[clang] [HLSL][RootSignature] Implement Lexing of DescriptorTables (PR #122981)

[Finn Plummer via cfe-commits]

================
@@ -0,0 +1,152 @@
+#include "clang/Parse/ParseHLSLRootSignature.h"
+
+namespace llvm {
+namespace hlsl {
+namespace root_signature {
+
+// Lexer Definitions
+
+static bool IsPreprocessorNumberChar(char C) {
+  // TODO: extend for float support with or without hexadecimal/exponent
+  return isdigit(C); // integer support
+}
+
+bool RootSignatureLexer::LexNumber(RootSignatureToken &Result) {
+  // NumericLiteralParser does not handle the sign so we will manually apply it
+  Result.Signed = Buffer.front() == '-';
+  if (Result.Signed)
+    AdvanceBuffer();
+
+  // Retrieve the possible number
+  StringRef NumSpelling = Buffer.take_while(IsPreprocessorNumberChar);
+
+  // Parse the numeric value and so semantic checks on its specification
+  clang::NumericLiteralParser Literal(NumSpelling, SourceLoc,
+                                      PP.getSourceManager(), PP.getLangOpts(),
+                                      PP.getTargetInfo(), PP.getDiagnostics());
+  if (Literal.hadError)
+    return true; // Error has already been reported so just return
+
+  // Retrieve the number value to store into the token
+  if (Literal.isIntegerLiteral()) {
+    Result.Kind = TokenKind::int_literal;
+
+    APSInt X = APSInt(32, Result.Signed);
+    if (Literal.GetIntegerValue(X))
+      return true; // TODO: Report overflow error
+
+    X = Result.Signed ? -X : X;
+    Result.IntLiteral = (uint32_t)X.getZExtValue();
+  } else {
+    return true; // TODO: report unsupported number literal specification
+  }
+
+  AdvanceBuffer(NumSpelling.size());
+  return false;
+}
+
+bool RootSignatureLexer::Lex(SmallVector<RootSignatureToken> &Tokens) {
+  // Discard any leading whitespace
+  AdvanceBuffer(Buffer.take_while(isspace).size());
+
+  while (!Buffer.empty()) {
+    RootSignatureToken Result;
+    if (LexToken(Result))
+      return true;
+
+    // Successfully Lexed the token so we can store it
+    Tokens.push_back(Result);
+
+    // Discard any trailing whitespace
+    AdvanceBuffer(Buffer.take_while(isspace).size());
+  }
+
+  return false;
+}
+
+bool RootSignatureLexer::LexToken(RootSignatureToken &Result) {
+  // Record where this token is in the text for usage in parser diagnostics
+  Result.TokLoc = SourceLoc;
+
+  char C = Buffer.front();
+
+  // Punctuators
+  switch (C) {
+#define PUNCTUATOR(X, Y)                                                       \
+  case Y: {                                                                    \
+    Result.Kind = TokenKind::pu_##X;                                           \
+    AdvanceBuffer();                                                           \
+    return false;                                                              \
+  }
+#include "clang/Parse/HLSLRootSignatureTokenKinds.def"
----------------
inbelic wrote:

Yes, it is defined this way so that the `PUNCUATOR` that is in the `.def` is the "default" macro. This is used when we do `#define TOK(X) X,` so then `PUNCUATOR(X) -> TOK(pu_X) -> X,`. It follows the pattern in some other `TokenKinds.def`.

https://github.com/llvm/llvm-project/pull/122981