fix: 将 libsql-client-go 作为普通目录提交，而非 gitlink

go_modules/libsql-client-go/sqliteparserutils/buffered_tokenizer.go

@@ -0,0 +1,51 @@
+package sqliteparserutils
+
+import (
+	"fmt"
+
+	"github.com/antlr4-go/antlr/v4"
+)
+
+// bufferedTokenizer preload few next tokens and put them in the tokens buffer to support look-ahead access with limited distance
+// Tokenizer maintains loaded tokens in ring buffer and use constant memory to support all requests (which is important for streaming mode)
+type bufferedTokenizer struct {
+	source antlr.TokenSource
+	tokens []antlr.Token
+	index  int
+}
+
+// createBufferedTokenizer initialize tokenizer which will store current token and bufferSize-1 next tokens for look-ahead accesses
+func createBufferedTokenizer(source antlr.TokenSource, bufferSize int) *bufferedTokenizer {
+	tokens := make([]antlr.Token, bufferSize)
+	stream := bufferedTokenizer{source: source, tokens: tokens, index: 0}
+	stream.load(bufferSize)
+	return &stream
+}
+
+func (stream *bufferedTokenizer) load(n int) {
+	i := 0
+	for i < n {
+		token := stream.source.NextToken()
+		if token.GetChannel() != antlr.TokenDefaultChannel {
+			continue
+		}
+		stream.tokens[(stream.index+i)%len(stream.tokens)] = token
+		i += 1
+	}
+	stream.index = (stream.index + i) % len(stream.tokens)
+}
+
+func (stream *bufferedTokenizer) Consume() {
+	stream.load(1)
+}
+
+func (stream *bufferedTokenizer) Get(k int) antlr.Token {
+	if k >= len(stream.tokens) {
+		panic(fmt.Errorf("out of buffer read attempts: %d >= %d", k, len(stream.tokens)))
+	}
+	return stream.tokens[(stream.index+k)%len(stream.tokens)]
+}
+
+func (stream *bufferedTokenizer) IsEOF() bool {
+	return stream.Get(0).GetTokenType() == antlr.TokenEOF
+}