aboutsummaryrefslogtreecommitdiff
path: root/src/syntax
diff options
context:
space:
mode:
Diffstat (limited to 'src/syntax')
-rw-r--r--src/syntax/token.cc122
1 files changed, 73 insertions, 49 deletions
diff --git a/src/syntax/token.cc b/src/syntax/token.cc
index b1db7ec1..fe3cd7a7 100644
--- a/src/syntax/token.cc
+++ b/src/syntax/token.cc
@@ -31,58 +31,82 @@
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include <open-axiom/token>
+#include <ostream>
+#include <iostream>
namespace OpenAxiom {
- namespace token {
- struct Keyword {
- const char* const key;
- const Value value;
- };
+ std::ostream&
+ operator<<(std::ostream& os, TokenCategory tc) {
+ switch (tc) {
+ case TokenCategory::Unclassified: os << "UNCLASSIFIED"; break;
+ case TokenCategory::Whitespace: os << "WHITESPACE"; break;
+ case TokenCategory::Comment: os << "COMMENT"; break;
+ case TokenCategory::Punctuator: os << "PUNCTUATOR"; break;
+ case TokenCategory::Operator: os << "OPERATOR"; break;
+ case TokenCategory::Integer: os << "INTEGER"; break;
+ case TokenCategory::FloatingPoint: os << "FLOATINGPOINT"; break;
+ case TokenCategory::String: os << "STRING"; break;
+ case TokenCategory::Keyword: os << "KEYWORD"; break;
+ case TokenCategory::Identifier: os << "IDENTIFIER"; break;
+ case TokenCategory::Formatting: os << "FORMATTING"; break;
+ case TokenCategory::Junk: os << "JUNK"; break;
+ default: os << "????"; break;
+ }
+ return os;
+ }
+
- const Keyword keyword_map[] = {
- { "add", Value::Add },
- { "and", Value::And },
- { "assume", Value::Assume },
- { "break", Value::Break },
- { "by", Value::By },
- { "case", Value::Case },
- { "catch", Value::Catch },
- { "do", Value::Do },
- { "else", Value::Else },
- { "exist", Value::Exists },
- { "finally", Value::Finally },
- { "for", Value::For },
- { "from", Value::From },
- { "function", Value::Function },
- { "has", Value::Has },
- { "if", Value::If },
- { "import", Value::Import },
- { "in", Value::In },
- { "inline", Value::Inline },
- { "is", Value::Is },
- { "isnt", Value::Isnt },
- { "iterate", Value::Iterate },
- { "leave", Value::Leave },
- { "macro", Value::Macro },
- { "mod", Value::Mod },
- { "module", Value::Module },
- { "namespace", Value::Namespace },
- { "of", Value::Of },
- { "or", Value::Or },
- { "pretend", Value::Pretend },
- { "quo", Value::Quo },
- { "rem", Value::Rem },
- { "repeat", Value::Repeat },
- { "return", Value::Return },
- { "rule", Value::Rule },
- { "structure", Value::Structure },
- { "then", Value::Then },
- { "throw", Value::Throw },
- { "try", Value::Try },
- { "until", Value::Until },
- { "with", Value::With },
- { "where", Value::Where },
- { "while", Value::While }
+ bool separator_or_punctuator(uint8_t c) {
+ switch (c) {
+ case '.': case '`': case '^': case '&': case '~': case '*':
+ case '-': case '+': case ';': case ',': case '@': case '|':
+ case '\'': case ':': case '=': case '\\': case '"': case '/':
+ case '(': case ')': case '{': case '}': case '[': case ']':
+ case '<': case '>': case '#': case ' ':
+ return true;
+ default:
+ return false;
+ }
+ }
+
+ namespace {
+ struct TokenMapEntry {
+ const char* const text;
+ const TokenCategory category;
+ const TokenValue value;
+ const Language dialect; // = Language::Spad
};
}
+
+ const TokenMapEntry token_map[] {
+#undef OPENAXIOM_DEFINE_TOKEN
+#define OPENAXIOM_DEFINE_TOKEN(T, N, C, ...) \
+ { N, TokenCategory::C, TokenValue::T, __VA_ARGS__ },
+#include <open-axiom/token-value>
+#undef OPENAXIOM_DEFINE_TOKEN
+ };
+
+ TokenClassification
+ classify(const std::string& s) {
+ for (auto& t : token_map) {
+ if (t.text == s)
+ return { t.category, t.value };
+ }
+ return { TokenCategory::Unclassified, TokenValue::Unknown };
+ }
+
+ std::ostream&
+ operator<<(std::ostream& os, TokenValue tv) {
+ if (tv < TokenValue::Artificial)
+ os << token_map[uint8_t(tv)].text;
+ else switch (tv) {
+ case TokenValue::Indent: os << "%INDENT"; break;
+ case TokenValue::Unindent: os << "%UNIDENT"; break;
+ case TokenValue::Justify: os << "%JUSTIFY"; break;
+ default: os << "%ALIEN"; break;
+ }
+
+ return os;
+ }
+
}