diff options
Diffstat (limited to 'vendor/github.com/openshift/imagebuilder/dockerfile')
5 files changed, 944 insertions, 0 deletions
diff --git a/vendor/github.com/openshift/imagebuilder/dockerfile/NOTICE b/vendor/github.com/openshift/imagebuilder/dockerfile/NOTICE new file mode 100644 index 000000000..519a7e995 --- /dev/null +++ b/vendor/github.com/openshift/imagebuilder/dockerfile/NOTICE @@ -0,0 +1,26 @@ +Source files in this directory and all sub-directories have been +copied from github.com/docker/docker/builder/dockerfile and are +Licensed under the Apache License Version 2.0. + +Note that the fork of github.com/docker/docker used commit +b68221c37ee597950364788204546f9c9d0e46a1. + +Docker +Copyright 2012-2017 Docker, Inc. + +This product includes software developed at Docker, Inc. (https://www.docker.com). + +This product contains software (https://github.com/kr/pty) developed +by Keith Rarick, licensed under the MIT License. + +The following is courtesy of our legal counsel: + + +Use and transfer of Docker may be subject to certain restrictions by the +United States and other governments. +It is your responsibility to ensure that your use and/or transfer does not +violate applicable laws. + +For more information, please see https://www.bis.doc.gov + +See also https://www.apache.org/dev/crypto.html and/or seek legal counsel. diff --git a/vendor/github.com/openshift/imagebuilder/dockerfile/command/command.go b/vendor/github.com/openshift/imagebuilder/dockerfile/command/command.go new file mode 100644 index 000000000..f23c6874b --- /dev/null +++ b/vendor/github.com/openshift/imagebuilder/dockerfile/command/command.go @@ -0,0 +1,46 @@ +// Package command contains the set of Dockerfile commands. +package command + +// Define constants for the command strings +const ( + Add = "add" + Arg = "arg" + Cmd = "cmd" + Copy = "copy" + Entrypoint = "entrypoint" + Env = "env" + Expose = "expose" + From = "from" + Healthcheck = "healthcheck" + Label = "label" + Maintainer = "maintainer" + Onbuild = "onbuild" + Run = "run" + Shell = "shell" + StopSignal = "stopsignal" + User = "user" + Volume = "volume" + Workdir = "workdir" +) + +// Commands is list of all Dockerfile commands +var Commands = map[string]struct{}{ + Add: {}, + Arg: {}, + Cmd: {}, + Copy: {}, + Entrypoint: {}, + Env: {}, + Expose: {}, + From: {}, + Healthcheck: {}, + Label: {}, + Maintainer: {}, + Onbuild: {}, + Run: {}, + Shell: {}, + StopSignal: {}, + User: {}, + Volume: {}, + Workdir: {}, +} diff --git a/vendor/github.com/openshift/imagebuilder/dockerfile/parser/line_parsers.go b/vendor/github.com/openshift/imagebuilder/dockerfile/parser/line_parsers.go new file mode 100644 index 000000000..82d912b26 --- /dev/null +++ b/vendor/github.com/openshift/imagebuilder/dockerfile/parser/line_parsers.go @@ -0,0 +1,399 @@ +package parser + +// line parsers are dispatch calls that parse a single unit of text into a +// Node object which contains the whole statement. Dockerfiles have varied +// (but not usually unique, see ONBUILD for a unique example) parsing rules +// per-command, and these unify the processing in a way that makes it +// manageable. + +import ( + "encoding/json" + "errors" + "fmt" + "sort" + "strings" + "unicode" + "unicode/utf8" + + "github.com/openshift/imagebuilder/dockerfile/command" +) + +var ( + errDockerfileNotStringArray = errors.New("when using JSON array syntax, arrays must be comprised of strings only") +) + +const ( + commandLabel = "LABEL" +) + +// ignore the current argument. This will still leave a command parsed, but +// will not incorporate the arguments into the ast. +func parseIgnore(rest string, d *Directive) (*Node, map[string]bool, error) { + return &Node{}, nil, nil +} + +// used for onbuild. Could potentially be used for anything that represents a +// statement with sub-statements. +// +// ONBUILD RUN foo bar -> (onbuild (run foo bar)) +// +func parseSubCommand(rest string, d *Directive) (*Node, map[string]bool, error) { + if rest == "" { + return nil, nil, nil + } + + child, err := newNodeFromLine(rest, d) + if err != nil { + return nil, nil, err + } + + return &Node{Children: []*Node{child}}, nil, nil +} + +// helper to parse words (i.e space delimited or quoted strings) in a statement. +// The quotes are preserved as part of this function and they are stripped later +// as part of processWords(). +func parseWords(rest string, d *Directive) []string { + const ( + inSpaces = iota // looking for start of a word + inWord + inQuote + ) + + words := []string{} + phase := inSpaces + word := "" + quote := '\000' + blankOK := false + var ch rune + var chWidth int + + for pos := 0; pos <= len(rest); pos += chWidth { + if pos != len(rest) { + ch, chWidth = utf8.DecodeRuneInString(rest[pos:]) + } + + if phase == inSpaces { // Looking for start of word + if pos == len(rest) { // end of input + break + } + if unicode.IsSpace(ch) { // skip spaces + continue + } + phase = inWord // found it, fall through + } + if (phase == inWord || phase == inQuote) && (pos == len(rest)) { + if blankOK || len(word) > 0 { + words = append(words, word) + } + break + } + if phase == inWord { + if unicode.IsSpace(ch) { + phase = inSpaces + if blankOK || len(word) > 0 { + words = append(words, word) + } + word = "" + blankOK = false + continue + } + if ch == '\'' || ch == '"' { + quote = ch + blankOK = true + phase = inQuote + } + if ch == d.escapeToken { + if pos+chWidth == len(rest) { + continue // just skip an escape token at end of line + } + // If we're not quoted and we see an escape token, then always just + // add the escape token plus the char to the word, even if the char + // is a quote. + word += string(ch) + pos += chWidth + ch, chWidth = utf8.DecodeRuneInString(rest[pos:]) + } + word += string(ch) + continue + } + if phase == inQuote { + if ch == quote { + phase = inWord + } + // The escape token is special except for ' quotes - can't escape anything for ' + if ch == d.escapeToken && quote != '\'' { + if pos+chWidth == len(rest) { + phase = inWord + continue // just skip the escape token at end + } + pos += chWidth + word += string(ch) + ch, chWidth = utf8.DecodeRuneInString(rest[pos:]) + } + word += string(ch) + } + } + + return words +} + +// parse environment like statements. Note that this does *not* handle +// variable interpolation, which will be handled in the evaluator. +func parseNameVal(rest string, key string, d *Directive) (*Node, error) { + // This is kind of tricky because we need to support the old + // variant: KEY name value + // as well as the new one: KEY name=value ... + // The trigger to know which one is being used will be whether we hit + // a space or = first. space ==> old, "=" ==> new + + words := parseWords(rest, d) + if len(words) == 0 { + return nil, nil + } + + // Old format (KEY name value) + if !strings.Contains(words[0], "=") { + parts := tokenWhitespace.Split(rest, 2) + if len(parts) < 2 { + return nil, fmt.Errorf(key + " must have two arguments") + } + return newKeyValueNode(parts[0], parts[1]), nil + } + + var rootNode *Node + var prevNode *Node + for _, word := range words { + if !strings.Contains(word, "=") { + return nil, fmt.Errorf("Syntax error - can't find = in %q. Must be of the form: name=value", word) + } + + parts := strings.SplitN(word, "=", 2) + node := newKeyValueNode(parts[0], parts[1]) + rootNode, prevNode = appendKeyValueNode(node, rootNode, prevNode) + } + + return rootNode, nil +} + +func newKeyValueNode(key, value string) *Node { + return &Node{ + Value: key, + Next: &Node{Value: value}, + } +} + +func appendKeyValueNode(node, rootNode, prevNode *Node) (*Node, *Node) { + if rootNode == nil { + rootNode = node + } + if prevNode != nil { + prevNode.Next = node + } + + prevNode = node.Next + return rootNode, prevNode +} + +func parseEnv(rest string, d *Directive) (*Node, map[string]bool, error) { + node, err := parseNameVal(rest, "ENV", d) + return node, nil, err +} + +func parseLabel(rest string, d *Directive) (*Node, map[string]bool, error) { + node, err := parseNameVal(rest, commandLabel, d) + return node, nil, err +} + +// NodeFromLabels returns a Node for the injected labels +func NodeFromLabels(labels map[string]string) *Node { + keys := []string{} + for key := range labels { + keys = append(keys, key) + } + // Sort the label to have a repeatable order + sort.Strings(keys) + + labelPairs := []string{} + var rootNode *Node + var prevNode *Node + for _, key := range keys { + value := labels[key] + labelPairs = append(labelPairs, fmt.Sprintf("%q='%s'", key, value)) + // Value must be single quoted to prevent env variable expansion + // See https://github.com/docker/docker/issues/26027 + node := newKeyValueNode(key, "'"+value+"'") + rootNode, prevNode = appendKeyValueNode(node, rootNode, prevNode) + } + + return &Node{ + Value: command.Label, + Original: commandLabel + " " + strings.Join(labelPairs, " "), + Next: rootNode, + } +} + +// parses a statement containing one or more keyword definition(s) and/or +// value assignments, like `name1 name2= name3="" name4=value`. +// Note that this is a stricter format than the old format of assignment, +// allowed by parseNameVal(), in a way that this only allows assignment of the +// form `keyword=[<value>]` like `name2=`, `name3=""`, and `name4=value` above. +// In addition, a keyword definition alone is of the form `keyword` like `name1` +// above. And the assignments `name2=` and `name3=""` are equivalent and +// assign an empty value to the respective keywords. +func parseNameOrNameVal(rest string, d *Directive) (*Node, map[string]bool, error) { + words := parseWords(rest, d) + if len(words) == 0 { + return nil, nil, nil + } + + var ( + rootnode *Node + prevNode *Node + ) + for i, word := range words { + node := &Node{} + node.Value = word + if i == 0 { + rootnode = node + } else { + prevNode.Next = node + } + prevNode = node + } + + return rootnode, nil, nil +} + +// parses a whitespace-delimited set of arguments. The result is effectively a +// linked list of string arguments. +func parseStringsWhitespaceDelimited(rest string, d *Directive) (*Node, map[string]bool, error) { + if rest == "" { + return nil, nil, nil + } + + node := &Node{} + rootnode := node + prevnode := node + for _, str := range tokenWhitespace.Split(rest, -1) { // use regexp + prevnode = node + node.Value = str + node.Next = &Node{} + node = node.Next + } + + // XXX to get around regexp.Split *always* providing an empty string at the + // end due to how our loop is constructed, nil out the last node in the + // chain. + prevnode.Next = nil + + return rootnode, nil, nil +} + +// parseString just wraps the string in quotes and returns a working node. +func parseString(rest string, d *Directive) (*Node, map[string]bool, error) { + if rest == "" { + return nil, nil, nil + } + n := &Node{} + n.Value = rest + return n, nil, nil +} + +// parseJSON converts JSON arrays to an AST. +func parseJSON(rest string, d *Directive) (*Node, map[string]bool, error) { + rest = strings.TrimLeftFunc(rest, unicode.IsSpace) + if !strings.HasPrefix(rest, "[") { + return nil, nil, fmt.Errorf(`Error parsing "%s" as a JSON array`, rest) + } + + var myJSON []interface{} + if err := json.NewDecoder(strings.NewReader(rest)).Decode(&myJSON); err != nil { + return nil, nil, err + } + + var top, prev *Node + for _, str := range myJSON { + s, ok := str.(string) + if !ok { + return nil, nil, errDockerfileNotStringArray + } + + node := &Node{Value: s} + if prev == nil { + top = node + } else { + prev.Next = node + } + prev = node + } + + return top, map[string]bool{"json": true}, nil +} + +// parseMaybeJSON determines if the argument appears to be a JSON array. If +// so, passes to parseJSON; if not, quotes the result and returns a single +// node. +func parseMaybeJSON(rest string, d *Directive) (*Node, map[string]bool, error) { + if rest == "" { + return nil, nil, nil + } + + node, attrs, err := parseJSON(rest, d) + + if err == nil { + return node, attrs, nil + } + if err == errDockerfileNotStringArray { + return nil, nil, err + } + + node = &Node{} + node.Value = rest + return node, nil, nil +} + +// parseMaybeJSONToList determines if the argument appears to be a JSON array. If +// so, passes to parseJSON; if not, attempts to parse it as a whitespace +// delimited string. +func parseMaybeJSONToList(rest string, d *Directive) (*Node, map[string]bool, error) { + node, attrs, err := parseJSON(rest, d) + + if err == nil { + return node, attrs, nil + } + if err == errDockerfileNotStringArray { + return nil, nil, err + } + + return parseStringsWhitespaceDelimited(rest, d) +} + +// The HEALTHCHECK command is like parseMaybeJSON, but has an extra type argument. +func parseHealthConfig(rest string, d *Directive) (*Node, map[string]bool, error) { + // Find end of first argument + var sep int + for ; sep < len(rest); sep++ { + if unicode.IsSpace(rune(rest[sep])) { + break + } + } + next := sep + for ; next < len(rest); next++ { + if !unicode.IsSpace(rune(rest[next])) { + break + } + } + + if sep == 0 { + return nil, nil, nil + } + + typ := rest[:sep] + cmd, attrs, err := parseMaybeJSON(rest[next:], d) + if err != nil { + return nil, nil, err + } + + return &Node{Value: typ, Next: cmd}, attrs, err +} diff --git a/vendor/github.com/openshift/imagebuilder/dockerfile/parser/parser.go b/vendor/github.com/openshift/imagebuilder/dockerfile/parser/parser.go new file mode 100644 index 000000000..0223963e1 --- /dev/null +++ b/vendor/github.com/openshift/imagebuilder/dockerfile/parser/parser.go @@ -0,0 +1,355 @@ +// Package parser implements a parser and parse tree dumper for Dockerfiles. +package parser + +import ( + "bufio" + "bytes" + "fmt" + "io" + "regexp" + "runtime" + "strconv" + "strings" + "unicode" + + "github.com/openshift/imagebuilder/dockerfile/command" + "github.com/docker/docker/pkg/system" + "github.com/pkg/errors" +) + +// Node is a structure used to represent a parse tree. +// +// In the node there are three fields, Value, Next, and Children. Value is the +// current token's string value. Next is always the next non-child token, and +// children contains all the children. Here's an example: +// +// (value next (child child-next child-next-next) next-next) +// +// This data structure is frankly pretty lousy for handling complex languages, +// but lucky for us the Dockerfile isn't very complicated. This structure +// works a little more effectively than a "proper" parse tree for our needs. +// +type Node struct { + Value string // actual content + Next *Node // the next item in the current sexp + Children []*Node // the children of this sexp + Attributes map[string]bool // special attributes for this node + Original string // original line used before parsing + Flags []string // only top Node should have this set + StartLine int // the line in the original dockerfile where the node begins + endLine int // the line in the original dockerfile where the node ends +} + +// Dump dumps the AST defined by `node` as a list of sexps. +// Returns a string suitable for printing. +func (node *Node) Dump() string { + str := "" + str += node.Value + + if len(node.Flags) > 0 { + str += fmt.Sprintf(" %q", node.Flags) + } + + for _, n := range node.Children { + str += "(" + n.Dump() + ")\n" + } + + for n := node.Next; n != nil; n = n.Next { + if len(n.Children) > 0 { + str += " " + n.Dump() + } else { + str += " " + strconv.Quote(n.Value) + } + } + + return strings.TrimSpace(str) +} + +func (node *Node) lines(start, end int) { + node.StartLine = start + node.endLine = end +} + +// AddChild adds a new child node, and updates line information +func (node *Node) AddChild(child *Node, startLine, endLine int) { + child.lines(startLine, endLine) + if node.StartLine < 0 { + node.StartLine = startLine + } + node.endLine = endLine + node.Children = append(node.Children, child) +} + +var ( + dispatch map[string]func(string, *Directive) (*Node, map[string]bool, error) + tokenWhitespace = regexp.MustCompile(`[\t\v\f\r ]+`) + tokenEscapeCommand = regexp.MustCompile(`^#[ \t]*escape[ \t]*=[ \t]*(?P<escapechar>.).*$`) + tokenPlatformCommand = regexp.MustCompile(`^#[ \t]*platform[ \t]*=[ \t]*(?P<platform>.*)$`) + tokenComment = regexp.MustCompile(`^#.*$`) +) + +// DefaultEscapeToken is the default escape token +const DefaultEscapeToken = '\\' + +// defaultPlatformToken is the platform assumed for the build if not explicitly provided +var defaultPlatformToken = runtime.GOOS + +// Directive is the structure used during a build run to hold the state of +// parsing directives. +type Directive struct { + escapeToken rune // Current escape token + platformToken string // Current platform token + lineContinuationRegex *regexp.Regexp // Current line continuation regex + processingComplete bool // Whether we are done looking for directives + escapeSeen bool // Whether the escape directive has been seen + platformSeen bool // Whether the platform directive has been seen +} + +// setEscapeToken sets the default token for escaping characters in a Dockerfile. +func (d *Directive) setEscapeToken(s string) error { + if s != "`" && s != "\\" { + return fmt.Errorf("invalid ESCAPE '%s'. Must be ` or \\", s) + } + d.escapeToken = rune(s[0]) + d.lineContinuationRegex = regexp.MustCompile(`\` + s + `[ \t]*$`) + return nil +} + +// setPlatformToken sets the default platform for pulling images in a Dockerfile. +func (d *Directive) setPlatformToken(s string) error { + s = strings.ToLower(s) + valid := []string{runtime.GOOS} + if system.LCOWSupported() { + valid = append(valid, "linux") + } + for _, item := range valid { + if s == item { + d.platformToken = s + return nil + } + } + return fmt.Errorf("invalid PLATFORM '%s'. Must be one of %v", s, valid) +} + +// possibleParserDirective looks for one or more parser directives '# escapeToken=<char>' and +// '# platform=<string>'. Parser directives must precede any builder instruction +// or other comments, and cannot be repeated. +func (d *Directive) possibleParserDirective(line string) error { + if d.processingComplete { + return nil + } + + tecMatch := tokenEscapeCommand.FindStringSubmatch(strings.ToLower(line)) + if len(tecMatch) != 0 { + for i, n := range tokenEscapeCommand.SubexpNames() { + if n == "escapechar" { + if d.escapeSeen == true { + return errors.New("only one escape parser directive can be used") + } + d.escapeSeen = true + return d.setEscapeToken(tecMatch[i]) + } + } + } + + // TODO @jhowardmsft LCOW Support: Eventually this check can be removed, + // but only recognise a platform token if running in LCOW mode. + if system.LCOWSupported() { + tpcMatch := tokenPlatformCommand.FindStringSubmatch(strings.ToLower(line)) + if len(tpcMatch) != 0 { + for i, n := range tokenPlatformCommand.SubexpNames() { + if n == "platform" { + if d.platformSeen == true { + return errors.New("only one platform parser directive can be used") + } + d.platformSeen = true + return d.setPlatformToken(tpcMatch[i]) + } + } + } + } + + d.processingComplete = true + return nil +} + +// NewDefaultDirective returns a new Directive with the default escapeToken token +func NewDefaultDirective() *Directive { + directive := Directive{} + directive.setEscapeToken(string(DefaultEscapeToken)) + directive.setPlatformToken(defaultPlatformToken) + return &directive +} + +func init() { + // Dispatch Table. see line_parsers.go for the parse functions. + // The command is parsed and mapped to the line parser. The line parser + // receives the arguments but not the command, and returns an AST after + // reformulating the arguments according to the rules in the parser + // functions. Errors are propagated up by Parse() and the resulting AST can + // be incorporated directly into the existing AST as a next. + dispatch = map[string]func(string, *Directive) (*Node, map[string]bool, error){ + command.Add: parseMaybeJSONToList, + command.Arg: parseNameOrNameVal, + command.Cmd: parseMaybeJSON, + command.Copy: parseMaybeJSONToList, + command.Entrypoint: parseMaybeJSON, + command.Env: parseEnv, + command.Expose: parseStringsWhitespaceDelimited, + command.From: parseStringsWhitespaceDelimited, + command.Healthcheck: parseHealthConfig, + command.Label: parseLabel, + command.Maintainer: parseString, + command.Onbuild: parseSubCommand, + command.Run: parseMaybeJSON, + command.Shell: parseMaybeJSON, + command.StopSignal: parseString, + command.User: parseString, + command.Volume: parseMaybeJSONToList, + command.Workdir: parseString, + } +} + +// newNodeFromLine splits the line into parts, and dispatches to a function +// based on the command and command arguments. A Node is created from the +// result of the dispatch. +func newNodeFromLine(line string, directive *Directive) (*Node, error) { + cmd, flags, args, err := splitCommand(line) + if err != nil { + return nil, err + } + + fn := dispatch[cmd] + // Ignore invalid Dockerfile instructions + if fn == nil { + fn = parseIgnore + } + next, attrs, err := fn(args, directive) + if err != nil { + return nil, err + } + + return &Node{ + Value: cmd, + Original: line, + Flags: flags, + Next: next, + Attributes: attrs, + }, nil +} + +// Result is the result of parsing a Dockerfile +type Result struct { + AST *Node + EscapeToken rune + Platform string + Warnings []string +} + +// PrintWarnings to the writer +func (r *Result) PrintWarnings(out io.Writer) { + if len(r.Warnings) == 0 { + return + } + fmt.Fprintf(out, strings.Join(r.Warnings, "\n")+"\n") +} + +// Parse reads lines from a Reader, parses the lines into an AST and returns +// the AST and escape token +func Parse(rwc io.Reader) (*Result, error) { + d := NewDefaultDirective() + currentLine := 0 + root := &Node{StartLine: -1} + scanner := bufio.NewScanner(rwc) + warnings := []string{} + + var err error + for scanner.Scan() { + bytesRead := scanner.Bytes() + if currentLine == 0 { + // First line, strip the byte-order-marker if present + bytesRead = bytes.TrimPrefix(bytesRead, utf8bom) + } + bytesRead, err = processLine(d, bytesRead, true) + if err != nil { + return nil, err + } + currentLine++ + + startLine := currentLine + line, isEndOfLine := trimContinuationCharacter(string(bytesRead), d) + if isEndOfLine && line == "" { + continue + } + + var hasEmptyContinuationLine bool + for !isEndOfLine && scanner.Scan() { + bytesRead, err := processLine(d, scanner.Bytes(), false) + if err != nil { + return nil, err + } + currentLine++ + + if isEmptyContinuationLine(bytesRead) { + hasEmptyContinuationLine = true + continue + } + + continuationLine := string(bytesRead) + continuationLine, isEndOfLine = trimContinuationCharacter(continuationLine, d) + line += continuationLine + } + + if hasEmptyContinuationLine { + warning := "[WARNING]: Empty continuation line found in:\n " + line + warnings = append(warnings, warning) + } + + child, err := newNodeFromLine(line, d) + if err != nil { + return nil, err + } + root.AddChild(child, startLine, currentLine) + } + + if len(warnings) > 0 { + warnings = append(warnings, "[WARNING]: Empty continuation lines will become errors in a future release.") + } + return &Result{ + AST: root, + Warnings: warnings, + EscapeToken: d.escapeToken, + Platform: d.platformToken, + }, nil +} + +func trimComments(src []byte) []byte { + return tokenComment.ReplaceAll(src, []byte{}) +} + +func trimWhitespace(src []byte) []byte { + return bytes.TrimLeftFunc(src, unicode.IsSpace) +} + +func isEmptyContinuationLine(line []byte) bool { + return len(trimComments(trimWhitespace(line))) == 0 +} + +var utf8bom = []byte{0xEF, 0xBB, 0xBF} + +func trimContinuationCharacter(line string, d *Directive) (string, bool) { + if d.lineContinuationRegex.MatchString(line) { + line = d.lineContinuationRegex.ReplaceAllString(line, "") + return line, false + } + return line, true +} + +// TODO: remove stripLeftWhitespace after deprecation period. It seems silly +// to preserve whitespace on continuation lines. Why is that done? +func processLine(d *Directive, token []byte, stripLeftWhitespace bool) ([]byte, error) { + if stripLeftWhitespace { + token = trimWhitespace(token) + } + return trimComments(token), d.possibleParserDirective(string(token)) +} diff --git a/vendor/github.com/openshift/imagebuilder/dockerfile/parser/split_command.go b/vendor/github.com/openshift/imagebuilder/dockerfile/parser/split_command.go new file mode 100644 index 000000000..171f454f6 --- /dev/null +++ b/vendor/github.com/openshift/imagebuilder/dockerfile/parser/split_command.go @@ -0,0 +1,118 @@ +package parser + +import ( + "strings" + "unicode" +) + +// splitCommand takes a single line of text and parses out the cmd and args, +// which are used for dispatching to more exact parsing functions. +func splitCommand(line string) (string, []string, string, error) { + var args string + var flags []string + + // Make sure we get the same results irrespective of leading/trailing spaces + cmdline := tokenWhitespace.Split(strings.TrimSpace(line), 2) + cmd := strings.ToLower(cmdline[0]) + + if len(cmdline) == 2 { + var err error + args, flags, err = extractBuilderFlags(cmdline[1]) + if err != nil { + return "", nil, "", err + } + } + + return cmd, flags, strings.TrimSpace(args), nil +} + +func extractBuilderFlags(line string) (string, []string, error) { + // Parses the BuilderFlags and returns the remaining part of the line + + const ( + inSpaces = iota // looking for start of a word + inWord + inQuote + ) + + words := []string{} + phase := inSpaces + word := "" + quote := '\000' + blankOK := false + var ch rune + + for pos := 0; pos <= len(line); pos++ { + if pos != len(line) { + ch = rune(line[pos]) + } + + if phase == inSpaces { // Looking for start of word + if pos == len(line) { // end of input + break + } + if unicode.IsSpace(ch) { // skip spaces + continue + } + + // Only keep going if the next word starts with -- + if ch != '-' || pos+1 == len(line) || rune(line[pos+1]) != '-' { + return line[pos:], words, nil + } + + phase = inWord // found something with "--", fall through + } + if (phase == inWord || phase == inQuote) && (pos == len(line)) { + if word != "--" && (blankOK || len(word) > 0) { + words = append(words, word) + } + break + } + if phase == inWord { + if unicode.IsSpace(ch) { + phase = inSpaces + if word == "--" { + return line[pos:], words, nil + } + if blankOK || len(word) > 0 { + words = append(words, word) + } + word = "" + blankOK = false + continue + } + if ch == '\'' || ch == '"' { + quote = ch + blankOK = true + phase = inQuote + continue + } + if ch == '\\' { + if pos+1 == len(line) { + continue // just skip \ at end + } + pos++ + ch = rune(line[pos]) + } + word += string(ch) + continue + } + if phase == inQuote { + if ch == quote { + phase = inWord + continue + } + if ch == '\\' { + if pos+1 == len(line) { + phase = inWord + continue // just skip \ at end + } + pos++ + ch = rune(line[pos]) + } + word += string(ch) + } + } + + return "", words, nil +} |