materialize-elasticsearch: extend index name normalization

Previously we would only normalize characters in index names that would be allowed by Flow
collection names, but that's not fully sufficient since the configured index name in the resource
configuration can be anything. It will usually be the collection name, but not always.

This extends normalization to all characters that pose problems in index names.

materialize-elasticsearch/driver.go

@@ -305,19 +305,24 @@ func normalizeIndexName(index string, truncateLimit int) string {
 	afterPrefix := false
 	var b strings.Builder
 	for _, r := range index {
-		// Replace disallowed characters with underscore. Of the various problematic characters in
-		// an Elasticsearch index name, a "." is the only one that Flow allows in a collection name.
-		if r == '.' {
-			r = '_'
+		// Strip disallowed prefixes.
+		if !afterPrefix && slices.Contains([]rune{'_', '-', '+', '.'}, r) {
+			continue
 		}
 
-		// Strip disallowed prefixes that may be present in a Flow collection name. A '.' is also a
-		// bad prefix, but those were already replaced with '_' above.
-		if !afterPrefix && (r == '-' || r == '_') {
-			continue
+		// Replace disallowed characters with underscore as long as we aren't still at the beginning
+		// of the string. Most of the characters in this list are named in the docs, but ES will
+		// also truncate index names that contain a '#' character to drop everything including &
+		// after that '#' character, so we'll normalize those too.
+		if slices.Contains([]rune{'*', '<', '"', ' ', '\\', '/', ',', '|', '>', '?', ':', '#'}, r) {
+			if !afterPrefix {
+				continue
+			}
+			r = '_'
 		}
 		afterPrefix = true
 
+		// Index names must be lowercase.
 		char := strings.ToLower(string(r))
 		if b.Len()+len(char) > truncateLimit {
 			// Truncate extremely long names. These must be less than 255 bytes.

materialize-elasticsearch/driver_test.go

@@ -118,7 +118,13 @@ func TestNormalizeIndexName(t *testing.T) {
 			name:       "dots",
 			byteLength: maxByteLength,
 			input:      ".some..collection",
-			want:       "some__collection",
+			want:       "some..collection",
+		},
+		{
+			name:       "normalized characters",
+			byteLength: maxByteLength,
+			input:      `??weird*<"\/,|>?:#index`,
+			want:       "weird___________index",
 		},
 		{
 			name:       "capitalized",