diff options
Diffstat (limited to 'forged/internal/scfg')
-rw-r--r-- | forged/internal/scfg/.golangci.yaml | 25 | ||||
-rw-r--r-- | forged/internal/scfg/reader.go | 157 | ||||
-rw-r--r-- | forged/internal/scfg/scfg.go | 59 | ||||
-rw-r--r-- | forged/internal/scfg/struct.go | 82 | ||||
-rw-r--r-- | forged/internal/scfg/unmarshal.go | 375 | ||||
-rw-r--r-- | forged/internal/scfg/writer.go | 112 |
6 files changed, 810 insertions, 0 deletions
diff --git a/forged/internal/scfg/.golangci.yaml b/forged/internal/scfg/.golangci.yaml new file mode 100644 index 0000000..7db9793 --- /dev/null +++ b/forged/internal/scfg/.golangci.yaml @@ -0,0 +1,25 @@ +linters: + enable-all: true + disable: + - perfsprint + - wsl + - varnamelen + - nlreturn + - exhaustruct + - wrapcheck + - lll + - exhaustive + - intrange + - godox + - nestif + - err113 + - staticcheck + - errorlint + - cyclop + - nonamedreturns + - funlen + - gochecknoglobals + +issues: + max-issues-per-linter: 0 + max-same-issues: 0 diff --git a/forged/internal/scfg/reader.go b/forged/internal/scfg/reader.go new file mode 100644 index 0000000..6a2bedc --- /dev/null +++ b/forged/internal/scfg/reader.go @@ -0,0 +1,157 @@ +// SPDX-License-Identifier: MIT +// SPDX-FileCopyrightText: Copyright (c) 2020 Simon Ser <https://emersion.fr> + +package scfg + +import ( + "bufio" + "fmt" + "io" + "os" + "strings" +) + +// This limits the max block nesting depth to prevent stack overflows. +const maxNestingDepth = 1000 + +// Load loads a configuration file. +func Load(path string) (Block, error) { + f, err := os.Open(path) + if err != nil { + return nil, err + } + defer f.Close() + + return Read(f) +} + +// Read parses a configuration file from an io.Reader. +func Read(r io.Reader) (Block, error) { + scanner := bufio.NewScanner(r) + + dec := decoder{scanner: scanner} + block, closingBrace, err := dec.readBlock() + if err != nil { + return nil, err + } else if closingBrace { + return nil, fmt.Errorf("line %v: unexpected '}'", dec.lineno) + } + + return block, scanner.Err() +} + +type decoder struct { + scanner *bufio.Scanner + lineno int + blockDepth int +} + +// readBlock reads a block. closingBrace is true if parsing stopped on '}' +// (otherwise, it stopped on Scanner.Scan). +func (dec *decoder) readBlock() (block Block, closingBrace bool, err error) { + dec.blockDepth++ + defer func() { + dec.blockDepth-- + }() + + if dec.blockDepth >= maxNestingDepth { + return nil, false, fmt.Errorf("exceeded max block depth") + } + + for dec.scanner.Scan() { + dec.lineno++ + + l := dec.scanner.Text() + words, err := splitWords(l) + if err != nil { + return nil, false, fmt.Errorf("line %v: %v", dec.lineno, err) + } else if len(words) == 0 { + continue + } + + if len(words) == 1 && l[len(l)-1] == '}' { + closingBrace = true + break + } + + var d *Directive + if words[len(words)-1] == "{" && l[len(l)-1] == '{' { + words = words[:len(words)-1] + + var name string + params := words + if len(words) > 0 { + name, params = words[0], words[1:] + } + + startLineno := dec.lineno + childBlock, childClosingBrace, err := dec.readBlock() + if err != nil { + return nil, false, err + } else if !childClosingBrace { + return nil, false, fmt.Errorf("line %v: unterminated block", startLineno) + } + + // Allows callers to tell apart "no block" and "empty block" + if childBlock == nil { + childBlock = Block{} + } + + d = &Directive{Name: name, Params: params, Children: childBlock, lineno: dec.lineno} + } else { + d = &Directive{Name: words[0], Params: words[1:], lineno: dec.lineno} + } + block = append(block, d) + } + + return block, closingBrace, nil +} + +func splitWords(l string) ([]string, error) { + var ( + words []string + sb strings.Builder + escape bool + quote rune + wantWSP bool + ) + for _, ch := range l { + switch { + case escape: + sb.WriteRune(ch) + escape = false + case wantWSP && (ch != ' ' && ch != '\t'): + return words, fmt.Errorf("atom not allowed after quoted string") + case ch == '\\': + escape = true + case quote != 0 && ch == quote: + quote = 0 + wantWSP = true + if sb.Len() == 0 { + words = append(words, "") + } + case quote == 0 && len(words) == 0 && sb.Len() == 0 && ch == '#': + return nil, nil + case quote == 0 && (ch == '\'' || ch == '"'): + if sb.Len() > 0 { + return words, fmt.Errorf("quoted string not allowed after atom") + } + quote = ch + case quote == 0 && (ch == ' ' || ch == '\t'): + if sb.Len() > 0 { + words = append(words, sb.String()) + } + sb.Reset() + wantWSP = false + default: + sb.WriteRune(ch) + } + } + if quote != 0 { + return words, fmt.Errorf("unterminated quoted string") + } + if sb.Len() > 0 { + words = append(words, sb.String()) + } + return words, nil +} diff --git a/forged/internal/scfg/scfg.go b/forged/internal/scfg/scfg.go new file mode 100644 index 0000000..4533e63 --- /dev/null +++ b/forged/internal/scfg/scfg.go @@ -0,0 +1,59 @@ +// SPDX-License-Identifier: MIT +// SPDX-FileCopyrightText: Copyright (c) 2020 Simon Ser <https://emersion.fr> + +// Package scfg parses and formats configuration files. +// Note that this fork of scfg behaves differently from upstream scfg. +package scfg + +import ( + "fmt" +) + +// Block is a list of directives. +type Block []*Directive + +// GetAll returns a list of directives with the provided name. +func (blk Block) GetAll(name string) []*Directive { + l := make([]*Directive, 0, len(blk)) + for _, child := range blk { + if child.Name == name { + l = append(l, child) + } + } + return l +} + +// Get returns the first directive with the provided name. +func (blk Block) Get(name string) *Directive { + for _, child := range blk { + if child.Name == name { + return child + } + } + return nil +} + +// Directive is a configuration directive. +type Directive struct { + Name string + Params []string + + Children Block + + lineno int +} + +// ParseParams extracts parameters from the directive. It errors out if the +// user hasn't provided enough parameters. +func (d *Directive) ParseParams(params ...*string) error { + if len(d.Params) < len(params) { + return fmt.Errorf("directive %q: want %v params, got %v", d.Name, len(params), len(d.Params)) + } + for i, ptr := range params { + if ptr == nil { + continue + } + *ptr = d.Params[i] + } + return nil +} diff --git a/forged/internal/scfg/struct.go b/forged/internal/scfg/struct.go new file mode 100644 index 0000000..98ec943 --- /dev/null +++ b/forged/internal/scfg/struct.go @@ -0,0 +1,82 @@ +// SPDX-License-Identifier: MIT +// SPDX-FileCopyrightText: Copyright (c) 2020 Simon Ser <https://emersion.fr> + +package scfg + +import ( + "fmt" + "reflect" + "strings" + "sync" +) + +// structInfo contains scfg metadata for structs. +type structInfo struct { + param int // index of field storing parameters + children map[string]int // indices of fields storing child directives +} + +var ( + structCacheMutex sync.Mutex + structCache = make(map[reflect.Type]*structInfo) +) + +func getStructInfo(t reflect.Type) (*structInfo, error) { + structCacheMutex.Lock() + defer structCacheMutex.Unlock() + + if info := structCache[t]; info != nil { + return info, nil + } + + info := &structInfo{ + param: -1, + children: make(map[string]int), + } + + for i := 0; i < t.NumField(); i++ { + f := t.Field(i) + if f.Anonymous { + return nil, fmt.Errorf("scfg: anonymous struct fields are not supported") + } else if !f.IsExported() { + continue + } + + tag := f.Tag.Get("scfg") + parts := strings.Split(tag, ",") + k, options := parts[0], parts[1:] + if k == "-" { + continue + } else if k == "" { + k = f.Name + } + + isParam := false + for _, opt := range options { + switch opt { + case "param": + isParam = true + default: + return nil, fmt.Errorf("scfg: invalid option %q in struct tag", opt) + } + } + + if isParam { + if info.param >= 0 { + return nil, fmt.Errorf("scfg: param option specified multiple times in struct tag in %v", t) + } + if parts[0] != "" { + return nil, fmt.Errorf("scfg: name must be empty when param option is specified in struct tag in %v", t) + } + info.param = i + } else { + if _, ok := info.children[k]; ok { + return nil, fmt.Errorf("scfg: key %q specified multiple times in struct tag in %v", k, t) + } + info.children[k] = i + } + } + + structCache[t] = info + return info, nil +} diff --git a/forged/internal/scfg/unmarshal.go b/forged/internal/scfg/unmarshal.go new file mode 100644 index 0000000..e9e1a52 --- /dev/null +++ b/forged/internal/scfg/unmarshal.go @@ -0,0 +1,375 @@ +// SPDX-License-Identifier: MIT +// SPDX-FileCopyrightText: Copyright (c) 2020 Simon Ser <https://emersion.fr> +// SPDX-FileCopyrightText: Copyright (c) 2025 Runxi Yu <https://runxiyu.org> + +package scfg + +import ( + "encoding" + "fmt" + "io" + "reflect" + "strconv" +) + +// Decoder reads and decodes an scfg document from an input stream. +type Decoder struct { + r io.Reader + unknownDirectives []*Directive +} + +// NewDecoder returns a new decoder which reads from r. +func NewDecoder(r io.Reader) *Decoder { + return &Decoder{r: r} +} + +// UnknownDirectives returns a slice of all unknown directives encountered +// during Decode. +func (dec *Decoder) UnknownDirectives() []*Directive { + return dec.unknownDirectives +} + +// Decode reads scfg document from the input and stores it in the value pointed +// to by v. +// +// If v is nil or not a pointer, Decode returns an error. +// +// Blocks can be unmarshaled to: +// +// - Maps. Each directive is unmarshaled into a map entry. The map key must +// be a string. +// - Structs. Each directive is unmarshaled into a struct field. +// +// Duplicate directives are not allowed, unless the struct field or map value +// is a slice of values representing a directive: structs or maps. +// +// Directives can be unmarshaled to: +// +// - Maps. The children block is unmarshaled into the map. Parameters are not +// allowed. +// - Structs. The children block is unmarshaled into the struct. Parameters +// are allowed if one of the struct fields contains the "param" option in +// its tag. +// - Slices. Parameters are unmarshaled into the slice. Children blocks are +// not allowed. +// - Arrays. Parameters are unmarshaled into the array. The number of +// parameters must match exactly the length of the array. Children blocks +// are not allowed. +// - Strings, booleans, integers, floating-point values, values implementing +// encoding.TextUnmarshaler. Only a single parameter is allowed and is +// unmarshaled into the value. Children blocks are not allowed. +// +// The decoding of each struct field can be customized by the format string +// stored under the "scfg" key in the struct field's tag. The tag contains the +// name of the field possibly followed by a comma-separated list of options. +// The name may be empty in order to specify options without overriding the +// default field name. As a special case, if the field name is "-", the field +// is ignored. The "param" option specifies that directive parameters are +// stored in this field (the name must be empty). +func (dec *Decoder) Decode(v interface{}) error { + block, err := Read(dec.r) + if err != nil { + return err + } + + rv := reflect.ValueOf(v) + if rv.Kind() != reflect.Ptr || rv.IsNil() { + return fmt.Errorf("scfg: invalid value for unmarshaling") + } + + return dec.unmarshalBlock(block, rv) +} + +func (dec *Decoder) unmarshalBlock(block Block, v reflect.Value) error { + v = unwrapPointers(v) + t := v.Type() + + dirsByName := make(map[string][]*Directive, len(block)) + for _, dir := range block { + dirsByName[dir.Name] = append(dirsByName[dir.Name], dir) + } + + switch v.Kind() { + case reflect.Map: + if t.Key().Kind() != reflect.String { + return fmt.Errorf("scfg: map key type must be string") + } + if v.IsNil() { + v.Set(reflect.MakeMap(t)) + } else if v.Len() > 0 { + clearMap(v) + } + + for name, dirs := range dirsByName { + mv := reflect.New(t.Elem()).Elem() + if err := dec.unmarshalDirectiveList(dirs, mv); err != nil { + return err + } + v.SetMapIndex(reflect.ValueOf(name), mv) + } + + case reflect.Struct: + si, err := getStructInfo(t) + if err != nil { + return err + } + + seen := make(map[int]bool) + + for name, dirs := range dirsByName { + fieldIndex, ok := si.children[name] + if !ok { + dec.unknownDirectives = append(dec.unknownDirectives, dirs...) + continue + } + fv := v.Field(fieldIndex) + if err := dec.unmarshalDirectiveList(dirs, fv); err != nil { + return err + } + seen[fieldIndex] = true + } + + for name, fieldIndex := range si.children { + if fieldIndex == si.param { + continue + } + if _, ok := seen[fieldIndex]; !ok { + return fmt.Errorf("scfg: missing required directive %q", name) + } + } + + default: + return fmt.Errorf("scfg: unsupported type for unmarshaling blocks: %v", t) + } + + return nil +} + +func (dec *Decoder) unmarshalDirectiveList(dirs []*Directive, v reflect.Value) error { + v = unwrapPointers(v) + t := v.Type() + + if v.Kind() != reflect.Slice || !isDirectiveType(t.Elem()) { + if len(dirs) > 1 { + return newUnmarshalDirectiveError(dirs[1], "directive must not be specified more than once") + } + return dec.unmarshalDirective(dirs[0], v) + } + + sv := reflect.MakeSlice(t, len(dirs), len(dirs)) + for i, dir := range dirs { + if err := dec.unmarshalDirective(dir, sv.Index(i)); err != nil { + return err + } + } + v.Set(sv) + return nil +} + +// isDirectiveType checks whether a type can only be unmarshaled as a +// directive, not as a parameter. Accepting too many types here would result in +// ambiguities, see: +// https://lists.sr.ht/~emersion/public-inbox/%3C20230629132458.152205-1-contact%40emersion.fr%3E#%3Ch4Y2peS_YBqY3ar4XlmPDPiNBFpYGns3EBYUx3_6zWEhV2o8_-fBQveRujGADWYhVVCucHBEryFGoPtpC3d3mQ-x10pWnFogfprbQTSvtxc=@emersion.fr%3E +func isDirectiveType(t reflect.Type) bool { + for t.Kind() == reflect.Ptr { + t = t.Elem() + } + + textUnmarshalerType := reflect.TypeOf((*encoding.TextUnmarshaler)(nil)).Elem() + if reflect.PtrTo(t).Implements(textUnmarshalerType) { + return false + } + + switch t.Kind() { + case reflect.Struct, reflect.Map: + return true + default: + return false + } +} + +func (dec *Decoder) unmarshalDirective(dir *Directive, v reflect.Value) error { + v = unwrapPointers(v) + t := v.Type() + + if v.CanAddr() { + if _, ok := v.Addr().Interface().(encoding.TextUnmarshaler); ok { + if len(dir.Children) != 0 { + return newUnmarshalDirectiveError(dir, "directive requires zero children") + } + return unmarshalParamList(dir, v) + } + } + + switch v.Kind() { + case reflect.Map: + if len(dir.Params) > 0 { + return newUnmarshalDirectiveError(dir, "directive requires zero parameters") + } + if err := dec.unmarshalBlock(dir.Children, v); err != nil { + return err + } + case reflect.Struct: + si, err := getStructInfo(t) + if err != nil { + return err + } + + if si.param >= 0 { + if err := unmarshalParamList(dir, v.Field(si.param)); err != nil { + return err + } + } else { + if len(dir.Params) > 0 { + return newUnmarshalDirectiveError(dir, "directive requires zero parameters") + } + } + + if err := dec.unmarshalBlock(dir.Children, v); err != nil { + return err + } + default: + if len(dir.Children) != 0 { + return newUnmarshalDirectiveError(dir, "directive requires zero children") + } + if err := unmarshalParamList(dir, v); err != nil { + return err + } + } + return nil +} + +func unmarshalParamList(dir *Directive, v reflect.Value) error { + switch v.Kind() { + case reflect.Slice: + t := v.Type() + sv := reflect.MakeSlice(t, len(dir.Params), len(dir.Params)) + for i, param := range dir.Params { + if err := unmarshalParam(param, sv.Index(i)); err != nil { + return newUnmarshalParamError(dir, i, err) + } + } + v.Set(sv) + case reflect.Array: + if len(dir.Params) != v.Len() { + return newUnmarshalDirectiveError(dir, fmt.Sprintf("directive requires exactly %v parameters", v.Len())) + } + for i, param := range dir.Params { + if err := unmarshalParam(param, v.Index(i)); err != nil { + return newUnmarshalParamError(dir, i, err) + } + } + default: + if len(dir.Params) != 1 { + return newUnmarshalDirectiveError(dir, "directive requires exactly one parameter") + } + if err := unmarshalParam(dir.Params[0], v); err != nil { + return newUnmarshalParamError(dir, 0, err) + } + } + + return nil +} + +func unmarshalParam(param string, v reflect.Value) error { + v = unwrapPointers(v) + t := v.Type() + + // TODO: improve our logic following: + // https://cs.opensource.google/go/go/+/refs/tags/go1.21.5:src/encoding/json/decode.go;drc=b9b8cecbfc72168ca03ad586cc2ed52b0e8db409;l=421 + if v.CanAddr() { + if v, ok := v.Addr().Interface().(encoding.TextUnmarshaler); ok { + return v.UnmarshalText([]byte(param)) + } + } + + switch v.Kind() { + case reflect.String: + v.Set(reflect.ValueOf(param)) + case reflect.Bool: + switch param { + case "true": + v.Set(reflect.ValueOf(true)) + case "false": + v.Set(reflect.ValueOf(false)) + default: + return fmt.Errorf("invalid bool parameter %q", param) + } + case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64: + i, err := strconv.ParseInt(param, 10, t.Bits()) + if err != nil { + return fmt.Errorf("invalid %v parameter: %v", t, err) + } + v.Set(reflect.ValueOf(i).Convert(t)) + case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64: + u, err := strconv.ParseUint(param, 10, t.Bits()) + if err != nil { + return fmt.Errorf("invalid %v parameter: %v", t, err) + } + v.Set(reflect.ValueOf(u).Convert(t)) + case reflect.Float32, reflect.Float64: + f, err := strconv.ParseFloat(param, t.Bits()) + if err != nil { + return fmt.Errorf("invalid %v parameter: %v", t, err) + } + v.Set(reflect.ValueOf(f).Convert(t)) + default: + return fmt.Errorf("unsupported type for unmarshaling parameter: %v", t) + } + + return nil +} + +func unwrapPointers(v reflect.Value) reflect.Value { + for v.Kind() == reflect.Ptr { + if v.IsNil() { + v.Set(reflect.New(v.Type().Elem())) + } + v = v.Elem() + } + return v +} + +func clearMap(v reflect.Value) { + for _, k := range v.MapKeys() { + v.SetMapIndex(k, reflect.Value{}) + } +} + +type unmarshalDirectiveError struct { + lineno int + name string + msg string +} + +func newUnmarshalDirectiveError(dir *Directive, msg string) *unmarshalDirectiveError { + return &unmarshalDirectiveError{ + name: dir.Name, + lineno: dir.lineno, + msg: msg, + } +} + +func (err *unmarshalDirectiveError) Error() string { + return fmt.Sprintf("line %v, directive %q: %v", err.lineno, err.name, err.msg) +} + +type unmarshalParamError struct { + lineno int + directive string + paramIndex int + err error +} + +func newUnmarshalParamError(dir *Directive, paramIndex int, err error) *unmarshalParamError { + return &unmarshalParamError{ + directive: dir.Name, + lineno: dir.lineno, + paramIndex: paramIndex, + err: err, + } +} + +func (err *unmarshalParamError) Error() string { + return fmt.Sprintf("line %v, directive %q, parameter %v: %v", err.lineno, err.directive, err.paramIndex+1, err.err) +} diff --git a/forged/internal/scfg/writer.go b/forged/internal/scfg/writer.go new file mode 100644 index 0000000..02a07fe --- /dev/null +++ b/forged/internal/scfg/writer.go @@ -0,0 +1,112 @@ +// SPDX-License-Identifier: MIT +// SPDX-FileCopyrightText: Copyright (c) 2020 Simon Ser <https://emersion.fr> + +package scfg + +import ( + "errors" + "io" + "strings" +) + +var errDirEmptyName = errors.New("scfg: directive with empty name") + +// Write writes a parsed configuration to the provided io.Writer. +func Write(w io.Writer, blk Block) error { + enc := newEncoder(w) + err := enc.encodeBlock(blk) + return err +} + +// encoder write SCFG directives to an output stream. +type encoder struct { + w io.Writer + lvl int + err error +} + +// newEncoder returns a new encoder that writes to w. +func newEncoder(w io.Writer) *encoder { + return &encoder{w: w} +} + +func (enc *encoder) push() { + enc.lvl++ +} + +func (enc *encoder) pop() { + enc.lvl-- +} + +func (enc *encoder) writeIndent() { + for i := 0; i < enc.lvl; i++ { + enc.write([]byte("\t")) + } +} + +func (enc *encoder) write(p []byte) { + if enc.err != nil { + return + } + _, enc.err = enc.w.Write(p) +} + +func (enc *encoder) encodeBlock(blk Block) error { + for _, dir := range blk { + if err := enc.encodeDir(*dir); err != nil { + return err + } + } + return enc.err +} + +func (enc *encoder) encodeDir(dir Directive) error { + if enc.err != nil { + return enc.err + } + + if dir.Name == "" { + enc.err = errDirEmptyName + return enc.err + } + + enc.writeIndent() + enc.write([]byte(maybeQuote(dir.Name))) + for _, p := range dir.Params { + enc.write([]byte(" ")) + enc.write([]byte(maybeQuote(p))) + } + + if len(dir.Children) > 0 { + enc.write([]byte(" {\n")) + enc.push() + if err := enc.encodeBlock(dir.Children); err != nil { + return err + } + enc.pop() + + enc.writeIndent() + enc.write([]byte("}")) + } + enc.write([]byte("\n")) + + return enc.err +} + +const specialChars = "\"\\\r\n'{} \t" + +func maybeQuote(s string) string { + if s == "" || strings.ContainsAny(s, specialChars) { + var sb strings.Builder + sb.WriteByte('"') + for _, ch := range s { + if strings.ContainsRune(`"\`, ch) { + sb.WriteByte('\\') + } + sb.WriteRune(ch) + } + sb.WriteByte('"') + return sb.String() + } + return s +} |