package yqlib import ( "fmt" "io" "strings" "unicode/utf8" "golang.org/x/text/unicode/norm" ) type shellVariablesEncoder struct { } func NewShellVariablesEncoder() Encoder { return &shellVariablesEncoder{} } func (pe *shellVariablesEncoder) CanHandleAliases() bool { return false } func (pe *shellVariablesEncoder) PrintDocumentSeparator(_ io.Writer) error { return nil } func (pe *shellVariablesEncoder) PrintLeadingContent(_ io.Writer, _ string) error { return nil } func (pe *shellVariablesEncoder) Encode(writer io.Writer, node *CandidateNode) error { node.ConvertKeysToStrings() err := pe.doEncode(&writer, node, "") if err != nil { return err } return err } func (pe *shellVariablesEncoder) doEncode(w *io.Writer, node *CandidateNode, path string) error { // Note this drops all comments. switch node.Kind { case ScalarNode: nonemptyPath := path if path == "" { // We can't assign an empty variable "=somevalue" because that would error out if sourced in a shell, // nor can we use "_" as a variable name ($_ is a special shell variable that can't be assigned)... // let's just pick a fallback key to use if we are encoding a single scalar nonemptyPath = "value" } _, err := io.WriteString(*w, nonemptyPath+"="+quoteValue(node.Value)+"\n") return err case SequenceNode: for index, child := range node.Content { err := pe.doEncode(w, child, appendPath(path, index)) if err != nil { return err } } return nil case MappingNode: for index := 0; index < len(node.Content); index = index + 2 { key := node.Content[index] value := node.Content[index+1] err := pe.doEncode(w, value, appendPath(path, key.Value)) if err != nil { return err } } return nil case AliasNode: return pe.doEncode(w, node.Alias, path) default: return fmt.Errorf("Unsupported node %v", node.Tag) } } func appendPath(cookedPath string, rawKey interface{}) string { // Shell variable names must match // [a-zA-Z_]+[a-zA-Z0-9_]* // // While this is not mandated by POSIX, which is quite lenient, it is // what shells (for example busybox ash *) allow in practice. // // Since yaml names can contain basically any character, we will process them according to these steps: // // 1. apply unicode compatibility decomposition NFKD (this will convert accented // letters to letters followed by accents, split ligatures, replace exponents // with the corresponding digit, etc. // // 2. discard non-ASCII characters as well as ASCII control characters (ie. anything // with code point < 32 or > 126), this will eg. discard accents but keep the base // unaccented letter because of NFKD above // // 3. replace all non-alphanumeric characters with _ // // Moreover, for the root key only, we will prepend an underscore if what results from the steps above // does not start with [a-zA-Z_] (ie. if the root key starts with a digit). // // Note this is NOT a 1:1 mapping. // // (*) see endofname.c from https://git.busybox.net/busybox/tag/?h=1_36_0 // XXX empty strings key := strings.Map(func(r rune) rune { if isAlphaNumericOrUnderscore(r) { return r } else if r < 32 || 126 < r { return -1 } return '_' }, norm.NFKD.String(fmt.Sprintf("%v", rawKey))) if cookedPath == "" { firstRune, _ := utf8.DecodeRuneInString(key) if !isAlphaOrUnderscore(firstRune) { return "_" + key } return key } return cookedPath + "_" + key } func quoteValue(value string) string { needsQuoting := false for _, r := range value { if !isAlphaNumericOrUnderscore(r) { needsQuoting = true break } } if needsQuoting { return "'" + strings.ReplaceAll(value, "'", "'\"'\"'") + "'" } return value } func isAlphaOrUnderscore(r rune) bool { return ('a' <= r && r <= 'z') || ('A' <= r && r <= 'Z') || r == '_' } func isAlphaNumericOrUnderscore(r rune) bool { return isAlphaOrUnderscore(r) || ('0' <= r && r <= '9') }