ( debug_enabled_ )
     {
         const ParserSymbol* symbol = node.symbol();
-        const std::string& lexeme = node.lexeme();
+        const std::basic_string<Char, Traits, Allocator>& lexeme = node.lexeme();
         int line = node.line();
         int column = node.column();
         fire_printf( "SHIFT: (%s %s %d:%d)\n", symbol ? symbol->identifier : "", lexeme.c_str(), line, column );
@@ -521,14 +521,14 @@ void Parser<Iterator, UserData, Char, Traits, Allocator>::debug_reduce( const Pa
 
     if ( debug_enabled_ )
     {
-        fire_printf( "REDUCE: %s <- ", reduced_symbol->identifier );        
+        fire_printf( "REDUCE: %s <- ", reduced_symbol->identifier );
 
         const ParserNode* node = nodes_.data() + start;
         const ParserNode* node_end = nodes_.data() + finish;
         if ( node != node_end )
         {
             const ParserSymbol* symbol = node->symbol();
-            const std::string& lexeme = node->lexeme();
+            const std::basic_string<Char, Traits, Allocator>& lexeme = node->lexeme();
             int line = node->line();
             int column = node->column();
             fire_printf( "(%s %s %d:%d)", symbol ? symbol->identifier : "", lexeme.c_str(), line, column );
@@ -538,7 +538,7 @@ void Parser<Iterator, UserData, Char, Traits, Allocator>::debug_reduce( const Pa
         while ( node != node_end )
         {
             const ParserSymbol* symbol = node->symbol();
-            const std::string& lexeme = node->lexeme();
+            const std::basic_string<Char, Traits, Allocator>& lexeme = node->lexeme();
             int line = node->line();
             int column = node->column();
             fire_printf( " (%s %s %d:%d)", symbol ? symbol->identifier : "", lexeme.c_str(), line, column );

diff --git a/src/lalr/lalr_examples/lalr_examples.forge b/src/lalr/lalr_examples/lalr_examples.forge
@@ -15,6 +15,9 @@ for _, cc in toolsets('^cc.*') do
         libraries = libraries;
         '${lib}/lalr_${platform}_${architecture}';
         cc:Cxx '${obj}/%1' {
+            defines = {
+                ([[LALR_EXAMPLES=\"%s/\"]]):format( pwd() );
+            };
             "lalr_examples.cpp",
             "lalr_error_handling_calculator_example.cpp",
             "lalr_hello_world_example.cpp",

diff --git a/src/lalr/lalr_examples/lalr_json_example.cpp b/src/lalr/lalr_examples/lalr_json_example.cpp
@@ -4,30 +4,41 @@
 #include <lalr/Parser.ipp>
 #include <lalr/PositionIterator.hpp>
 #include <lalr/string_literal.hpp>
+#include <fstream>
 #include <string.h>
+#include <locale>
+#include <clocale>
+#include <codecvt>
+#ifdef __APPLE__
+#include <stdint.h>
+#else
+#include <uchar.h>
+#endif
 
 using namespace std;
 using namespace lalr;
 
 namespace 
 {
 
+typedef std::basic_string<char32_t> String;
+
 struct Attribute;
 
 struct Value
 {
-    std::string value_;
+    String value_;
     std::vector<shared_ptr<Attribute>> attributes_;
     std::vector<shared_ptr<Value>> elements_;
 
     Value()
     : value_()
     , attributes_()
     , elements_()
-    {        
+    {
     }
 
-    Value( const std::string& value )
+    Value( const String& value )
     : value_( value )
     , attributes_()
     , elements_()
@@ -37,7 +48,7 @@ struct Value
 
 struct Attribute
 {
-    std::string name_;
+    String name_;
     shared_ptr<Value> value_;
 
     Attribute()
@@ -46,7 +57,7 @@ struct Attribute
     {
     }
 
-    Attribute( const std::string& name, const shared_ptr<Value>& value )
+    Attribute( const String& name, const shared_ptr<Value>& value )
     : name_( name )
     , value_( value )
     {
@@ -55,7 +66,7 @@ struct Attribute
 
 struct JsonUserData
 {
-    std::string name_;
+    String name_;
     shared_ptr<Value> value_;
 
     JsonUserData()
@@ -70,66 +81,66 @@ struct JsonUserData
     {
     }    
 
-    JsonUserData( const std::string& name, shared_ptr<Value> value )
+    JsonUserData( const String& name, shared_ptr<Value> value )
     : name_( name )
     , value_( value )
     {
     }    
 };
 
-static JsonUserData document( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData document( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     return start[1];
 }
 
-static JsonUserData attribute( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData attribute( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {   
     const shared_ptr<Value>& attribute = start[2].value_;
     return JsonUserData( nodes[0].lexeme(), attribute );
 }
 
-static JsonUserData null( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData null( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     shared_ptr<Value> null_value = make_shared<Value>();
     return JsonUserData( null_value );
 }
 
-static JsonUserData value( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData value( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     shared_ptr<Value> value = make_shared<Value>( nodes[0].lexeme() );
     return JsonUserData( value );
 }
 
-static JsonUserData object( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData object( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     return start[1];
 }
 
-static JsonUserData add_to_object( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData add_to_object( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     const shared_ptr<Value>& object = start[0].value_;
     shared_ptr<Attribute> attribute = make_shared<Attribute>( start[2].name_, start[2].value_ );
     object->attributes_.push_back( attribute );
     return JsonUserData( object );
 }
 
-static JsonUserData create_object( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData create_object( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     shared_ptr<Value> object = make_shared<Value>();
     shared_ptr<Attribute> attribute = make_shared<Attribute>( start[0].name_, start[0].value_ );
     object->attributes_.push_back( attribute );
     return JsonUserData( object );
 }
 
-static JsonUserData add_to_array( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData add_to_array( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     const shared_ptr<Value>& array = start[0].value_;
     const shared_ptr<Value>& element = start[2].value_;
     array->elements_.push_back( element );
     return JsonUserData( array );
 }
 
-static JsonUserData create_array( const JsonUserData* start, const ParserNode<char>* nodes, size_t length )
+static JsonUserData create_array( const JsonUserData* start, const ParserNode<char32_t>* nodes, size_t length )
 {
     shared_ptr<Value> array = make_shared<Value>();
     const shared_ptr<Value>& element = start[0].value_;
@@ -147,20 +158,28 @@ static void indent( int level )
 
 static void print( const Value& value, int level )
 {
+    std::setlocale( LC_ALL, "en_US.UTF-8" );
+    std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> utf8;
+
     for ( const shared_ptr<Attribute>& attribute : value.attributes_ )
     {
         LALR_ASSERT( attribute );
-        const string& name = attribute->name_;
+        const String& name = attribute->name_;
         const Value& value = *attribute->value_;
         if ( value.attributes_.empty() && value.elements_.empty() )
         {
             indent( level + 1 );
-            printf( "%s='%s'\n", attribute->name_.c_str(), attribute->value_->value_.c_str() );
+            printf( "%s='%s'\n", 
+                utf8.to_bytes(attribute->name_).c_str(),
+                utf8.to_bytes(attribute->value_->value_).c_str()
+            );
         }
         else
         {
             indent( level + 1 );
-            printf( "%s:\n", attribute->name_.c_str() );
+            printf( "%s:\n",
+                utf8.to_bytes(attribute->name_).c_str()
+            );
             print( value, level + 1 );
         }
     }
@@ -173,7 +192,10 @@ static void print( const Value& value, int level )
         if ( value.attributes_.empty() && value.elements_.empty() )
         {
             indent( level + 1 );
-            printf( "%d: '%s'\n", index, value.value_.c_str() );
+            printf( "%d: '%s'\n",
+                index,
+                utf8.to_bytes(value.value_).c_str()
+            );
         }
         else
         {
@@ -190,8 +212,8 @@ static void print( const Value& value, int level )
 void lalr_json_example()
 {
     extern const lalr::ParserStateMachine* json_parser_state_machine;
-    Parser<const char*, JsonUserData> parser( json_parser_state_machine );
-    parser.set_lexer_action_handler( "string", &string_literal<const char*> );
+    Parser<istreambuf_iterator<char32_t>, JsonUserData> parser( json_parser_state_machine );
+    parser.set_lexer_action_handler( "string", &string_literal<istreambuf_iterator<char32_t>> );
     parser.parser_action_handlers()
         ( "document", &document )
         ( "add_to_object", &add_to_object )
@@ -205,20 +227,16 @@ void lalr_json_example()
         ( "array", &object )
     ;
 
-    const char* input =
-    "{\n"
-    "    \"model\": {\n"
-    "        \"format\": \"Model\",\n"
-    "        \"version\": 1,\n"
-    "        \"address\": \"0017FAB0\",\n"
-    "        \"items\": {\n"
-    "            \"name\": \"Albert\"\n"
-    "        },\n"
-    "        \"more_items\": ['one', 2, 3]\n"
-    "    }\n"
-    "}\n";
-
-    parser.parse( input, input + strlen(input) );
+    using std::locale;
+    using std::codecvt;
+    using std::basic_ifstream;
+    using std::istreambuf_iterator;
+    std::basic_ifstream<char32_t> file( LALR_EXAMPLES "lalr_json_example.json", std::ios_base::binary );
+    file.imbue( locale(file.getloc(), new codecvt<char32_t, char, std::mbstate_t>) );
+    istreambuf_iterator<char32_t> input( file );
+    istreambuf_iterator<char32_t> input_end;
+
+    parser.parse( input, input_end );
     LALR_ASSERT( parser.accepted() );
     LALR_ASSERT( parser.full() );
     print( *parser.user_data().value_, 0 );

diff --git a/src/lalr/lalr_examples/lalr_json_example.json b/src/lalr/lalr_examples/lalr_json_example.json
@@ -0,0 +1,11 @@
+{
+    "model": {
+        "format": "Model",
+        "version": 1,
+        "address": "0017FAB0",
+        "items": {
+            "name": "Albert"
+        },
+        "more_items": ["one", 2, 3, "to prove that it's really UTF-8... 😁 😄!"]
+    }
+}