cache: replace the all-in-one query parser by a complete one with AST/lexer/parser

1 files changed, 71 insertions, 0 deletions

diff --git a/query/lexer.go b/query/lexer.go
new file mode 100644
index 00000000..ec05f44e
--- /dev/null
+++ b/query/lexer.go
@@ -0,0 +1,71 @@
+package query
+
+import (
+	"fmt"
+	"strings"
+	"unicode"
+)
+
+type token struct {
+	qualifier string
+	value     string
+}
+
+// TODO: this lexer implementation behave badly with unmatched quotes.
+// A hand written one would be better instead of relying on strings.FieldsFunc()
+
+// tokenize parse and break a input into tokens ready to be
+// interpreted later by a parser to get the semantic.
+func tokenize(query string) ([]token, error) {
+	fields := splitQuery(query)
+
+	var tokens []token
+	for _, field := range fields {
+		split := strings.Split(field, ":")
+		if len(split) != 2 {
+			return nil, fmt.Errorf("can't tokenize \"%s\"", field)
+		}
+
+		if len(split[0]) == 0 {
+			return nil, fmt.Errorf("can't tokenize \"%s\": empty qualifier", field)
+		}
+		if len(split[1]) == 0 {
+			return nil, fmt.Errorf("empty value for qualifier \"%s\"", split[0])
+		}
+
+		tokens = append(tokens, token{
+			qualifier: split[0],
+			value:     removeQuote(split[1]),
+		})
+	}
+	return tokens, nil
+}
+
+func splitQuery(query string) []string {
+	lastQuote := rune(0)
+	f := func(c rune) bool {
+		switch {
+		case c == lastQuote:
+			lastQuote = rune(0)
+			return false
+		case lastQuote != rune(0):
+			return false
+		case unicode.In(c, unicode.Quotation_Mark):
+			lastQuote = c
+			return false
+		default:
+			return unicode.IsSpace(c)
+		}
+	}
+
+	return strings.FieldsFunc(query, f)
+}
+
+func removeQuote(field string) string {
+	if len(field) >= 2 {
+		if field[0] == '"' && field[len(field)-1] == '"' {
+			return field[1 : len(field)-1]
+		}
+	}
+	return field
+}