code reorganization
[epoint] / pkg / document / document.go
diff --git a/pkg/document/document.go b/pkg/document/document.go
new file mode 100644 (file)
index 0000000..0c8b0f4
--- /dev/null
@@ -0,0 +1,695 @@
+// Package document implements epoint document parsing and creation.
+//
+// An epoint document is an OpenPGP (RFC 4880) clear signed
+// utf-8 text of key-value pairs.
+// The body contains a content-type MIME header so the document
+// can be used in OpenPGP/MIME (RFC 3156) emails.
+// The format of the key-value pairs are similar to MIME header
+// fields: keys and values are separated by ": ", repeated keys
+// are not allowed, long values can be split before a space.
+//
+// Example:
+//
+// -----BEGIN PGP SIGNED MESSAGE-----
+// Hash: SHA1
+//
+// Content-Type: text/plain.epoint.type; charset=utf-8
+//
+// Key: Value1
+// Another-Key: Value2
+// Last-Key: Long
+//  value that spans
+//  multiple lines
+// -----BEGIN PGP SIGNATURE-----
+//
+// pgp signature
+// -----END PGP SIGNATURE-----
+package document
+
+// TODO: error wrapper (so reporting to user or creating bounce cert is simple)
+// TODO: optional fields: exact semantics ("" vs "-" vs nil)
+// TODO: trailing space handling in ParseFields
+// TODO: fields of notice (last notice, serial, failure notice,..)
+// TODO: limits and cert type specific input validation
+// TODO: fix Cert mess
+// TODO: nonce is id, id is even number of hex digits (require only drawer.nonce to be uniq)
+// TODO: denom, issuer from key (key representation: armor?)
+
+import (
+       "bytes"
+       "crypto"
+       "crypto/openpgp"
+       "crypto/openpgp/armor"
+       "crypto/openpgp/packet"
+       "crypto/sha1"
+       "encoding/hex"
+       "fmt"
+       "reflect"
+       "strconv"
+       "strings"
+       "time"
+)
+
+// limits
+const (
+       MaxFields             = 20
+       MaxLineLength         = 160  // 1 sha512 + 1 key (without \n)
+       MaxValueLength        = 1300 // 20 sha256 space separated (without \n)
+       MaxNonceLength        = 20
+       MaxDenominationLength = 100
+)
+
+const ClearSignedHeader = "-----BEGIN PGP SIGNED MESSAGE-----"
+
+// MIME type for epoint documents, see RFC 4288
+var ContentType = map[string]string{
+       "Draft":      "text/vnd.epoint.draft; charset=utf-8",
+       "Notice":     "text/vnd.epoint.notice; charset=utf-8",
+       "DebitCert":  "text/vnd.epoint.debit; charset=utf-8",
+       "CreditCert": "text/vnd.epoint.credit; charset=utf-8",
+       "BounceCert": "text/vnd.epoint.bounce; charset=utf-8",
+}
+
+// OpenPGP signed cleartext document representation
+type Signed struct {
+       // Sign and CleanSigned sets Hash for FormatSigned
+       // TODO: CreationDate
+       Hash string
+       // Signed text (no dash escape, no trailing space, \n new lines)
+       Body []byte
+       // Armored detached text signature of the Body
+       Signature []byte
+}
+
+// parsed epoint document
+type Document struct {
+       Type   string
+       Fields map[string]string
+       Order  []string
+}
+
+var fieldtype = map[string]string{
+       "Amount":             "int",
+       "Authorized-By":      "id",
+       "Balance":            "int",
+       "Beneficiary":        "id",
+       "Date":               "date",
+       "Debit-Cert":         "id",
+       "Denomination":       "text",
+       "Difference":         "int",
+       "Draft":              "id",
+       "Drawer":             "id",
+       "Expiry-Date":        "date",
+       "Holder":             "id",
+       "Issuer":             "id",
+       "Last-Cert":          "id",
+       "Last-Credit-Serial": "int",
+       "Last-Debit-Serial":  "int",
+       "Maturity-Date":      "date",
+       "Nonce":              "id",
+       "Notes":              "text",
+       "References":         "ids",
+       "Serial":             "int",
+}
+
+var fieldname = map[string]string{
+       "AuthorizedBy":     "Authorized-By",
+       "DebitCert":        "Debit-Cert",
+       "ExpiryDate":       "Expiry-Date",
+       "LastCert":         "Last-Cert",
+       "LastCreditSerial": "Last-Credit-Serial",
+       "LastDebitSerial":  "Last-Debit-Serial",
+       "MaturityDate":     "Maturity-Date",
+}
+
+type Draft struct {
+       Drawer       string
+       Beneficiary  string
+       Amount       int64
+       Denomination string
+       Issuer       string
+       AuthorizedBy string
+       MaturityDate *int64 // optional
+       ExpiryDate   *int64 // optional
+       Nonce        string
+       Notes        *string // optional
+}
+
+type Notice struct {
+       Date         int64
+       AuthorizedBy string
+       Notes        *string  // optional
+       References   []string // may be empty (startup notice)
+}
+
+type Cert struct {
+       Holder           string
+       Serial           int64
+       Balance          int64
+       Denomination     string
+       Issuer           string
+       Date             int64
+       AuthorizedBy     string
+       Notes            *string // optional
+       LastDebitSerial  int64   // 0 if none
+       LastCreditSerial int64   // 0 if none
+       LastCert         *string // nil if serial == 1
+       References       []string
+       Difference       int64
+       Draft            string
+}
+
+type DebitCert struct {
+       Cert
+       Beneficiary      string
+}
+
+type CreditCert struct {
+       Cert
+       Drawer           string
+       DebitCert        string
+}
+
+type BounceCert struct {
+       Drawer       string
+       Draft        string
+       LastCert     *string // optional
+       Balance      int64   // 0 if none
+       Date         int64
+       AuthorizedBy string
+       Notes        *string // optional
+       References   []string
+}
+
+func ToCert(v interface{}) (cert *Cert, err error) {
+       cert = new(Cert)
+       switch x := v.(type) {
+       case *DebitCert:
+               cert = &x.Cert
+       case *CreditCert:
+               cert = &x.Cert
+       default:
+               err = fmt.Errorf("ToCert: only debit or credit document can be converted to cert")
+       }
+       return
+}
+
+func cleanBody(s []byte) []byte {
+       nl := []byte{'\n'}
+       a := bytes.Split(s, nl)
+       for i := range a {
+               a[i] = bytes.TrimRight(a[i], " \t")
+       }
+       return bytes.Join(a, nl)
+}
+
+// sha1 sum of the (cleaned) document body as uppercase hex string
+func Id(c *Signed) string {
+       h := sha1.New()
+       h.Write(c.Body)
+       return fmt.Sprintf("%040X", h.Sum())
+}
+
+// parse an epoint document without checking the signature and format details
+func Parse(s []byte) (iv interface{}, c *Signed, err error) {
+       c, err = ParseSigned(s)
+       if err != nil {
+               return
+       }
+       doc, err := ParseDocument(c.Body)
+       if err != nil {
+               return
+       }
+       iv, err = ParseStruct(doc)
+       return
+}
+
+// format and sign an epoint document
+func Format(iv interface{}, key *openpgp.Entity) (s []byte, c *Signed, err error) {
+       doc, err := FormatStruct(iv)
+       if err != nil {
+               return
+       }
+       body, err := FormatDocument(doc)
+       if err != nil {
+               return
+       }
+       c, err = Sign(body, key)
+       if err != nil {
+               return
+       }
+       s, err = FormatSigned(c)
+       return
+}
+
+// verify an epoint document, return the cleaned version as well
+func Verify(c *Signed, key openpgp.KeyRing) (err error) {
+       msg := bytes.NewBuffer(c.Body)
+       sig := bytes.NewBuffer(c.Signature)
+       // TODO: verify signature
+       _, _ = msg, sig
+       //      _, err = openpgp.CheckArmoredDetachedSignature(key, msg, sig)
+       return
+}
+
+// sign body with given secret key
+func Sign(body []byte, key *openpgp.Entity) (c *Signed, err error) {
+       c = new(Signed)
+       c.Hash = "SHA256"
+       c.Body = cleanBody(body)
+       w := new(bytes.Buffer)
+       err = openpgp.ArmoredDetachSignText(w, key, bytes.NewBuffer(c.Body))
+       if err != nil {
+               return
+       }
+       // close armored document with a \n
+       _, _ = w.Write([]byte{'\n'})
+       c.Signature = w.Bytes()
+       return
+}
+
+// split a clear signed document into body and armored signature
+func ParseSigned(s []byte) (c *Signed, err error) {
+       // look for clear signed header
+       for !bytes.HasPrefix(s, []byte(ClearSignedHeader)) {
+               _, s = getLine(s)
+               if len(s) == 0 {
+                       err = fmt.Errorf("ParseSigned: clear signed header is missing")
+                       return
+               }
+       }
+       s = s[len(ClearSignedHeader):]
+       // end of line after the header
+       empty, s := getLine(s)
+       if len(empty) != 0 {
+               err = fmt.Errorf("ParseSigned: bad clear signed header")
+               return
+       }
+       // skip all hash headers, section 7.
+       for bytes.HasPrefix(s, []byte("Hash: ")) {
+               _, s = getLine(s)
+       }
+       // skip empty line
+       empty, s = getLine(s)
+       if len(empty) != 0 {
+               err = fmt.Errorf("ParseSigned: expected an empty line after armor headers")
+               return
+       }
+       lines := [][]byte{}
+       for !bytes.HasPrefix(s, []byte("-----BEGIN")) {
+               var line []byte
+               line, s = getLine(s)
+               // dash unescape, section 7.1.
+               if bytes.HasPrefix(line, []byte("- ")) {
+                       line = line[2:]
+               }
+               // empty values are not supported: "Key: \n"
+               lines = append(lines, bytes.TrimRight(line, " \t"))
+       }
+       c = new(Signed)
+       // last line is not closed by \n
+       c.Body = bytes.Join(lines, []byte("\n"))
+       // signature is just the rest of the input data
+       c.Signature = s
+       return
+}
+
+// clean up, check and reencode signature
+// used on drafts before calculating the signed document hash
+func CleanSigned(c *Signed) (err error) {
+       b, err := armor.Decode(bytes.NewBuffer(c.Signature))
+       if err != nil {
+               return
+       }
+       if b.Type != openpgp.SignatureType {
+               err = fmt.Errorf("CleanSigned: invalid armored signature type")
+               return
+       }
+       p, err := packet.Read(b.Body)
+       if err != nil {
+               return
+       }
+       sig, ok := p.(*packet.Signature)
+       if !ok {
+               err = fmt.Errorf("CleanSigned: invalid signature packet")
+               return
+       }
+       // section 5.2.3
+       if sig.SigType != packet.SigTypeText {
+               err = fmt.Errorf("CleanSigned: expected text signature")
+               return
+       }
+       switch sig.Hash {
+       case crypto.SHA1:
+               c.Hash = "SHA1"
+       case crypto.SHA256:
+               c.Hash = "SHA256"
+       default:
+               err = fmt.Errorf("CleanSigned: expected SHA1 or SHA256 signature hash")
+               return
+       }
+       // TODO: check CreationTime and other subpackets
+       if sig.SigLifetimeSecs != nil && *sig.SigLifetimeSecs != 0 {
+               err = fmt.Errorf("CleanSigned: signature must not expire")
+               return
+       }
+       out := new(bytes.Buffer)
+       w, err := armor.Encode(out, openpgp.SignatureType, nil)
+       if err != nil {
+               return
+       }
+       err = sig.Serialize(w)
+       if err != nil {
+               return
+       }
+       err = w.Close()
+       if err != nil {
+               return
+       }
+       c.Signature = out.Bytes()
+       return
+}
+
+// create clear signed document
+func FormatSigned(c *Signed) (data []byte, err error) {
+       s := ClearSignedHeader + "\n"
+       if c.Hash != "" {
+               s += "Hash: " + c.Hash + "\n"
+       }
+       s += "\n"
+       s += string(c.Body)
+       s += "\n"
+       s += string(c.Signature)
+       data = []byte(s)
+       return
+}
+
+// parse type and fields of a document body
+func ParseDocument(body []byte) (doc *Document, err error) {
+       // parse content type header first
+       fields, s, err := ParseFields(body)
+       if err != nil {
+               return
+       }
+       ctype, ok := fields["Content-Type"]
+       if len(fields) != 1 || !ok {
+               return nil, fmt.Errorf("ParseBody: expected a single Content-Type header field")
+       }
+       doc = new(Document)
+       for k, v := range ContentType {
+               if ctype == v {
+                       doc.Type = k
+                       break
+               }
+       }
+       if doc.Type == "" {
+               return nil, fmt.Errorf("ParseBody: unknown Content-Type: %s", ctype)
+       }
+       // TODO: doc.Order
+       doc.Fields, s, err = ParseFields(s)
+       if err == nil && len(s) > 0 {
+               err = fmt.Errorf("ParseBody: extra data after fields: %q", s)
+       }
+       return
+}
+
+// create document body
+func FormatDocument(doc *Document) (body []byte, err error) {
+       ctype, ok := ContentType[doc.Type]
+       if !ok {
+               err = fmt.Errorf("FormatDocument: unknown document type: %s", doc.Type)
+               return
+       }
+       s := "Content-Type: " + ctype + "\n\n"
+       for _, k := range doc.Order {
+               s += k + ": " + doc.Fields[k] + "\n"
+       }
+       return []byte(s), nil
+}
+
+// parse doc fields into a struct according to the document type
+func parseStruct(v reflect.Value, fields map[string]string, seen map[string]bool) (err error) {
+       t := v.Type()
+       n := v.NumField()
+       for i := 0; i < n && err == nil; i++ {
+               ft := t.Field(i)
+               fv := v.Field(i)
+               if ft.Anonymous && fv.Kind() == reflect.Struct {
+                       err = parseStruct(fv, fields, seen)
+                       continue
+               }
+               key := fieldname[ft.Name]
+               if key == "" {
+                       key = ft.Name
+               }
+               s, ok := fields[key]
+               if !ok {
+                       if fv.Kind() == reflect.Ptr {
+                               // missing optional key: leave the pointer as nil
+                               continue
+                       }
+                       return fmt.Errorf("ParseStruct: field %s of %s is missing\n", key, t.Name())
+               }
+               seen[key] = true
+               if fv.Kind() == reflect.Ptr {
+                       if s == "" || s == "-" {
+                               // TODO
+                               // empty optional key: same as missing
+                               continue
+                       }
+                       fv.Set(reflect.New(fv.Type().Elem()))
+                       fv = fv.Elem()
+               }
+               switch fieldtype[key] {
+               case "id":
+                       var val string
+                       val, err = parseId(s)
+                       fv.SetString(val)
+               case "text":
+                       var val string
+                       val, err = parseString(s)
+                       fv.SetString(val)
+               case "int":
+                       var val int64
+                       val, err = strconv.Atoi64(s)
+                       fv.SetInt(val)
+               case "date":
+                       var val int64
+                       val, err = parseDate(s)
+                       fv.SetInt(val)
+               case "ids":
+                       // TODO: empty slice?
+                       ids := strings.Split(s, " ")
+                       val := make([]string, len(ids))
+                       for j, id := range ids {
+                               val[j], err = parseId(id)
+                               if err != nil {
+                                       return
+                               }
+                       }
+                       fv.Set(reflect.ValueOf(val))
+               default:
+                       panic("bad field type " + key + " " + fieldtype[key])
+               }
+       }
+       return
+}
+
+func ParseStruct(doc *Document) (iv interface{}, err error) {
+       switch doc.Type {
+       case "Draft":
+               iv = new(Draft)
+       case "Notice":
+               iv = new(Notice)
+       case "DebitCert":
+               iv = new(DebitCert)
+       case "CreditCert":
+               iv = new(CreditCert)
+       case "BounceCert":
+               iv = new(BounceCert)
+       default:
+               err = fmt.Errorf("ParseStruct: unkown doc type: %s", doc.Type)
+               return
+       }
+       seen := make(map[string]bool)
+       err = parseStruct(reflect.ValueOf(iv).Elem(), doc.Fields, seen)
+       if err != nil {
+               return
+       }
+       if len(doc.Fields) != len(seen) {
+               for f := range doc.Fields {
+                       if !seen[f] {
+                               err = fmt.Errorf("ParseStruct: unknown field %s in %s", f, doc.Type)
+                               return
+                       }
+               }
+       }
+       return
+}
+
+// turn a struct into a document
+func formatStruct(v reflect.Value, doc *Document) (err error) {
+       t := v.Type()
+       n := v.NumField()
+       for i := 0; i < n; i++ {
+               ft := t.Field(i)
+               fv := v.Field(i)
+               if ft.Anonymous && fv.Kind() == reflect.Struct {
+                       err = formatStruct(fv, doc)
+                       if err != nil {
+                               return
+                       }
+                       continue
+               }
+               key := fieldname[ft.Name]
+               if key == "" {
+                       key = ft.Name
+               }
+               val := ""
+               if fv.Kind() == reflect.Ptr {
+                       if fv.IsNil() {
+                               // keep empty optional fields but mark them
+                               val = "-"
+                               goto setval
+                       }
+                       fv = fv.Elem()
+               }
+               switch fieldtype[key] {
+               case "id":
+                       val = formatId(fv.String())
+               case "text":
+                       val = formatString(fv.String())
+               case "int":
+                       val = strconv.Itoa64(fv.Int())
+               case "date":
+                       val = formatDate(fv.Int())
+               case "ids":
+                       k := fv.Len()
+                       for j := 0; j < k; j++ {
+                               if j > 0 {
+                                       val += "\n "
+                               }
+                               val += formatId(fv.Index(j).String())
+                       }
+               default:
+                       panic("bad field type " + key + " " + fieldtype[key])
+               }
+       setval:
+               doc.Fields[key] = val
+               doc.Order = append(doc.Order, key)
+       }
+       return
+}
+
+// turn a struct into a document
+func FormatStruct(iv interface{}) (doc *Document, err error) {
+       v := reflect.ValueOf(iv)
+       if v.Kind() != reflect.Ptr || v.IsNil() || v.Elem().Kind() != reflect.Struct {
+               panic("input is not a pointer to struct")
+       }
+       doc = new(Document)
+       doc.Type = v.Elem().Type().Name()
+       doc.Fields = make(map[string]string)
+       err = formatStruct(v.Elem(), doc)
+       return
+}
+
+func ParseFields(s []byte) (fields map[string]string, rest []byte, err error) {
+       rest = s
+       fields = make(map[string]string)
+       key := ""
+       // \n is optional after the last field and an extra \n is allowed as well
+       for len(rest) > 0 {
+               var line []byte
+               line, rest = getLine(rest)
+               // empty line after the last field is consumed
+               if len(line) == 0 {
+                       break
+               }
+               if line[0] == ' ' && key != "" {
+                       // "Key: v1\n v2\n" is equivalent to "Key: v1 v2\n"
+                       fields[key] += string(line)
+                       continue
+               }
+               if line[0] < 'A' || line[0] > 'Z' {
+                       err = fmt.Errorf("ParseFields: field name must start with an upper-case ascii letter")
+                       return
+               }
+               i := bytes.IndexByte(line, ':')
+               if i < 0 {
+                       err = fmt.Errorf("ParseFields: missing ':'")
+                       return
+               }
+               key = string(line[:i])
+               if _, ok := fields[key]; ok {
+                       err = fmt.Errorf("ParseFields: repeated fields are not allowed")
+                       return
+               }
+               fields[key] = string(line[i+1:])
+       }
+       for key, v := range fields {
+               // either a single space follows ':' or the value is empty
+               // good: "Key:\n", "Key:\n value\n", "Key: value\n", "Key: v1\n v2\n"
+               // bad: "Key:value\n", "Key: \nvalue\n"
+               // bad but not checked here: "Key: \n", "Key: value \n", "Key:\n \n value\n"
+               if len(v) == 0 {
+                       continue
+               }
+               if v[0] != ' ' {
+                       err = fmt.Errorf("ParseFields: ':' is not followed by ' '")
+                       return
+               }
+               fields[key] = v[1:]
+       }
+       return
+}
+
+// TODO: limit errors
+
+func parseId(s string) (string, error) {
+       // check if hex decodable
+       // TODO: length check
+       dst := make([]byte, len(s)/2)
+       _, err := hex.Decode(dst, []byte(s))
+       return s, err
+}
+
+func formatId(s string) string {
+       return s
+}
+
+func parseString(s string) (string, error) {
+       if len(s) > MaxValueLength {
+               return "", fmt.Errorf("parseString: length limit is exceeded")
+       }
+       return s, nil
+}
+
+func formatString(s string) string {
+       return s
+}
+
+func parseDate(s string) (int64, error) {
+       // TODO: fractional seconds?
+       t, err := time.Parse(time.RFC3339, s)
+       if err != nil {
+               return 0, err
+       }
+       return t.Seconds(), nil
+}
+
+func formatDate(i int64) string {
+       return time.SecondsToUTC(i).Format(time.RFC3339)
+}
+
+func getLine(data []byte) (line, rest []byte) {
+       i := bytes.IndexByte(data, '\n')
+       j := i + 1
+       if i < 0 {
+               i = len(data)
+               j = i
+       } else if i > 0 && data[i-1] == '\r' {
+               i--
+       }
+       return data[:i], data[j:]
+}