Update toml dependency

This commit is contained in:
Frank Denis 2021-08-08 10:21:25 +02:00
parent 1052fa6323
commit c9d5d81e6a
43 changed files with 1455 additions and 825 deletions

View file

@ -1,5 +1,2 @@
TAGS
tags
.*.swp
tomlcheck/tomlcheck
toml.test
/toml-test

View file

@ -1,15 +0,0 @@
language: go
go:
- 1.1
- 1.2
- 1.3
- 1.4
- 1.5
- 1.6
- tip
install:
- go install ./...
- go get github.com/BurntSushi/toml-test
script:
- export PATH="$PATH:$HOME/gopath/bin"
- make test

View file

@ -1,3 +1 @@
Compatible with TOML version
[v0.4.0](https://github.com/toml-lang/toml/blob/v0.4.0/versions/en/toml-v0.4.0.md)
Compatible with TOML version [v1.0.0](https://toml.io/en/v1.0.0).

View file

@ -1,19 +0,0 @@
install:
go install ./...
test: install
go test -v
toml-test toml-test-decoder
toml-test -encoder toml-test-encoder
fmt:
gofmt -w *.go */*.go
colcheck *.go */*.go
tags:
find ./ -name '*.go' -print0 | xargs -0 gotags > TAGS
push:
git push origin master
git push github master

View file

@ -6,27 +6,22 @@ packages. This package also supports the `encoding.TextUnmarshaler` and
`encoding.TextMarshaler` interfaces so that you can define custom data
representations. (There is an example of this below.)
Spec: https://github.com/toml-lang/toml
Compatible with TOML version [v1.0.0](https://toml.io/en/v1.0.0).
Compatible with TOML version
[v0.4.0](https://github.com/toml-lang/toml/blob/master/versions/en/toml-v0.4.0.md)
Documentation: https://godocs.io/github.com/BurntSushi/toml
Documentation: https://godoc.org/github.com/BurntSushi/toml
See the [releases page](https://github.com/BurntSushi/toml/releases) for a
changelog; this information is also in the git tag annotations (e.g. `git show
v0.4.0`).
Installation:
This library requires Go 1.13 or newer; install it with:
```bash
go get github.com/BurntSushi/toml
```
$ go get github.com/BurntSushi/toml
Try the toml validator:
It also comes with a TOML validator CLI tool:
```bash
go get github.com/BurntSushi/toml/cmd/tomlv
tomlv some-toml-file.toml
```
[![Build Status](https://travis-ci.org/BurntSushi/toml.svg?branch=master)](https://travis-ci.org/BurntSushi/toml) [![GoDoc](https://godoc.org/github.com/BurntSushi/toml?status.svg)](https://godoc.org/github.com/BurntSushi/toml)
$ go get github.com/BurntSushi/toml/cmd/tomlv
$ tomlv some-toml-file.toml
### Testing
@ -36,8 +31,8 @@ and the encoder.
### Examples
This package works similarly to how the Go standard library handles `XML`
and `JSON`. Namely, data is loaded into Go values via reflection.
This package works similarly to how the Go standard library handles XML and
JSON. Namely, data is loaded into Go values via reflection.
For the simplest example, consider some TOML file as just a list of keys
and values:
@ -54,11 +49,11 @@ Which could be defined in Go as:
```go
type Config struct {
Age int
Cats []string
Pi float64
Perfection []int
DOB time.Time // requires `import time`
Age int
Cats []string
Pi float64
Perfection []int
DOB time.Time // requires `import time`
}
```
@ -84,6 +79,9 @@ type TOML struct {
}
```
Beware that like other most other decoders **only exported fields** are
considered when encoding and decoding; private fields are silently ignored.
### Using the `encoding.TextUnmarshaler` interface
Here's an example that automatically parses duration strings into
@ -103,19 +101,19 @@ Which can be decoded with:
```go
type song struct {
Name string
Duration duration
Name string
Duration duration
}
type songs struct {
Song []song
Song []song
}
var favorites songs
if _, err := toml.Decode(blob, &favorites); err != nil {
log.Fatal(err)
log.Fatal(err)
}
for _, s := range favorites.Song {
fmt.Printf("%s (%s)\n", s.Name, s.Duration)
fmt.Printf("%s (%s)\n", s.Name, s.Duration)
}
```
@ -134,6 +132,9 @@ func (d *duration) UnmarshalText(text []byte) error {
}
```
To target TOML specifically you can implement `UnmarshalTOML` TOML interface in
a similar way.
### More complex usage
Here's an example of how to load the example from the official spec page:
@ -180,23 +181,23 @@ And the corresponding Go types are:
```go
type tomlConfig struct {
Title string
Owner ownerInfo
DB database `toml:"database"`
Title string
Owner ownerInfo
DB database `toml:"database"`
Servers map[string]server
Clients clients
}
type ownerInfo struct {
Name string
Org string `toml:"organization"`
Bio string
DOB time.Time
Org string `toml:"organization"`
Bio string
DOB time.Time
}
type database struct {
Server string
Ports []int
Server string
Ports []int
ConnMax int `toml:"connection_max"`
Enabled bool
}
@ -207,7 +208,7 @@ type server struct {
}
type clients struct {
Data [][]interface{}
Data [][]interface{}
Hosts []string
}
```
@ -216,3 +217,4 @@ Note that a case insensitive match will be tried if an exact match can't be
found.
A working example of the above can be found in `_examples/example.{go,toml}`.

View file

@ -1,19 +1,17 @@
package toml
import (
"encoding"
"fmt"
"io"
"io/ioutil"
"math"
"os"
"reflect"
"strings"
"time"
)
func e(format string, args ...interface{}) error {
return fmt.Errorf("toml: "+format, args...)
}
// Unmarshaler is the interface implemented by objects that can unmarshal a
// TOML description of themselves.
type Unmarshaler interface {
@ -27,30 +25,21 @@ func Unmarshal(p []byte, v interface{}) error {
}
// Primitive is a TOML value that hasn't been decoded into a Go value.
// When using the various `Decode*` functions, the type `Primitive` may
// be given to any value, and its decoding will be delayed.
//
// A `Primitive` value can be decoded using the `PrimitiveDecode` function.
// This type can be used for any value, which will cause decoding to be delayed.
// You can use the PrimitiveDecode() function to "manually" decode these values.
//
// The underlying representation of a `Primitive` value is subject to change.
// Do not rely on it.
// NOTE: The underlying representation of a `Primitive` value is subject to
// change. Do not rely on it.
//
// N.B. Primitive values are still parsed, so using them will only avoid
// the overhead of reflection. They can be useful when you don't know the
// exact type of TOML data until run time.
// NOTE: Primitive values are still parsed, so using them will only avoid the
// overhead of reflection. They can be useful when you don't know the exact type
// of TOML data until runtime.
type Primitive struct {
undecoded interface{}
context Key
}
// DEPRECATED!
//
// Use MetaData.PrimitiveDecode instead.
func PrimitiveDecode(primValue Primitive, v interface{}) error {
md := MetaData{decoded: make(map[string]bool)}
return md.unify(primValue.undecoded, rvalue(v))
}
// PrimitiveDecode is just like the other `Decode*` functions, except it
// decodes a TOML value that has already been parsed. Valid primitive values
// can *only* be obtained from values filled by the decoder functions,
@ -68,43 +57,51 @@ func (md *MetaData) PrimitiveDecode(primValue Primitive, v interface{}) error {
return md.unify(primValue.undecoded, rvalue(v))
}
// Decode will decode the contents of `data` in TOML format into a pointer
// `v`.
// Decoder decodes TOML data.
//
// TOML hashes correspond to Go structs or maps. (Dealer's choice. They can be
// used interchangeably.)
// TOML tables correspond to Go structs or maps (dealer's choice – they can be
// used interchangeably).
//
// TOML arrays of tables correspond to either a slice of structs or a slice
// of maps.
// TOML table arrays correspond to either a slice of structs or a slice of maps.
//
// TOML datetimes correspond to Go `time.Time` values.
// TOML datetimes correspond to Go time.Time values. Local datetimes are parsed
// in the local timezone.
//
// All other TOML types (float, string, int, bool and array) correspond
// to the obvious Go types.
// All other TOML types (float, string, int, bool and array) correspond to the
// obvious Go types.
//
// An exception to the above rules is if a type implements the
// encoding.TextUnmarshaler interface. In this case, any primitive TOML value
// (floats, strings, integers, booleans and datetimes) will be converted to
// a byte string and given to the value's UnmarshalText method. See the
// Unmarshaler example for a demonstration with time duration strings.
// An exception to the above rules is if a type implements the TextUnmarshaler
// interface, in which case any primitive TOML value (floats, strings, integers,
// booleans, datetimes) will be converted to a []byte and given to the value's
// UnmarshalText method. See the Unmarshaler example for a demonstration with
// time duration strings.
//
// Key mapping
//
// TOML keys can map to either keys in a Go map or field names in a Go
// struct. The special `toml` struct tag may be used to map TOML keys to
// struct fields that don't match the key name exactly. (See the example.)
// A case insensitive match to struct names will be tried if an exact match
// can't be found.
// TOML keys can map to either keys in a Go map or field names in a Go struct.
// The special `toml` struct tag can be used to map TOML keys to struct fields
// that don't match the key name exactly (see the example). A case insensitive
// match to struct names will be tried if an exact match can't be found.
//
// The mapping between TOML values and Go values is loose. That is, there
// may exist TOML values that cannot be placed into your representation, and
// there may be parts of your representation that do not correspond to
// TOML values. This loose mapping can be made stricter by using the IsDefined
// and/or Undecoded methods on the MetaData returned.
// The mapping between TOML values and Go values is loose. That is, there may
// exist TOML values that cannot be placed into your representation, and there
// may be parts of your representation that do not correspond to TOML values.
// This loose mapping can be made stricter by using the IsDefined and/or
// Undecoded methods on the MetaData returned.
//
// This decoder will not handle cyclic types. If a cyclic type is passed,
// `Decode` will not terminate.
func Decode(data string, v interface{}) (MetaData, error) {
// This decoder does not handle cyclic types. Decode will not terminate if a
// cyclic type is passed.
type Decoder struct {
r io.Reader
}
// NewDecoder creates a new Decoder.
func NewDecoder(r io.Reader) *Decoder {
return &Decoder{r: r}
}
// Decode TOML data in to the pointer `v`.
func (dec *Decoder) Decode(v interface{}) (MetaData, error) {
rv := reflect.ValueOf(v)
if rv.Kind() != reflect.Ptr {
return MetaData{}, e("Decode of non-pointer %s", reflect.TypeOf(v))
@ -112,7 +109,15 @@ func Decode(data string, v interface{}) (MetaData, error) {
if rv.IsNil() {
return MetaData{}, e("Decode of nil %s", reflect.TypeOf(v))
}
p, err := parse(data)
// TODO: have parser should read from io.Reader? Or at the very least, make
// it read from []byte rather than string
data, err := ioutil.ReadAll(dec.r)
if err != nil {
return MetaData{}, err
}
p, err := parse(string(data))
if err != nil {
return MetaData{}, err
}
@ -123,24 +128,22 @@ func Decode(data string, v interface{}) (MetaData, error) {
return md, md.unify(p.mapping, indirect(rv))
}
// DecodeFile is just like Decode, except it will automatically read the
// contents of the file at `fpath` and decode it for you.
func DecodeFile(fpath string, v interface{}) (MetaData, error) {
bs, err := ioutil.ReadFile(fpath)
if err != nil {
return MetaData{}, err
}
return Decode(string(bs), v)
// Decode the TOML data in to the pointer v.
//
// See the documentation on Decoder for a description of the decoding process.
func Decode(data string, v interface{}) (MetaData, error) {
return NewDecoder(strings.NewReader(data)).Decode(v)
}
// DecodeReader is just like Decode, except it will consume all bytes
// from the reader and decode it for you.
func DecodeReader(r io.Reader, v interface{}) (MetaData, error) {
bs, err := ioutil.ReadAll(r)
// DecodeFile is just like Decode, except it will automatically read the
// contents of the file at path and decode it for you.
func DecodeFile(path string, v interface{}) (MetaData, error) {
fp, err := os.Open(path)
if err != nil {
return MetaData{}, err
}
return Decode(string(bs), v)
defer fp.Close()
return NewDecoder(fp).Decode(v)
}
// unify performs a sort of type unification based on the structure of `rv`,
@ -149,8 +152,8 @@ func DecodeReader(r io.Reader, v interface{}) (MetaData, error) {
// Any type mismatch produces an error. Finding a type that we don't know
// how to handle produces an unsupported type error.
func (md *MetaData) unify(data interface{}, rv reflect.Value) error {
// Special case. Look for a `Primitive` value.
// TODO: #76 would make this superfluous after implemented.
if rv.Type() == reflect.TypeOf((*Primitive)(nil)).Elem() {
// Save the undecoded data and the key context into the primitive
// value.
@ -170,25 +173,17 @@ func (md *MetaData) unify(data interface{}, rv reflect.Value) error {
}
}
// Special case. Handle time.Time values specifically.
// TODO: Remove this code when we decide to drop support for Go 1.1.
// This isn't necessary in Go 1.2 because time.Time satisfies the encoding
// interfaces.
if rv.Type().AssignableTo(rvalue(time.Time{}).Type()) {
return md.unifyDatetime(data, rv)
}
// Special case. Look for a value satisfying the TextUnmarshaler interface.
if v, ok := rv.Interface().(TextUnmarshaler); ok {
if v, ok := rv.Interface().(encoding.TextUnmarshaler); ok {
return md.unifyText(data, v)
}
// BUG(burntsushi)
// TODO:
// The behavior here is incorrect whenever a Go type satisfies the
// encoding.TextUnmarshaler interface but also corresponds to a TOML
// hash or array. In particular, the unmarshaler should only be applied
// to primitive TOML values. But at this point, it will be applied to
// all kinds of values and produce an incorrect error whenever those values
// are hashes or arrays (including arrays of tables).
// encoding.TextUnmarshaler interface but also corresponds to a TOML hash or
// array. In particular, the unmarshaler should only be applied to primitive
// TOML values. But at this point, it will be applied to all kinds of values
// and produce an incorrect error whenever those values are hashes or arrays
// (including arrays of tables).
k := rv.Kind()
@ -277,6 +272,12 @@ func (md *MetaData) unifyStruct(mapping interface{}, rv reflect.Value) error {
}
func (md *MetaData) unifyMap(mapping interface{}, rv reflect.Value) error {
if k := rv.Type().Key().Kind(); k != reflect.String {
return fmt.Errorf(
"toml: cannot decode to a map with non-string key type (%s in %q)",
k, rv.Type())
}
tmap, ok := mapping.(map[string]interface{})
if !ok {
if tmap == nil {
@ -312,10 +313,8 @@ func (md *MetaData) unifyArray(data interface{}, rv reflect.Value) error {
}
return badtype("slice", data)
}
sliceLen := datav.Len()
if sliceLen != rv.Len() {
return e("expected array length %d; got TOML array of length %d",
rv.Len(), sliceLen)
if l := datav.Len(); l != rv.Len() {
return e("expected array length %d; got TOML array of length %d", rv.Len(), l)
}
return md.unifySliceArray(datav, rv)
}
@ -337,11 +336,10 @@ func (md *MetaData) unifySlice(data interface{}, rv reflect.Value) error {
}
func (md *MetaData) unifySliceArray(data, rv reflect.Value) error {
sliceLen := data.Len()
for i := 0; i < sliceLen; i++ {
v := data.Index(i).Interface()
sliceval := indirect(rv.Index(i))
if err := md.unify(v, sliceval); err != nil {
l := data.Len()
for i := 0; i < l; i++ {
err := md.unify(data.Index(i).Interface(), indirect(rv.Index(i)))
if err != nil {
return err
}
}
@ -439,7 +437,7 @@ func (md *MetaData) unifyAnything(data interface{}, rv reflect.Value) error {
return nil
}
func (md *MetaData) unifyText(data interface{}, v TextUnmarshaler) error {
func (md *MetaData) unifyText(data interface{}, v encoding.TextUnmarshaler) error {
var s string
switch sdata := data.(type) {
case TextMarshaler:
@ -482,7 +480,7 @@ func indirect(v reflect.Value) reflect.Value {
if v.Kind() != reflect.Ptr {
if v.CanSet() {
pv := v.Addr()
if _, ok := pv.Interface().(TextUnmarshaler); ok {
if _, ok := pv.Interface().(encoding.TextUnmarshaler); ok {
return pv
}
}
@ -498,12 +496,16 @@ func isUnifiable(rv reflect.Value) bool {
if rv.CanSet() {
return true
}
if _, ok := rv.Interface().(TextUnmarshaler); ok {
if _, ok := rv.Interface().(encoding.TextUnmarshaler); ok {
return true
}
return false
}
func e(format string, args ...interface{}) error {
return fmt.Errorf("toml: "+format, args...)
}
func badtype(expected string, data interface{}) error {
return e("cannot load TOML value of type %T into a Go %s", data, expected)
}

18
vendor/github.com/BurntSushi/toml/decode_go116.go generated vendored Normal file
View file

@ -0,0 +1,18 @@
// +build go1.16
package toml
import (
"io/fs"
)
// DecodeFS is just like Decode, except it will automatically read the contents
// of the file at `path` from a fs.FS instance.
func DecodeFS(fsys fs.FS, path string, v interface{}) (MetaData, error) {
fp, err := fsys.Open(path)
if err != nil {
return MetaData{}, err
}
defer fp.Close()
return NewDecoder(fp).Decode(v)
}

View file

@ -2,9 +2,9 @@ package toml
import "strings"
// MetaData allows access to meta information about TOML data that may not
// be inferrable via reflection. In particular, whether a key has been defined
// and the TOML type of a key.
// MetaData allows access to meta information about TOML data that may not be
// inferable via reflection. In particular, whether a key has been defined and
// the TOML type of a key.
type MetaData struct {
mapping map[string]interface{}
types map[string]tomlType
@ -13,10 +13,11 @@ type MetaData struct {
context Key // Used only during decoding.
}
// IsDefined returns true if the key given exists in the TOML data. The key
// should be specified hierarchially. e.g.,
// IsDefined reports if the key exists in the TOML data.
//
// The key should be specified hierarchically, for example to access the TOML
// key "a.b.c" you would use:
//
// // access the TOML key 'a.b.c'
// IsDefined("a", "b", "c")
//
// IsDefined will return false if an empty key given. Keys are case sensitive.
@ -41,8 +42,8 @@ func (md *MetaData) IsDefined(key ...string) bool {
// Type returns a string representation of the type of the key specified.
//
// Type will return the empty string if given an empty key or a key that
// does not exist. Keys are case sensitive.
// Type will return the empty string if given an empty key or a key that does
// not exist. Keys are case sensitive.
func (md *MetaData) Type(key ...string) string {
fullkey := strings.Join(key, ".")
if typ, ok := md.types[fullkey]; ok {
@ -51,13 +52,11 @@ func (md *MetaData) Type(key ...string) string {
return ""
}
// Key is the type of any TOML key, including key groups. Use (MetaData).Keys
// to get values of this type.
// Key represents any TOML key, including key groups. Use (MetaData).Keys to get
// values of this type.
type Key []string
func (k Key) String() string {
return strings.Join(k, ".")
}
func (k Key) String() string { return strings.Join(k, ".") }
func (k Key) maybeQuotedAll() string {
var ss []string
@ -68,6 +67,9 @@ func (k Key) maybeQuotedAll() string {
}
func (k Key) maybeQuoted(i int) string {
if k[i] == "" {
return `""`
}
quote := false
for _, c := range k[i] {
if !isBareKeyChar(c) {
@ -76,7 +78,7 @@ func (k Key) maybeQuoted(i int) string {
}
}
if quote {
return "\"" + strings.Replace(k[i], "\"", "\\\"", -1) + "\""
return `"` + quotedReplacer.Replace(k[i]) + `"`
}
return k[i]
}
@ -89,10 +91,10 @@ func (k Key) add(piece string) Key {
}
// Keys returns a slice of every key in the TOML data, including key groups.
// Each key is itself a slice, where the first element is the top of the
// hierarchy and the last is the most specific.
//
// The list will have the same order as the keys appeared in the TOML data.
// Each key is itself a slice, where the first element is the top of the
// hierarchy and the last is the most specific. The list will have the same
// order as the keys appeared in the TOML data.
//
// All keys returned are non-empty.
func (md *MetaData) Keys() []Key {

33
vendor/github.com/BurntSushi/toml/deprecated.go generated vendored Normal file
View file

@ -0,0 +1,33 @@
package toml
import (
"encoding"
"io"
)
// DEPRECATED!
//
// Use the identical encoding.TextMarshaler instead. It is defined here to
// support Go 1.1 and older.
type TextMarshaler encoding.TextMarshaler
// DEPRECATED!
//
// Use the identical encoding.TextUnmarshaler instead. It is defined here to
// support Go 1.1 and older.
type TextUnmarshaler encoding.TextUnmarshaler
// DEPRECATED!
//
// Use MetaData.PrimitiveDecode instead.
func PrimitiveDecode(primValue Primitive, v interface{}) error {
md := MetaData{decoded: make(map[string]bool)}
return md.unify(primValue.undecoded, rvalue(v))
}
// DEPRECATED!
//
// Use NewDecoder(reader).Decode(&v) instead.
func DecodeReader(r io.Reader, v interface{}) (MetaData, error) {
return NewDecoder(r).Decode(v)
}

View file

@ -1,27 +1,13 @@
/*
Package toml provides facilities for decoding and encoding TOML configuration
files via reflection. There is also support for delaying decoding with
the Primitive type, and querying the set of keys in a TOML document with the
MetaData type.
Package toml implements decoding and encoding of TOML files.
The specification implemented: https://github.com/toml-lang/toml
This package supports TOML v1.0.0, as listed on https://toml.io
The sub-command github.com/BurntSushi/toml/cmd/tomlv can be used to verify
whether a file is a valid TOML document. It can also be used to print the
type of each key in a TOML document.
There is also support for delaying decoding with the Primitive type, and
querying the set of keys in a TOML document with the MetaData type.
Testing
There are two important types of tests used for this package. The first is
contained inside '*_test.go' files and uses the standard Go unit testing
framework. These tests are primarily devoted to holistically testing the
decoder and encoder.
The second type of testing is used to verify the implementation's adherence
to the TOML specification. These tests have been factored into their own
project: https://github.com/BurntSushi/toml-test
The reason the tests are in a separate project is so that they can be used by
any implementation of TOML. Namely, it is language agnostic.
The github.com/BurntSushi/toml/cmd/tomlv package implements a TOML validator,
and can be used to verify if TOML document is valid. It can also be used to
print the type of each key.
*/
package toml

View file

@ -2,48 +2,92 @@ package toml
import (
"bufio"
"encoding"
"errors"
"fmt"
"io"
"math"
"reflect"
"sort"
"strconv"
"strings"
"time"
"github.com/BurntSushi/toml/internal"
)
type tomlEncodeError struct{ error }
var (
errArrayMixedElementTypes = errors.New(
"toml: cannot encode array with mixed element types")
errArrayNilElement = errors.New(
"toml: cannot encode array with nil element")
errNonString = errors.New(
"toml: cannot encode a map with non-string key type")
errAnonNonStruct = errors.New(
"toml: cannot encode an anonymous field that is not a struct")
errArrayNoTable = errors.New(
"toml: TOML array element cannot contain a table")
errNoKey = errors.New(
"toml: top-level values must be Go maps or structs")
errAnything = errors.New("") // used in testing
errArrayNilElement = errors.New("toml: cannot encode array with nil element")
errNonString = errors.New("toml: cannot encode a map with non-string key type")
errAnonNonStruct = errors.New("toml: cannot encode an anonymous field that is not a struct")
errNoKey = errors.New("toml: top-level values must be Go maps or structs")
errAnything = errors.New("") // used in testing
)
var quotedReplacer = strings.NewReplacer(
"\t", "\\t",
"\n", "\\n",
"\r", "\\r",
"\"", "\\\"",
"\\", "\\\\",
"\x00", `\u0000`,
"\x01", `\u0001`,
"\x02", `\u0002`,
"\x03", `\u0003`,
"\x04", `\u0004`,
"\x05", `\u0005`,
"\x06", `\u0006`,
"\x07", `\u0007`,
"\b", `\b`,
"\t", `\t`,
"\n", `\n`,
"\x0b", `\u000b`,
"\f", `\f`,
"\r", `\r`,
"\x0e", `\u000e`,
"\x0f", `\u000f`,
"\x10", `\u0010`,
"\x11", `\u0011`,
"\x12", `\u0012`,
"\x13", `\u0013`,
"\x14", `\u0014`,
"\x15", `\u0015`,
"\x16", `\u0016`,
"\x17", `\u0017`,
"\x18", `\u0018`,
"\x19", `\u0019`,
"\x1a", `\u001a`,
"\x1b", `\u001b`,
"\x1c", `\u001c`,
"\x1d", `\u001d`,
"\x1e", `\u001e`,
"\x1f", `\u001f`,
"\x7f", `\u007f`,
)
// Encoder controls the encoding of Go values to a TOML document to some
// io.Writer.
// Encoder encodes a Go to a TOML document.
//
// The indentation level can be controlled with the Indent field.
// The mapping between Go values and TOML values should be precisely the same as
// for the Decode* functions. Similarly, the TextMarshaler interface is
// supported by encoding the resulting bytes as strings. If you want to write
// arbitrary binary data then you will need to use something like base64 since
// TOML does not have any binary types.
//
// When encoding TOML hashes (Go maps or structs), keys without any sub-hashes
// are encoded first.
//
// Go maps will be sorted alphabetically by key for deterministic output.
//
// Encoding Go values without a corresponding TOML representation will return an
// error. Examples of this includes maps with non-string keys, slices with nil
// elements, embedded non-struct types, and nested slices containing maps or
// structs. (e.g. [][]map[string]string is not allowed but []map[string]string
// is okay, as is []map[string][]string).
//
// NOTE: Only exported keys are encoded due to the use of reflection. Unexported
// keys are silently discarded.
type Encoder struct {
// A single indentation level. By default it is two spaces.
// The string to use for a single indentation level. The default is two
// spaces.
Indent string
// hasWritten is whether we have written any output to w yet.
@ -51,8 +95,7 @@ type Encoder struct {
w *bufio.Writer
}
// NewEncoder returns a TOML encoder that encodes Go values to the io.Writer
// given. By default, a single indentation level is 2 spaces.
// NewEncoder create a new Encoder.
func NewEncoder(w io.Writer) *Encoder {
return &Encoder{
w: bufio.NewWriter(w),
@ -60,29 +103,10 @@ func NewEncoder(w io.Writer) *Encoder {
}
}
// Encode writes a TOML representation of the Go value to the underlying
// io.Writer. If the value given cannot be encoded to a valid TOML document,
// then an error is returned.
// Encode writes a TOML representation of the Go value to the Encoder's writer.
//
// The mapping between Go values and TOML values should be precisely the same
// as for the Decode* functions. Similarly, the TextMarshaler interface is
// supported by encoding the resulting bytes as strings. (If you want to write
// arbitrary binary data then you will need to use something like base64 since
// TOML does not have any binary types.)
//
// When encoding TOML hashes (i.e., Go maps or structs), keys without any
// sub-hashes are encoded first.
//
// If a Go map is encoded, then its keys are sorted alphabetically for
// deterministic output. More control over this behavior may be provided if
// there is demand for it.
//
// Encoding Go values without a corresponding TOML representation---like map
// types with non-string keys---will cause an error to be returned. Similarly
// for mixed arrays/slices, arrays/slices with nil elements, embedded
// non-struct types and nested slices containing maps or structs.
// (e.g., [][]map[string]string is not allowed but []map[string]string is OK
// and so is []map[string][]string.)
// An error is returned if the value given cannot be encoded to a valid TOML
// document.
func (enc *Encoder) Encode(v interface{}) error {
rv := eindirect(reflect.ValueOf(v))
if err := enc.safeEncode(Key([]string{}), rv); err != nil {
@ -110,9 +134,13 @@ func (enc *Encoder) encode(key Key, rv reflect.Value) {
// Special case. If we can marshal the type to text, then we used that.
// Basically, this prevents the encoder for handling these types as
// generic structs (or whatever the underlying type of a TextMarshaler is).
switch rv.Interface().(type) {
case time.Time, TextMarshaler:
enc.keyEqElement(key, rv)
switch t := rv.Interface().(type) {
case time.Time, encoding.TextMarshaler:
enc.writeKeyValue(key, rv, false)
return
// TODO: #76 would make this superfluous after implemented.
case Primitive:
enc.encode(key, reflect.ValueOf(t.undecoded))
return
}
@ -123,12 +151,12 @@ func (enc *Encoder) encode(key Key, rv reflect.Value) {
reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32,
reflect.Uint64,
reflect.Float32, reflect.Float64, reflect.String, reflect.Bool:
enc.keyEqElement(key, rv)
enc.writeKeyValue(key, rv, false)
case reflect.Array, reflect.Slice:
if typeEqual(tomlArrayHash, tomlTypeOfGo(rv)) {
enc.eArrayOfTables(key, rv)
} else {
enc.keyEqElement(key, rv)
enc.writeKeyValue(key, rv, false)
}
case reflect.Interface:
if rv.IsNil() {
@ -148,22 +176,32 @@ func (enc *Encoder) encode(key Key, rv reflect.Value) {
case reflect.Struct:
enc.eTable(key, rv)
default:
panic(e("unsupported type for key '%s': %s", key, k))
encPanic(fmt.Errorf("unsupported type for key '%s': %s", key, k))
}
}
// eElement encodes any value that can be an array element (primitives and
// arrays).
// eElement encodes any value that can be an array element.
func (enc *Encoder) eElement(rv reflect.Value) {
switch v := rv.Interface().(type) {
case time.Time:
// Special case time.Time as a primitive. Has to come before
// TextMarshaler below because time.Time implements
// encoding.TextMarshaler, but we need to always use UTC.
enc.wf(v.UTC().Format("2006-01-02T15:04:05Z"))
case time.Time: // Using TextMarshaler adds extra quotes, which we don't want.
format := time.RFC3339Nano
switch v.Location() {
case internal.LocalDatetime:
format = "2006-01-02T15:04:05.999999999"
case internal.LocalDate:
format = "2006-01-02"
case internal.LocalTime:
format = "15:04:05.999999999"
}
switch v.Location() {
default:
enc.wf(v.Format(format))
case internal.LocalDatetime, internal.LocalDate, internal.LocalTime:
enc.wf(v.In(time.UTC).Format(format))
}
return
case TextMarshaler:
// Special case. Use text marshaler if it's available for this value.
case encoding.TextMarshaler:
// Use text marshaler if it's available for this value.
if s, err := v.MarshalText(); err != nil {
encPanic(err)
} else {
@ -171,32 +209,49 @@ func (enc *Encoder) eElement(rv reflect.Value) {
}
return
}
switch rv.Kind() {
case reflect.Bool:
enc.wf(strconv.FormatBool(rv.Bool()))
case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32,
reflect.Int64:
enc.wf(strconv.FormatInt(rv.Int(), 10))
case reflect.Uint, reflect.Uint8, reflect.Uint16,
reflect.Uint32, reflect.Uint64:
enc.wf(strconv.FormatUint(rv.Uint(), 10))
case reflect.Float32:
enc.wf(floatAddDecimal(strconv.FormatFloat(rv.Float(), 'f', -1, 32)))
case reflect.Float64:
enc.wf(floatAddDecimal(strconv.FormatFloat(rv.Float(), 'f', -1, 64)))
case reflect.Array, reflect.Slice:
enc.eArrayOrSliceElement(rv)
case reflect.Interface:
enc.eElement(rv.Elem())
case reflect.String:
enc.writeQuoted(rv.String())
case reflect.Bool:
enc.wf(strconv.FormatBool(rv.Bool()))
case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
enc.wf(strconv.FormatInt(rv.Int(), 10))
case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64:
enc.wf(strconv.FormatUint(rv.Uint(), 10))
case reflect.Float32:
f := rv.Float()
if math.IsNaN(f) {
enc.wf("nan")
} else if math.IsInf(f, 0) {
enc.wf("%cinf", map[bool]byte{true: '-', false: '+'}[math.Signbit(f)])
} else {
enc.wf(floatAddDecimal(strconv.FormatFloat(f, 'f', -1, 32)))
}
case reflect.Float64:
f := rv.Float()
if math.IsNaN(f) {
enc.wf("nan")
} else if math.IsInf(f, 0) {
enc.wf("%cinf", map[bool]byte{true: '-', false: '+'}[math.Signbit(f)])
} else {
enc.wf(floatAddDecimal(strconv.FormatFloat(f, 'f', -1, 64)))
}
case reflect.Array, reflect.Slice:
enc.eArrayOrSliceElement(rv)
case reflect.Struct:
enc.eStruct(nil, rv, true)
case reflect.Map:
enc.eMap(nil, rv, true)
case reflect.Interface:
enc.eElement(rv.Elem())
default:
panic(e("unexpected primitive type: %s", rv.Kind()))
encPanic(fmt.Errorf("unexpected primitive type: %T", rv.Interface()))
}
}
// By the TOML spec, all floats must have a decimal with at least one
// number on either side.
// By the TOML spec, all floats must have a decimal with at least one number on
// either side.
func floatAddDecimal(fstr string) string {
if !strings.Contains(fstr, ".") {
return fstr + ".0"
@ -230,16 +285,14 @@ func (enc *Encoder) eArrayOfTables(key Key, rv reflect.Value) {
if isNil(trv) {
continue
}
panicIfInvalidKey(key)
enc.newline()
enc.wf("%s[[%s]]", enc.indentStr(key), key.maybeQuotedAll())
enc.newline()
enc.eMapOrStruct(key, trv)
enc.eMapOrStruct(key, trv, false)
}
}
func (enc *Encoder) eTable(key Key, rv reflect.Value) {
panicIfInvalidKey(key)
if len(key) == 1 {
// Output an extra newline between top-level tables.
// (The newline isn't written if nothing else has been written though.)
@ -249,21 +302,22 @@ func (enc *Encoder) eTable(key Key, rv reflect.Value) {
enc.wf("%s[%s]", enc.indentStr(key), key.maybeQuotedAll())
enc.newline()
}
enc.eMapOrStruct(key, rv)
enc.eMapOrStruct(key, rv, false)
}
func (enc *Encoder) eMapOrStruct(key Key, rv reflect.Value) {
func (enc *Encoder) eMapOrStruct(key Key, rv reflect.Value, inline bool) {
switch rv := eindirect(rv); rv.Kind() {
case reflect.Map:
enc.eMap(key, rv)
enc.eMap(key, rv, inline)
case reflect.Struct:
enc.eStruct(key, rv)
enc.eStruct(key, rv, inline)
default:
// Should never happen?
panic("eTable: unhandled reflect.Value Kind: " + rv.Kind().String())
}
}
func (enc *Encoder) eMap(key Key, rv reflect.Value) {
func (enc *Encoder) eMap(key Key, rv reflect.Value, inline bool) {
rt := rv.Type()
if rt.Key().Kind() != reflect.String {
encPanic(errNonString)
@ -281,57 +335,76 @@ func (enc *Encoder) eMap(key Key, rv reflect.Value) {
}
}
var writeMapKeys = func(mapKeys []string) {
var writeMapKeys = func(mapKeys []string, trailC bool) {
sort.Strings(mapKeys)
for _, mapKey := range mapKeys {
mrv := rv.MapIndex(reflect.ValueOf(mapKey))
if isNil(mrv) {
// Don't write anything for nil fields.
for i, mapKey := range mapKeys {
val := rv.MapIndex(reflect.ValueOf(mapKey))
if isNil(val) {
continue
}
enc.encode(key.add(mapKey), mrv)
if inline {
enc.writeKeyValue(Key{mapKey}, val, true)
if trailC || i != len(mapKeys)-1 {
enc.wf(", ")
}
} else {
enc.encode(key.add(mapKey), val)
}
}
}
writeMapKeys(mapKeysDirect)
writeMapKeys(mapKeysSub)
if inline {
enc.wf("{")
}
writeMapKeys(mapKeysDirect, len(mapKeysSub) > 0)
writeMapKeys(mapKeysSub, false)
if inline {
enc.wf("}")
}
}
func (enc *Encoder) eStruct(key Key, rv reflect.Value) {
func (enc *Encoder) eStruct(key Key, rv reflect.Value, inline bool) {
// Write keys for fields directly under this key first, because if we write
// a field that creates a new table, then all keys under it will be in that
// a field that creates a new table then all keys under it will be in that
// table (not the one we're writing here).
rt := rv.Type()
var fieldsDirect, fieldsSub [][]int
var addFields func(rt reflect.Type, rv reflect.Value, start []int)
//
// Fields is a [][]int: for fieldsDirect this always has one entry (the
// struct index). For fieldsSub it contains two entries: the parent field
// index from tv, and the field indexes for the fields of the sub.
var (
rt = rv.Type()
fieldsDirect, fieldsSub [][]int
addFields func(rt reflect.Type, rv reflect.Value, start []int)
)
addFields = func(rt reflect.Type, rv reflect.Value, start []int) {
for i := 0; i < rt.NumField(); i++ {
f := rt.Field(i)
// skip unexported fields
if f.PkgPath != "" && !f.Anonymous {
if f.PkgPath != "" && !f.Anonymous { /// Skip unexported fields.
continue
}
frv := rv.Field(i)
// Treat anonymous struct fields with tag names as though they are
// not anonymous, like encoding/json does.
//
// Non-struct anonymous fields use the normal encoding logic.
if f.Anonymous {
t := f.Type
switch t.Kind() {
case reflect.Struct:
// Treat anonymous struct fields with
// tag names as though they are not
// anonymous, like encoding/json does.
if getOptions(f.Tag).name == "" {
addFields(t, frv, f.Index)
addFields(t, frv, append(start, f.Index...))
continue
}
case reflect.Ptr:
if t.Elem().Kind() == reflect.Struct &&
getOptions(f.Tag).name == "" {
if t.Elem().Kind() == reflect.Struct && getOptions(f.Tag).name == "" {
if !frv.IsNil() {
addFields(t.Elem(), frv.Elem(), f.Index)
addFields(t.Elem(), frv.Elem(), append(start, f.Index...))
}
continue
}
// Fall through to the normal field encoding logic below
// for non-struct anonymous fields.
}
}
@ -344,35 +417,49 @@ func (enc *Encoder) eStruct(key Key, rv reflect.Value) {
}
addFields(rt, rv, nil)
var writeFields = func(fields [][]int) {
writeFields := func(fields [][]int) {
for _, fieldIndex := range fields {
sft := rt.FieldByIndex(fieldIndex)
sf := rv.FieldByIndex(fieldIndex)
if isNil(sf) {
// Don't write anything for nil fields.
fieldType := rt.FieldByIndex(fieldIndex)
fieldVal := rv.FieldByIndex(fieldIndex)
if isNil(fieldVal) { /// Don't write anything for nil fields.
continue
}
opts := getOptions(sft.Tag)
opts := getOptions(fieldType.Tag)
if opts.skip {
continue
}
keyName := sft.Name
keyName := fieldType.Name
if opts.name != "" {
keyName = opts.name
}
if opts.omitempty && isEmpty(sf) {
if opts.omitempty && isEmpty(fieldVal) {
continue
}
if opts.omitzero && isZero(sf) {
if opts.omitzero && isZero(fieldVal) {
continue
}
enc.encode(key.add(keyName), sf)
if inline {
enc.writeKeyValue(Key{keyName}, fieldVal, true)
if fieldIndex[0] != len(fields)-1 {
enc.wf(", ")
}
} else {
enc.encode(key.add(keyName), fieldVal)
}
}
}
if inline {
enc.wf("{")
}
writeFields(fieldsDirect)
writeFields(fieldsSub)
if inline {
enc.wf("}")
}
}
// tomlTypeName returns the TOML type name of the Go value's type. It is
@ -411,13 +498,26 @@ func tomlTypeOfGo(rv reflect.Value) tomlType {
switch rv.Interface().(type) {
case time.Time:
return tomlDatetime
case TextMarshaler:
case encoding.TextMarshaler:
return tomlString
default:
// Someone used a pointer receiver: we can make it work for pointer
// values.
if rv.CanAddr() {
_, ok := rv.Addr().Interface().(encoding.TextMarshaler)
if ok {
return tomlString
}
}
return tomlHash
}
default:
panic("unexpected reflect.Kind: " + rv.Kind().String())
_, ok := rv.Interface().(encoding.TextMarshaler)
if ok {
return tomlString
}
encPanic(errors.New("unsupported type: " + rv.Kind().String()))
panic("") // Need *some* return value
}
}
@ -430,30 +530,19 @@ func tomlArrayType(rv reflect.Value) tomlType {
if isNil(rv) || !rv.IsValid() || rv.Len() == 0 {
return nil
}
/// Don't allow nil.
rvlen := rv.Len()
for i := 1; i < rvlen; i++ {
if tomlTypeOfGo(rv.Index(i)) == nil {
encPanic(errArrayNilElement)
}
}
firstType := tomlTypeOfGo(rv.Index(0))
if firstType == nil {
encPanic(errArrayNilElement)
}
rvlen := rv.Len()
for i := 1; i < rvlen; i++ {
elem := rv.Index(i)
switch elemType := tomlTypeOfGo(elem); {
case elemType == nil:
encPanic(errArrayNilElement)
case !typeEqual(firstType, elemType):
encPanic(errArrayMixedElementTypes)
}
}
// If we have a nested array, then we must make sure that the nested
// array contains ONLY primitives.
// This checks arbitrarily nested arrays.
if typeEqual(firstType, tomlArray) || typeEqual(firstType, tomlArrayHash) {
nest := tomlArrayType(eindirect(rv.Index(0)))
if typeEqual(nest, tomlHash) || typeEqual(nest, tomlArrayHash) {
encPanic(errArrayNoTable)
}
}
return firstType
}
@ -511,14 +600,20 @@ func (enc *Encoder) newline() {
}
}
func (enc *Encoder) keyEqElement(key Key, val reflect.Value) {
// Write a key/value pair:
//
// key = <any value>
//
// If inline is true it won't add a newline at the end.
func (enc *Encoder) writeKeyValue(key Key, val reflect.Value, inline bool) {
if len(key) == 0 {
encPanic(errNoKey)
}
panicIfInvalidKey(key)
enc.wf("%s%s = ", enc.indentStr(key), key.maybeQuoted(len(key)-1))
enc.eElement(val)
enc.newline()
if !inline {
enc.newline()
}
}
func (enc *Encoder) wf(format string, v ...interface{}) {
@ -553,16 +648,3 @@ func isNil(rv reflect.Value) bool {
return false
}
}
func panicIfInvalidKey(key Key) {
for _, k := range key {
if len(k) == 0 {
encPanic(e("Key '%s' is not a valid table name. Key names "+
"cannot be empty.", key.maybeQuotedAll()))
}
}
}
func isValidKeyName(s string) bool {
return len(s) != 0
}

View file

@ -1,19 +0,0 @@
// +build go1.2
package toml
// In order to support Go 1.1, we define our own TextMarshaler and
// TextUnmarshaler types. For Go 1.2+, we just alias them with the
// standard library interfaces.
import (
"encoding"
)
// TextMarshaler is a synonym for encoding.TextMarshaler. It is defined here
// so that Go 1.1 can be supported.
type TextMarshaler encoding.TextMarshaler
// TextUnmarshaler is a synonym for encoding.TextUnmarshaler. It is defined
// here so that Go 1.1 can be supported.
type TextUnmarshaler encoding.TextUnmarshaler

View file

@ -1,18 +0,0 @@
// +build !go1.2
package toml
// These interfaces were introduced in Go 1.2, so we add them manually when
// compiling for Go 1.1.
// TextMarshaler is a synonym for encoding.TextMarshaler. It is defined here
// so that Go 1.1 can be supported.
type TextMarshaler interface {
MarshalText() (text []byte, err error)
}
// TextUnmarshaler is a synonym for encoding.TextUnmarshaler. It is defined
// here so that Go 1.1 can be supported.
type TextUnmarshaler interface {
UnmarshalText(text []byte) error
}

3
vendor/github.com/BurntSushi/toml/go.mod generated vendored Normal file
View file

@ -0,0 +1,3 @@
module github.com/BurntSushi/toml
go 1.16

0
vendor/github.com/BurntSushi/toml/go.sum generated vendored Normal file
View file

36
vendor/github.com/BurntSushi/toml/internal/tz.go generated vendored Normal file
View file

@ -0,0 +1,36 @@
package internal
import "time"
// Timezones used for local datetime, date, and time TOML types.
//
// The exact way times and dates without a timezone should be interpreted is not
// well-defined in the TOML specification and left to the implementation. These
// defaults to current local timezone offset of the computer, but this can be
// changed by changing these variables before decoding.
//
// TODO:
// Ideally we'd like to offer people the ability to configure the used timezone
// by setting Decoder.Timezone and Encoder.Timezone; however, this is a bit
// tricky: the reason we use three different variables for this is to support
// round-tripping – without these specific TZ names we wouldn't know which
// format to use.
//
// There isn't a good way to encode this right now though, and passing this sort
// of information also ties in to various related issues such as string format
// encoding, encoding of comments, etc.
//
// So, for the time being, just put this in internal until we can write a good
// comprehensive API for doing all of this.
//
// The reason they're exported is because they're referred from in e.g.
// internal/tag.
//
// Note that this behaviour is valid according to the TOML spec as the exact
// behaviour is left up to implementations.
var (
localOffset = func() int { _, o := time.Now().Zone(); return o }()
LocalDatetime = time.FixedZone("datetime-local", localOffset)
LocalDate = time.FixedZone("date-local", localOffset)
LocalTime = time.FixedZone("time-local", localOffset)
)

View file

@ -2,6 +2,8 @@ package toml
import (
"fmt"
"reflect"
"runtime"
"strings"
"unicode"
"unicode/utf8"
@ -29,6 +31,7 @@ const (
itemArrayTableStart
itemArrayTableEnd
itemKeyStart
itemKeyEnd
itemCommentStart
itemInlineTableStart
itemInlineTableEnd
@ -64,9 +67,9 @@ type lexer struct {
state stateFn
items chan item
// Allow for backing up up to three runes.
// Allow for backing up up to four runes.
// This is necessary because TOML contains 3-rune tokens (""" and ''').
prevWidths [3]int
prevWidths [4]int
nprev int // how many of prevWidths are in use
// If we emit an eof, we can still back up, but it is not OK to call
// next again.
@ -93,6 +96,7 @@ func (lx *lexer) nextItem() item {
return item
default:
lx.state = lx.state(lx)
//fmt.Printf(" STATE %-24s current: %-10q stack: %s\n", lx.state, lx.current(), lx.stack)
}
}
}
@ -137,7 +141,7 @@ func (lx *lexer) emitTrim(typ itemType) {
func (lx *lexer) next() (r rune) {
if lx.atEOF {
panic("next called after EOF")
panic("BUG in lexer: next called after EOF")
}
if lx.pos >= len(lx.input) {
lx.atEOF = true
@ -147,12 +151,19 @@ func (lx *lexer) next() (r rune) {
if lx.input[lx.pos] == '\n' {
lx.line++
}
lx.prevWidths[3] = lx.prevWidths[2]
lx.prevWidths[2] = lx.prevWidths[1]
lx.prevWidths[1] = lx.prevWidths[0]
if lx.nprev < 3 {
if lx.nprev < 4 {
lx.nprev++
}
r, w := utf8.DecodeRuneInString(lx.input[lx.pos:])
if r == utf8.RuneError {
lx.errorf("invalid UTF-8 byte at position %d (line %d): 0x%02x", lx.pos, lx.line, lx.input[lx.pos])
return utf8.RuneError
}
lx.prevWidths[0] = w
lx.pos += w
return r
@ -163,18 +174,19 @@ func (lx *lexer) ignore() {
lx.start = lx.pos
}
// backup steps back one rune. Can be called only twice between calls to next.
// backup steps back one rune. Can be called 4 times between calls to next.
func (lx *lexer) backup() {
if lx.atEOF {
lx.atEOF = false
return
}
if lx.nprev < 1 {
panic("backed up too far")
panic("BUG in lexer: backed up too far")
}
w := lx.prevWidths[0]
lx.prevWidths[0] = lx.prevWidths[1]
lx.prevWidths[1] = lx.prevWidths[2]
lx.prevWidths[2] = lx.prevWidths[3]
lx.nprev--
lx.pos -= w
if lx.pos < len(lx.input) && lx.input[lx.pos] == '\n' {
@ -269,8 +281,9 @@ func lexTopEnd(lx *lexer) stateFn {
lx.emit(itemEOF)
return nil
}
return lx.errorf("expected a top-level item to end with a newline, "+
"comment, or EOF, but got %q instead", r)
return lx.errorf(
"expected a top-level item to end with a newline, comment, or EOF, but got %q instead",
r)
}
// lexTable lexes the beginning of a table. Namely, it makes sure that
@ -297,8 +310,9 @@ func lexTableEnd(lx *lexer) stateFn {
func lexArrayTableEnd(lx *lexer) stateFn {
if r := lx.next(); r != arrayTableEnd {
return lx.errorf("expected end of table array name delimiter %q, "+
"but got %q instead", arrayTableEnd, r)
return lx.errorf(
"expected end of table array name delimiter %q, but got %q instead",
arrayTableEnd, r)
}
lx.emit(itemArrayTableEnd)
return lexTopEnd
@ -308,32 +322,19 @@ func lexTableNameStart(lx *lexer) stateFn {
lx.skip(isWhitespace)
switch r := lx.peek(); {
case r == tableEnd || r == eof:
return lx.errorf("unexpected end of table name " +
"(table names cannot be empty)")
return lx.errorf("unexpected end of table name (table names cannot be empty)")
case r == tableSep:
return lx.errorf("unexpected table separator " +
"(table names cannot be empty)")
return lx.errorf("unexpected table separator (table names cannot be empty)")
case r == stringStart || r == rawStringStart:
lx.ignore()
lx.push(lexTableNameEnd)
return lexValue // reuse string lexing
return lexQuotedName
default:
return lexBareTableName
lx.push(lexTableNameEnd)
return lexBareName
}
}
// lexBareTableName lexes the name of a table. It assumes that at least one
// valid character for the table has already been read.
func lexBareTableName(lx *lexer) stateFn {
r := lx.next()
if isBareKeyChar(r) {
return lexBareTableName
}
lx.backup()
lx.emit(itemText)
return lexTableNameEnd
}
// lexTableNameEnd reads the end of a piece of a table name, optionally
// consuming whitespace.
func lexTableNameEnd(lx *lexer) stateFn {
@ -347,63 +348,101 @@ func lexTableNameEnd(lx *lexer) stateFn {
case r == tableEnd:
return lx.pop()
default:
return lx.errorf("expected '.' or ']' to end table name, "+
"but got %q instead", r)
return lx.errorf("expected '.' or ']' to end table name, but got %q instead", r)
}
}
// lexKeyStart consumes a key name up until the first non-whitespace character.
// lexKeyStart will ignore whitespace.
func lexKeyStart(lx *lexer) stateFn {
r := lx.peek()
// lexBareName lexes one part of a key or table.
//
// It assumes that at least one valid character for the table has already been
// read.
//
// Lexes only one part, e.g. only 'a' inside 'a.b'.
func lexBareName(lx *lexer) stateFn {
r := lx.next()
if isBareKeyChar(r) {
return lexBareName
}
lx.backup()
lx.emit(itemText)
return lx.pop()
}
// lexBareName lexes one part of a key or table.
//
// It assumes that at least one valid character for the table has already been
// read.
//
// Lexes only one part, e.g. only '"a"' inside '"a".b'.
func lexQuotedName(lx *lexer) stateFn {
r := lx.next()
switch {
case r == keySep:
return lx.errorf("unexpected key separator %q", keySep)
case isWhitespace(r) || isNL(r):
lx.next()
return lexSkip(lx, lexKeyStart)
case isWhitespace(r):
return lexSkip(lx, lexValue)
case r == stringStart:
lx.ignore() // ignore the '"'
return lexString
case r == rawStringStart:
lx.ignore() // ignore the "'"
return lexRawString
case r == eof:
return lx.errorf("unexpected EOF; expected value")
default:
return lx.errorf("expected value but found %q instead", r)
}
}
// lexKeyStart consumes all key parts until a '='.
func lexKeyStart(lx *lexer) stateFn {
lx.skip(isWhitespace)
switch r := lx.peek(); {
case r == '=' || r == eof:
return lx.errorf("unexpected '=': key name appears blank")
case r == '.':
return lx.errorf("unexpected '.': keys cannot start with a '.'")
case r == stringStart || r == rawStringStart:
lx.ignore()
fallthrough
default: // Bare key
lx.emit(itemKeyStart)
lx.push(lexKeyEnd)
return lexValue // reuse string lexing
default:
lx.ignore()
lx.emit(itemKeyStart)
return lexBareKey
return lexKeyNameStart
}
}
// lexBareKey consumes the text of a bare key. Assumes that the first character
// (which is not whitespace) has not yet been consumed.
func lexBareKey(lx *lexer) stateFn {
switch r := lx.next(); {
case isBareKeyChar(r):
return lexBareKey
case isWhitespace(r):
lx.backup()
lx.emit(itemText)
return lexKeyEnd
case r == keySep:
lx.backup()
lx.emit(itemText)
return lexKeyEnd
func lexKeyNameStart(lx *lexer) stateFn {
lx.skip(isWhitespace)
switch r := lx.peek(); {
case r == '=' || r == eof:
return lx.errorf("unexpected '='")
case r == '.':
return lx.errorf("unexpected '.'")
case r == stringStart || r == rawStringStart:
lx.ignore()
lx.push(lexKeyEnd)
return lexQuotedName
default:
return lx.errorf("bare keys cannot contain %q", r)
lx.push(lexKeyEnd)
return lexBareName
}
}
// lexKeyEnd consumes the end of a key and trims whitespace (up to the key
// separator).
func lexKeyEnd(lx *lexer) stateFn {
lx.skip(isWhitespace)
switch r := lx.next(); {
case r == keySep:
return lexSkip(lx, lexValue)
case isWhitespace(r):
return lexSkip(lx, lexKeyEnd)
case r == eof:
return lx.errorf("unexpected EOF; expected key separator %q", keySep)
case r == '.':
lx.ignore()
return lexKeyNameStart
case r == '=':
lx.emit(itemKeyEnd)
return lexSkip(lx, lexValue)
default:
return lx.errorf("expected key separator %q, but got %q instead",
keySep, r)
return lx.errorf("expected '.' or '=', but got %q instead", r)
}
}
@ -450,10 +489,15 @@ func lexValue(lx *lexer) stateFn {
}
lx.ignore() // ignore the "'"
return lexRawString
case '+', '-':
return lexNumberStart
case '.': // special error case, be kind to users
return lx.errorf("floats must start with a digit, not '.'")
case 'i', 'n':
if (lx.accept('n') && lx.accept('f')) || (lx.accept('a') && lx.accept('n')) {
lx.emit(itemFloat)
return lx.pop()
}
case '-', '+':
return lexDecimalNumberStart
}
if unicode.IsLetter(r) {
// Be permissive here; lexBool will give a nice error if the
@ -463,6 +507,9 @@ func lexValue(lx *lexer) stateFn {
lx.backup()
return lexBool
}
if r == eof {
return lx.errorf("unexpected EOF; expected value")
}
return lx.errorf("expected value but found %q instead", r)
}
@ -507,9 +554,8 @@ func lexArrayValueEnd(lx *lexer) stateFn {
return lexArrayEnd
}
return lx.errorf(
"expected a comma or array terminator %q, but got %q instead",
arrayEnd, r,
)
"expected a comma or array terminator %q, but got %s instead",
arrayEnd, runeOrEOF(r))
}
// lexArrayEnd finishes the lexing of an array.
@ -546,8 +592,7 @@ func lexInlineTableValue(lx *lexer) stateFn {
// key/value pair and the next pair (or the end of the table):
// it ignores whitespace and expects either a ',' or a '}'.
func lexInlineTableValueEnd(lx *lexer) stateFn {
r := lx.next()
switch {
switch r := lx.next(); {
case isWhitespace(r):
return lexSkip(lx, lexInlineTableValueEnd)
case isNL(r):
@ -557,12 +602,25 @@ func lexInlineTableValueEnd(lx *lexer) stateFn {
return lexCommentStart
case r == comma:
lx.ignore()
lx.skip(isWhitespace)
if lx.peek() == '}' {
return lx.errorf("trailing comma not allowed in inline tables")
}
return lexInlineTableValue
case r == inlineTableEnd:
return lexInlineTableEnd
default:
return lx.errorf(
"expected a comma or an inline table terminator %q, but got %s instead",
inlineTableEnd, runeOrEOF(r))
}
return lx.errorf("expected a comma or an inline table terminator %q, "+
"but got %q instead", inlineTableEnd, r)
}
func runeOrEOF(r rune) string {
if r == eof {
return "end of file"
}
return "'" + string(r) + "'"
}
// lexInlineTableEnd finishes the lexing of an inline table.
@ -579,7 +637,9 @@ func lexString(lx *lexer) stateFn {
r := lx.next()
switch {
case r == eof:
return lx.errorf("unexpected EOF")
return lx.errorf(`unexpected EOF; expected '"'`)
case isControl(r) || r == '\r':
return lx.errorf("control characters are not allowed inside strings: '0x%02x'", r)
case isNL(r):
return lx.errorf("strings cannot contain newlines")
case r == '\\':
@ -598,19 +658,40 @@ func lexString(lx *lexer) stateFn {
// lexMultilineString consumes the inner contents of a string. It assumes that
// the beginning '"""' has already been consumed and ignored.
func lexMultilineString(lx *lexer) stateFn {
switch lx.next() {
r := lx.next()
switch r {
case eof:
return lx.errorf("unexpected EOF")
return lx.errorf(`unexpected EOF; expected '"""'`)
case '\r':
if lx.peek() != '\n' {
return lx.errorf("control characters are not allowed inside strings: '0x%02x'", r)
}
return lexMultilineString
case '\\':
return lexMultilineStringEscape
case stringEnd:
/// Found " → try to read two more "".
if lx.accept(stringEnd) {
if lx.accept(stringEnd) {
lx.backup()
/// Peek ahead: the string can contain " and "", including at the
/// end: """str"""""
/// 6 or more at the end, however, is an error.
if lx.peek() == stringEnd {
/// Check if we already lexed 5 's; if so we have 6 now, and
/// that's just too many man!
if strings.HasSuffix(lx.current(), `"""""`) {
return lx.errorf(`unexpected '""""""'`)
}
lx.backup()
lx.backup()
return lexMultilineString
}
lx.backup() /// backup: don't include the """ in the item.
lx.backup()
lx.backup()
lx.emit(itemMultilineString)
lx.next()
lx.next() /// Read over ''' again and discard it.
lx.next()
lx.next()
lx.ignore()
@ -619,6 +700,10 @@ func lexMultilineString(lx *lexer) stateFn {
lx.backup()
}
}
if isControl(r) {
return lx.errorf("control characters are not allowed inside strings: '0x%02x'", r)
}
return lexMultilineString
}
@ -628,7 +713,9 @@ func lexRawString(lx *lexer) stateFn {
r := lx.next()
switch {
case r == eof:
return lx.errorf("unexpected EOF")
return lx.errorf(`unexpected EOF; expected "'"`)
case isControl(r) || r == '\r':
return lx.errorf("control characters are not allowed inside strings: '0x%02x'", r)
case isNL(r):
return lx.errorf("strings cannot contain newlines")
case r == rawStringEnd:
@ -645,17 +732,38 @@ func lexRawString(lx *lexer) stateFn {
// a string. It assumes that the beginning "'''" has already been consumed and
// ignored.
func lexMultilineRawString(lx *lexer) stateFn {
switch lx.next() {
r := lx.next()
switch r {
case eof:
return lx.errorf("unexpected EOF")
return lx.errorf(`unexpected EOF; expected "'''"`)
case '\r':
if lx.peek() != '\n' {
return lx.errorf("control characters are not allowed inside strings: '0x%02x'", r)
}
return lexMultilineRawString
case rawStringEnd:
/// Found ' → try to read two more ''.
if lx.accept(rawStringEnd) {
if lx.accept(rawStringEnd) {
lx.backup()
/// Peek ahead: the string can contain ' and '', including at the
/// end: '''str'''''
/// 6 or more at the end, however, is an error.
if lx.peek() == rawStringEnd {
/// Check if we already lexed 5 's; if so we have 6 now, and
/// that's just too many man!
if strings.HasSuffix(lx.current(), "'''''") {
return lx.errorf(`unexpected "''''''"`)
}
lx.backup()
lx.backup()
return lexMultilineRawString
}
lx.backup() /// backup: don't include the ''' in the item.
lx.backup()
lx.backup()
lx.emit(itemRawMultilineString)
lx.next()
lx.next() /// Read over ''' again and discard it.
lx.next()
lx.next()
lx.ignore()
@ -664,6 +772,10 @@ func lexMultilineRawString(lx *lexer) stateFn {
lx.backup()
}
}
if isControl(r) {
return lx.errorf("control characters are not allowed inside strings: '0x%02x'", r)
}
return lexMultilineRawString
}
@ -694,6 +806,10 @@ func lexStringEscape(lx *lexer) stateFn {
fallthrough
case '"':
fallthrough
case ' ', '\t':
// Inside """ .. """ strings you can use \ to escape newlines, and any
// amount of whitespace can be between the \ and \n.
fallthrough
case '\\':
return lx.pop()
case 'u':
@ -701,8 +817,7 @@ func lexStringEscape(lx *lexer) stateFn {
case 'U':
return lexLongUnicodeEscape
}
return lx.errorf("invalid escape character %q; only the following "+
"escape characters are allowed: "+
return lx.errorf("invalid escape character %q; only the following escape characters are allowed: "+
`\b, \t, \n, \f, \r, \", \\, \uXXXX, and \UXXXXXXXX`, r)
}
@ -711,8 +826,9 @@ func lexShortUnicodeEscape(lx *lexer) stateFn {
for i := 0; i < 4; i++ {
r = lx.next()
if !isHexadecimal(r) {
return lx.errorf(`expected four hexadecimal digits after '\u', `+
"but got %q instead", lx.current())
return lx.errorf(
`expected four hexadecimal digits after '\u', but got %q instead`,
lx.current())
}
}
return lx.pop()
@ -723,28 +839,33 @@ func lexLongUnicodeEscape(lx *lexer) stateFn {
for i := 0; i < 8; i++ {
r = lx.next()
if !isHexadecimal(r) {
return lx.errorf(`expected eight hexadecimal digits after '\U', `+
"but got %q instead", lx.current())
return lx.errorf(
`expected eight hexadecimal digits after '\U', but got %q instead`,
lx.current())
}
}
return lx.pop()
}
// lexNumberOrDateStart consumes either an integer, a float, or datetime.
// lexNumberOrDateStart processes the first character of a value which begins
// with a digit. It exists to catch values starting with '0', so that
// lexBaseNumberOrDate can differentiate base prefixed integers from other
// types.
func lexNumberOrDateStart(lx *lexer) stateFn {
r := lx.next()
if isDigit(r) {
return lexNumberOrDate
}
switch r {
case '_':
return lexNumber
case 'e', 'E':
return lexFloat
case '.':
return lx.errorf("floats must start with a digit, not '.'")
case '0':
return lexBaseNumberOrDate
}
return lx.errorf("expected a digit but got %q", r)
if !isDigit(r) {
// The only way to reach this state is if the value starts
// with a digit, so specifically treat anything else as an
// error.
return lx.errorf("expected a digit but got %q", r)
}
return lexNumberOrDate
}
// lexNumberOrDate consumes either an integer, float or datetime.
@ -754,10 +875,10 @@ func lexNumberOrDate(lx *lexer) stateFn {
return lexNumberOrDate
}
switch r {
case '-':
case '-', ':':
return lexDatetime
case '_':
return lexNumber
return lexDecimalNumber
case '.', 'e', 'E':
return lexFloat
}
@ -775,41 +896,156 @@ func lexDatetime(lx *lexer) stateFn {
return lexDatetime
}
switch r {
case '-', 'T', ':', '.', 'Z', '+':
case '-', ':', 'T', 't', ' ', '.', 'Z', 'z', '+':
return lexDatetime
}
lx.backup()
lx.emit(itemDatetime)
lx.emitTrim(itemDatetime)
return lx.pop()
}
// lexNumberStart consumes either an integer or a float. It assumes that a sign
// has already been read, but that *no* digits have been consumed.
// lexNumberStart will move to the appropriate integer or float states.
func lexNumberStart(lx *lexer) stateFn {
// We MUST see a digit. Even floats have to start with a digit.
// lexHexInteger consumes a hexadecimal integer after seeing the '0x' prefix.
func lexHexInteger(lx *lexer) stateFn {
r := lx.next()
if !isDigit(r) {
if r == '.' {
return lx.errorf("floats must start with a digit, not '.'")
}
return lx.errorf("expected a digit but got %q", r)
}
return lexNumber
}
// lexNumber consumes an integer or a float after seeing the first digit.
func lexNumber(lx *lexer) stateFn {
r := lx.next()
if isDigit(r) {
return lexNumber
if isHexadecimal(r) {
return lexHexInteger
}
switch r {
case '_':
return lexNumber
return lexHexInteger
}
lx.backup()
lx.emit(itemInteger)
return lx.pop()
}
// lexOctalInteger consumes an octal integer after seeing the '0o' prefix.
func lexOctalInteger(lx *lexer) stateFn {
r := lx.next()
if isOctal(r) {
return lexOctalInteger
}
switch r {
case '_':
return lexOctalInteger
}
lx.backup()
lx.emit(itemInteger)
return lx.pop()
}
// lexBinaryInteger consumes a binary integer after seeing the '0b' prefix.
func lexBinaryInteger(lx *lexer) stateFn {
r := lx.next()
if isBinary(r) {
return lexBinaryInteger
}
switch r {
case '_':
return lexBinaryInteger
}
lx.backup()
lx.emit(itemInteger)
return lx.pop()
}
// lexDecimalNumber consumes a decimal float or integer.
func lexDecimalNumber(lx *lexer) stateFn {
r := lx.next()
if isDigit(r) {
return lexDecimalNumber
}
switch r {
case '.', 'e', 'E':
return lexFloat
case '_':
return lexDecimalNumber
}
lx.backup()
lx.emit(itemInteger)
return lx.pop()
}
// lexDecimalNumber consumes the first digit of a number beginning with a sign.
// It assumes the sign has already been consumed. Values which start with a sign
// are only allowed to be decimal integers or floats.
//
// The special "nan" and "inf" values are also recognized.
func lexDecimalNumberStart(lx *lexer) stateFn {
r := lx.next()
// Special error cases to give users better error messages
switch r {
case 'i':
if !lx.accept('n') || !lx.accept('f') {
return lx.errorf("invalid float: '%s'", lx.current())
}
lx.emit(itemFloat)
return lx.pop()
case 'n':
if !lx.accept('a') || !lx.accept('n') {
return lx.errorf("invalid float: '%s'", lx.current())
}
lx.emit(itemFloat)
return lx.pop()
case '0':
p := lx.peek()
switch p {
case 'b', 'o', 'x':
return lx.errorf("cannot use sign with non-decimal numbers: '%s%c'", lx.current(), p)
}
case '.':
return lx.errorf("floats must start with a digit, not '.'")
}
if isDigit(r) {
return lexDecimalNumber
}
return lx.errorf("expected a digit but got %q", r)
}
// lexBaseNumberOrDate differentiates between the possible values which
// start with '0'. It assumes that before reaching this state, the initial '0'
// has been consumed.
func lexBaseNumberOrDate(lx *lexer) stateFn {
r := lx.next()
// Note: All datetimes start with at least two digits, so we don't
// handle date characters (':', '-', etc.) here.
if isDigit(r) {
return lexNumberOrDate
}
switch r {
case '_':
// Can only be decimal, because there can't be an underscore
// between the '0' and the base designator, and dates can't
// contain underscores.
return lexDecimalNumber
case '.', 'e', 'E':
return lexFloat
case 'b':
r = lx.peek()
if !isBinary(r) {
lx.errorf("not a binary number: '%s%c'", lx.current(), r)
}
return lexBinaryInteger
case 'o':
r = lx.peek()
if !isOctal(r) {
lx.errorf("not an octal number: '%s%c'", lx.current(), r)
}
return lexOctalInteger
case 'x':
r = lx.peek()
if !isHexadecimal(r) {
lx.errorf("not a hexidecimal number: '%s%c'", lx.current(), r)
}
return lexHexInteger
}
lx.backup()
@ -867,21 +1103,22 @@ func lexCommentStart(lx *lexer) stateFn {
// It will consume *up to* the first newline character, and pass control
// back to the last state on the stack.
func lexComment(lx *lexer) stateFn {
r := lx.peek()
if isNL(r) || r == eof {
switch r := lx.next(); {
case isNL(r) || r == eof:
lx.backup()
lx.emit(itemText)
return lx.pop()
case isControl(r):
return lx.errorf("control characters are not allowed inside comments: '0x%02x'", r)
default:
return lexComment
}
lx.next()
return lexComment
}
// lexSkip ignores all slurped input and moves on to the next state.
func lexSkip(lx *lexer, nextState stateFn) stateFn {
return func(lx *lexer) stateFn {
lx.ignore()
return nextState
}
lx.ignore()
return nextState
}
// isWhitespace returns true if `r` is a whitespace character according
@ -894,6 +1131,16 @@ func isNL(r rune) bool {
return r == '\n' || r == '\r'
}
// Control characters except \n, \t
func isControl(r rune) bool {
switch r {
case '\t', '\r', '\n':
return false
default:
return (r >= 0x00 && r <= 0x1f) || r == 0x7f
}
}
func isDigit(r rune) bool {
return r >= '0' && r <= '9'
}
@ -904,6 +1151,14 @@ func isHexadecimal(r rune) bool {
(r >= 'A' && r <= 'F')
}
func isOctal(r rune) bool {
return r >= '0' && r <= '7'
}
func isBinary(r rune) bool {
return r == '0' || r == '1'
}
func isBareKeyChar(r rune) bool {
return (r >= 'A' && r <= 'Z') ||
(r >= 'a' && r <= 'z') ||
@ -912,6 +1167,17 @@ func isBareKeyChar(r rune) bool {
r == '-'
}
func (s stateFn) String() string {
name := runtime.FuncForPC(reflect.ValueOf(s).Pointer()).Name()
if i := strings.LastIndexByte(name, '.'); i > -1 {
name = name[i+1:]
}
if s == nil {
name = "<nil>"
}
return name + "()"
}
func (itype itemType) String() string {
switch itype {
case itemError:
@ -938,12 +1204,18 @@ func (itype itemType) String() string {
return "TableEnd"
case itemKeyStart:
return "KeyStart"
case itemKeyEnd:
return "KeyEnd"
case itemArray:
return "Array"
case itemArrayEnd:
return "ArrayEnd"
case itemCommentStart:
return "CommentStart"
case itemInlineTableStart:
return "InlineTableStart"
case itemInlineTableEnd:
return "InlineTableEnd"
}
panic(fmt.Sprintf("BUG: Unknown type '%d'.", int(itype)))
}

View file

@ -1,12 +1,14 @@
package toml
import (
"errors"
"fmt"
"strconv"
"strings"
"time"
"unicode"
"unicode/utf8"
"github.com/BurntSushi/toml/internal"
)
type parser struct {
@ -14,39 +16,54 @@ type parser struct {
types map[string]tomlType
lx *lexer
// A list of keys in the order that they appear in the TOML data.
ordered []Key
// the full key for the current hash in scope
context Key
// the base key name for everything except hashes
currentKey string
// rough approximation of line number
approxLine int
// A map of 'key.group.names' to whether they were created implicitly.
implicits map[string]bool
ordered []Key // List of keys in the order that they appear in the TOML data.
context Key // Full key for the current hash in scope.
currentKey string // Base key name for everything except hashes.
approxLine int // Rough approximation of line number
implicits map[string]bool // Record implied keys (e.g. 'key.group.names').
}
type parseError string
// ParseError is used when a file can't be parsed: for example invalid integer
// literals, duplicate keys, etc.
type ParseError struct {
Message string
Line int
LastKey string
}
func (pe parseError) Error() string {
return string(pe)
func (pe ParseError) Error() string {
return fmt.Sprintf("Near line %d (last key parsed '%s'): %s",
pe.Line, pe.LastKey, pe.Message)
}
func parse(data string) (p *parser, err error) {
defer func() {
if r := recover(); r != nil {
var ok bool
if err, ok = r.(parseError); ok {
if err, ok = r.(ParseError); ok {
return
}
panic(r)
}
}()
// Read over BOM; do this here as the lexer calls utf8.DecodeRuneInString()
// which mangles stuff.
if strings.HasPrefix(data, "\xff\xfe") || strings.HasPrefix(data, "\xfe\xff") {
data = data[2:]
}
// Examine first few bytes for NULL bytes; this probably means it's a UTF-16
// file (second byte in surrogate pair being NULL). Again, do this here to
// avoid having to deal with UTF-8/16 stuff in the lexer.
ex := 6
if len(data) < 6 {
ex = len(data)
}
if strings.ContainsRune(data[:ex], 0) {
return nil, errors.New("files cannot contain NULL bytes; probably using UTF-16; TOML files must be UTF-8")
}
p = &parser{
mapping: make(map[string]interface{}),
types: make(map[string]tomlType),
@ -66,13 +83,17 @@ func parse(data string) (p *parser, err error) {
}
func (p *parser) panicf(format string, v ...interface{}) {
msg := fmt.Sprintf("Near line %d (last key parsed '%s'): %s",
p.approxLine, p.current(), fmt.Sprintf(format, v...))
panic(parseError(msg))
msg := fmt.Sprintf(format, v...)
panic(ParseError{
Message: msg,
Line: p.approxLine,
LastKey: p.current(),
})
}
func (p *parser) next() item {
it := p.lx.nextItem()
//fmt.Printf("ITEM %-18s line %-3d │ %q\n", it.typ, it.line, it.val)
if it.typ == itemError {
p.panicf("%s", it.val)
}
@ -97,44 +118,63 @@ func (p *parser) assertEqual(expected, got itemType) {
func (p *parser) topLevel(item item) {
switch item.typ {
case itemCommentStart:
case itemCommentStart: // # ..
p.approxLine = item.line
p.expect(itemText)
case itemTableStart:
kg := p.next()
p.approxLine = kg.line
case itemTableStart: // [ .. ]
name := p.next()
p.approxLine = name.line
var key Key
for ; kg.typ != itemTableEnd && kg.typ != itemEOF; kg = p.next() {
key = append(key, p.keyString(kg))
for ; name.typ != itemTableEnd && name.typ != itemEOF; name = p.next() {
key = append(key, p.keyString(name))
}
p.assertEqual(itemTableEnd, kg.typ)
p.assertEqual(itemTableEnd, name.typ)
p.establishContext(key, false)
p.addContext(key, false)
p.setType("", tomlHash)
p.ordered = append(p.ordered, key)
case itemArrayTableStart:
kg := p.next()
p.approxLine = kg.line
case itemArrayTableStart: // [[ .. ]]
name := p.next()
p.approxLine = name.line
var key Key
for ; kg.typ != itemArrayTableEnd && kg.typ != itemEOF; kg = p.next() {
key = append(key, p.keyString(kg))
for ; name.typ != itemArrayTableEnd && name.typ != itemEOF; name = p.next() {
key = append(key, p.keyString(name))
}
p.assertEqual(itemArrayTableEnd, kg.typ)
p.assertEqual(itemArrayTableEnd, name.typ)
p.establishContext(key, true)
p.addContext(key, true)
p.setType("", tomlArrayHash)
p.ordered = append(p.ordered, key)
case itemKeyStart:
kname := p.next()
p.approxLine = kname.line
p.currentKey = p.keyString(kname)
case itemKeyStart: // key = ..
outerContext := p.context
/// Read all the key parts (e.g. 'a' and 'b' in 'a.b')
k := p.next()
p.approxLine = k.line
var key Key
for ; k.typ != itemKeyEnd && k.typ != itemEOF; k = p.next() {
key = append(key, p.keyString(k))
}
p.assertEqual(itemKeyEnd, k.typ)
val, typ := p.value(p.next())
p.setValue(p.currentKey, val)
p.setType(p.currentKey, typ)
/// The current key is the last part.
p.currentKey = key[len(key)-1]
/// All the other parts (if any) are the context; need to set each part
/// as implicit.
context := key[:len(key)-1]
for i := range context {
p.addImplicitContext(append(p.context, context[i:i+1]...))
}
/// Set value.
val, typ := p.value(p.next(), false)
p.set(p.currentKey, val, typ)
p.ordered = append(p.ordered, p.context.add(p.currentKey))
/// Remove the context we added (preserving any context from [tbl] lines).
p.context = outerContext
p.currentKey = ""
default:
p.bug("Unexpected type at top level: %s", item.typ)
@ -148,180 +188,253 @@ func (p *parser) keyString(it item) string {
return it.val
case itemString, itemMultilineString,
itemRawString, itemRawMultilineString:
s, _ := p.value(it)
s, _ := p.value(it, false)
return s.(string)
default:
p.bug("Unexpected key type: %s", it.typ)
panic("unreachable")
}
panic("unreachable")
}
var datetimeRepl = strings.NewReplacer(
"z", "Z",
"t", "T",
" ", "T")
// value translates an expected value from the lexer into a Go value wrapped
// as an empty interface.
func (p *parser) value(it item) (interface{}, tomlType) {
func (p *parser) value(it item, parentIsArray bool) (interface{}, tomlType) {
switch it.typ {
case itemString:
return p.replaceEscapes(it.val), p.typeOfPrimitive(it)
case itemMultilineString:
trimmed := stripFirstNewline(stripEscapedWhitespace(it.val))
return p.replaceEscapes(trimmed), p.typeOfPrimitive(it)
return p.replaceEscapes(stripFirstNewline(stripEscapedNewlines(it.val))), p.typeOfPrimitive(it)
case itemRawString:
return it.val, p.typeOfPrimitive(it)
case itemRawMultilineString:
return stripFirstNewline(it.val), p.typeOfPrimitive(it)
case itemInteger:
return p.valueInteger(it)
case itemFloat:
return p.valueFloat(it)
case itemBool:
switch it.val {
case "true":
return true, p.typeOfPrimitive(it)
case "false":
return false, p.typeOfPrimitive(it)
default:
p.bug("Expected boolean value, but got '%s'.", it.val)
}
p.bug("Expected boolean value, but got '%s'.", it.val)
case itemInteger:
if !numUnderscoresOK(it.val) {
p.panicf("Invalid integer %q: underscores must be surrounded by digits",
it.val)
}
val := strings.Replace(it.val, "_", "", -1)
num, err := strconv.ParseInt(val, 10, 64)
if err != nil {
// Distinguish integer values. Normally, it'd be a bug if the lexer
// provides an invalid integer, but it's possible that the number is
// out of range of valid values (which the lexer cannot determine).
// So mark the former as a bug but the latter as a legitimate user
// error.
if e, ok := err.(*strconv.NumError); ok &&
e.Err == strconv.ErrRange {
p.panicf("Integer '%s' is out of the range of 64-bit "+
"signed integers.", it.val)
} else {
p.bug("Expected integer value, but got '%s'.", it.val)
}
}
return num, p.typeOfPrimitive(it)
case itemFloat:
parts := strings.FieldsFunc(it.val, func(r rune) bool {
switch r {
case '.', 'e', 'E':
return true
}
return false
})
for _, part := range parts {
if !numUnderscoresOK(part) {
p.panicf("Invalid float %q: underscores must be "+
"surrounded by digits", it.val)
}
}
if !numPeriodsOK(it.val) {
// As a special case, numbers like '123.' or '1.e2',
// which are valid as far as Go/strconv are concerned,
// must be rejected because TOML says that a fractional
// part consists of '.' followed by 1+ digits.
p.panicf("Invalid float %q: '.' must be followed "+
"by one or more digits", it.val)
}
val := strings.Replace(it.val, "_", "", -1)
num, err := strconv.ParseFloat(val, 64)
if err != nil {
if e, ok := err.(*strconv.NumError); ok &&
e.Err == strconv.ErrRange {
p.panicf("Float '%s' is out of the range of 64-bit "+
"IEEE-754 floating-point numbers.", it.val)
} else {
p.panicf("Invalid float value: %q", it.val)
}
}
return num, p.typeOfPrimitive(it)
case itemDatetime:
var t time.Time
var ok bool
var err error
for _, format := range []string{
"2006-01-02T15:04:05Z07:00",
"2006-01-02T15:04:05",
"2006-01-02",
} {
t, err = time.ParseInLocation(format, it.val, time.Local)
if err == nil {
ok = true
break
}
}
if !ok {
p.panicf("Invalid TOML Datetime: %q.", it.val)
}
return t, p.typeOfPrimitive(it)
return p.valueDatetime(it)
case itemArray:
array := make([]interface{}, 0)
types := make([]tomlType, 0)
for it = p.next(); it.typ != itemArrayEnd; it = p.next() {
if it.typ == itemCommentStart {
p.expect(itemText)
continue
}
val, typ := p.value(it)
array = append(array, val)
types = append(types, typ)
}
return array, p.typeOfArray(types)
return p.valueArray(it)
case itemInlineTableStart:
var (
hash = make(map[string]interface{})
outerContext = p.context
outerKey = p.currentKey
)
p.context = append(p.context, p.currentKey)
p.currentKey = ""
for it := p.next(); it.typ != itemInlineTableEnd; it = p.next() {
if it.typ != itemKeyStart {
p.bug("Expected key start but instead found %q, around line %d",
it.val, p.approxLine)
}
if it.typ == itemCommentStart {
p.expect(itemText)
continue
}
// retrieve key
k := p.next()
p.approxLine = k.line
kname := p.keyString(k)
// retrieve value
p.currentKey = kname
val, typ := p.value(p.next())
// make sure we keep metadata up to date
p.setType(kname, typ)
p.ordered = append(p.ordered, p.context.add(p.currentKey))
hash[kname] = val
}
p.context = outerContext
p.currentKey = outerKey
return hash, tomlHash
return p.valueInlineTable(it, parentIsArray)
default:
p.bug("Unexpected value type: %s", it.typ)
}
p.bug("Unexpected value type: %s", it.typ)
panic("unreachable")
}
func (p *parser) valueInteger(it item) (interface{}, tomlType) {
if !numUnderscoresOK(it.val) {
p.panicf("Invalid integer %q: underscores must be surrounded by digits", it.val)
}
if numHasLeadingZero(it.val) {
p.panicf("Invalid integer %q: cannot have leading zeroes", it.val)
}
num, err := strconv.ParseInt(it.val, 0, 64)
if err != nil {
// Distinguish integer values. Normally, it'd be a bug if the lexer
// provides an invalid integer, but it's possible that the number is
// out of range of valid values (which the lexer cannot determine).
// So mark the former as a bug but the latter as a legitimate user
// error.
if e, ok := err.(*strconv.NumError); ok && e.Err == strconv.ErrRange {
p.panicf("Integer '%s' is out of the range of 64-bit signed integers.", it.val)
} else {
p.bug("Expected integer value, but got '%s'.", it.val)
}
}
return num, p.typeOfPrimitive(it)
}
func (p *parser) valueFloat(it item) (interface{}, tomlType) {
parts := strings.FieldsFunc(it.val, func(r rune) bool {
switch r {
case '.', 'e', 'E':
return true
}
return false
})
for _, part := range parts {
if !numUnderscoresOK(part) {
p.panicf("Invalid float %q: underscores must be surrounded by digits", it.val)
}
}
if len(parts) > 0 && numHasLeadingZero(parts[0]) {
p.panicf("Invalid float %q: cannot have leading zeroes", it.val)
}
if !numPeriodsOK(it.val) {
// As a special case, numbers like '123.' or '1.e2',
// which are valid as far as Go/strconv are concerned,
// must be rejected because TOML says that a fractional
// part consists of '.' followed by 1+ digits.
p.panicf("Invalid float %q: '.' must be followed by one or more digits", it.val)
}
val := strings.Replace(it.val, "_", "", -1)
if val == "+nan" || val == "-nan" { // Go doesn't support this, but TOML spec does.
val = "nan"
}
num, err := strconv.ParseFloat(val, 64)
if err != nil {
if e, ok := err.(*strconv.NumError); ok && e.Err == strconv.ErrRange {
p.panicf("Float '%s' is out of the range of 64-bit IEEE-754 floating-point numbers.", it.val)
} else {
p.panicf("Invalid float value: %q", it.val)
}
}
return num, p.typeOfPrimitive(it)
}
var dtTypes = []struct {
fmt string
zone *time.Location
}{
{time.RFC3339Nano, time.Local},
{"2006-01-02T15:04:05.999999999", internal.LocalDatetime},
{"2006-01-02", internal.LocalDate},
{"15:04:05.999999999", internal.LocalTime},
}
func (p *parser) valueDatetime(it item) (interface{}, tomlType) {
it.val = datetimeRepl.Replace(it.val)
var (
t time.Time
ok bool
err error
)
for _, dt := range dtTypes {
t, err = time.ParseInLocation(dt.fmt, it.val, dt.zone)
if err == nil {
ok = true
break
}
}
if !ok {
p.panicf("Invalid TOML Datetime: %q.", it.val)
}
return t, p.typeOfPrimitive(it)
}
func (p *parser) valueArray(it item) (interface{}, tomlType) {
p.setType(p.currentKey, tomlArray)
// p.setType(p.currentKey, typ)
var (
array []interface{}
types []tomlType
)
for it = p.next(); it.typ != itemArrayEnd; it = p.next() {
if it.typ == itemCommentStart {
p.expect(itemText)
continue
}
val, typ := p.value(it, true)
array = append(array, val)
types = append(types, typ)
}
return array, tomlArray
}
func (p *parser) valueInlineTable(it item, parentIsArray bool) (interface{}, tomlType) {
var (
hash = make(map[string]interface{})
outerContext = p.context
outerKey = p.currentKey
)
p.context = append(p.context, p.currentKey)
prevContext := p.context
p.currentKey = ""
p.addImplicit(p.context)
p.addContext(p.context, parentIsArray)
/// Loop over all table key/value pairs.
for it := p.next(); it.typ != itemInlineTableEnd; it = p.next() {
if it.typ == itemCommentStart {
p.expect(itemText)
continue
}
/// Read all key parts.
k := p.next()
p.approxLine = k.line
var key Key
for ; k.typ != itemKeyEnd && k.typ != itemEOF; k = p.next() {
key = append(key, p.keyString(k))
}
p.assertEqual(itemKeyEnd, k.typ)
/// The current key is the last part.
p.currentKey = key[len(key)-1]
/// All the other parts (if any) are the context; need to set each part
/// as implicit.
context := key[:len(key)-1]
for i := range context {
p.addImplicitContext(append(p.context, context[i:i+1]...))
}
/// Set the value.
val, typ := p.value(p.next(), false)
p.set(p.currentKey, val, typ)
p.ordered = append(p.ordered, p.context.add(p.currentKey))
hash[p.currentKey] = val
/// Restore context.
p.context = prevContext
}
p.context = outerContext
p.currentKey = outerKey
return hash, tomlHash
}
// numHasLeadingZero checks if this number has leading zeroes, allowing for '0',
// +/- signs, and base prefixes.
func numHasLeadingZero(s string) bool {
if len(s) > 1 && s[0] == '0' && isDigit(rune(s[1])) { // >1 to allow "0" and isDigit to allow 0x
return true
}
if len(s) > 2 && (s[0] == '-' || s[0] == '+') && s[1] == '0' {
return true
}
return false
}
// numUnderscoresOK checks whether each underscore in s is surrounded by
// characters that are not underscores.
func numUnderscoresOK(s string) bool {
switch s {
case "nan", "+nan", "-nan", "inf", "-inf", "+inf":
return true
}
accept := false
for _, r := range s {
if r == '_' {
if !accept {
return false
}
accept = false
continue
}
accept = true
// isHexadecimal is a superset of all the permissable characters
// surrounding an underscore.
accept = isHexadecimal(r)
}
return accept
}
@ -338,13 +451,12 @@ func numPeriodsOK(s string) bool {
return !period
}
// establishContext sets the current context of the parser,
// where the context is either a hash or an array of hashes. Which one is
// set depends on the value of the `array` parameter.
// Set the current context of the parser, where the context is either a hash or
// an array of hashes, depending on the value of the `array` parameter.
//
// Establishing the context also makes sure that the key isn't a duplicate, and
// will create implicit hashes automatically.
func (p *parser) establishContext(key Key, array bool) {
func (p *parser) addContext(key Key, array bool) {
var ok bool
// Always start at the top level and drill down for our context.
@ -383,7 +495,7 @@ func (p *parser) establishContext(key Key, array bool) {
// list of tables for it.
k := key[len(key)-1]
if _, ok := hashContext[k]; !ok {
hashContext[k] = make([]map[string]interface{}, 0, 5)
hashContext[k] = make([]map[string]interface{}, 0, 4)
}
// Add a new table. But make sure the key hasn't already been used
@ -391,8 +503,7 @@ func (p *parser) establishContext(key Key, array bool) {
if hash, ok := hashContext[k].([]map[string]interface{}); ok {
hashContext[k] = append(hash, make(map[string]interface{}))
} else {
p.panicf("Key '%s' was already created and cannot be used as "+
"an array.", keyContext)
p.panicf("Key '%s' was already created and cannot be used as an array.", keyContext)
}
} else {
p.setValue(key[len(key)-1], make(map[string]interface{}))
@ -400,15 +511,22 @@ func (p *parser) establishContext(key Key, array bool) {
p.context = append(p.context, key[len(key)-1])
}
// set calls setValue and setType.
func (p *parser) set(key string, val interface{}, typ tomlType) {
p.setValue(p.currentKey, val)
p.setType(p.currentKey, typ)
}
// setValue sets the given key to the given value in the current context.
// It will make sure that the key hasn't already been defined, account for
// implicit key groups.
func (p *parser) setValue(key string, value interface{}) {
var tmpHash interface{}
var ok bool
hash := p.mapping
keyContext := make(Key, 0)
var (
tmpHash interface{}
ok bool
hash = p.mapping
keyContext Key
)
for _, k := range p.context {
keyContext = append(keyContext, k)
if tmpHash, ok = hash[k]; !ok {
@ -422,24 +540,26 @@ func (p *parser) setValue(key string, value interface{}) {
case map[string]interface{}:
hash = t
default:
p.bug("Expected hash to have type 'map[string]interface{}', but "+
"it has '%T' instead.", tmpHash)
p.panicf("Key '%s' has already been defined.", keyContext)
}
}
keyContext = append(keyContext, key)
if _, ok := hash[key]; ok {
// Typically, if the given key has already been set, then we have
// to raise an error since duplicate keys are disallowed. However,
// it's possible that a key was previously defined implicitly. In this
// case, it is allowed to be redefined concretely. (See the
// `tests/valid/implicit-and-explicit-after.toml` test in `toml-test`.)
// Normally redefining keys isn't allowed, but the key could have been
// defined implicitly and it's allowed to be redefined concretely. (See
// the `valid/implicit-and-explicit-after.toml` in toml-test)
//
// But we have to make sure to stop marking it as an implicit. (So that
// another redefinition provokes an error.)
//
// Note that since it has already been defined (as a hash), we don't
// want to overwrite it. So our business is done.
if p.isArray(keyContext) {
p.removeImplicit(keyContext)
hash[key] = value
return
}
if p.isImplicit(keyContext) {
p.removeImplicit(keyContext)
return
@ -449,6 +569,7 @@ func (p *parser) setValue(key string, value interface{}) {
// key, which is *always* wrong.
p.panicf("Key '%s' has already been defined.", keyContext)
}
hash[key] = value
}
@ -468,21 +589,15 @@ func (p *parser) setType(key string, typ tomlType) {
p.types[keyContext.String()] = typ
}
// addImplicit sets the given Key as having been created implicitly.
func (p *parser) addImplicit(key Key) {
p.implicits[key.String()] = true
}
// removeImplicit stops tagging the given key as having been implicitly
// created.
func (p *parser) removeImplicit(key Key) {
p.implicits[key.String()] = false
}
// isImplicit returns true if the key group pointed to by the key was created
// implicitly.
func (p *parser) isImplicit(key Key) bool {
return p.implicits[key.String()]
// Implicit keys need to be created when tables are implied in "a.b.c.d = 1" and
// "[a.b.c]" (the "a", "b", and "c" hashes are never created explicitly).
func (p *parser) addImplicit(key Key) { p.implicits[key.String()] = true }
func (p *parser) removeImplicit(key Key) { p.implicits[key.String()] = false }
func (p *parser) isImplicit(key Key) bool { return p.implicits[key.String()] }
func (p *parser) isArray(key Key) bool { return p.types[key.String()] == tomlArray }
func (p *parser) addImplicitContext(key Key) {
p.addImplicit(key)
p.addContext(key, false)
}
// current returns the full key name of the current context.
@ -497,20 +612,54 @@ func (p *parser) current() string {
}
func stripFirstNewline(s string) string {
if len(s) == 0 || s[0] != '\n' {
return s
if len(s) > 0 && s[0] == '\n' {
return s[1:]
}
return s[1:]
if len(s) > 1 && s[0] == '\r' && s[1] == '\n' {
return s[2:]
}
return s
}
func stripEscapedWhitespace(s string) string {
esc := strings.Split(s, "\\\n")
if len(esc) > 1 {
for i := 1; i < len(esc); i++ {
esc[i] = strings.TrimLeftFunc(esc[i], unicode.IsSpace)
// Remove newlines inside triple-quoted strings if a line ends with "\".
func stripEscapedNewlines(s string) string {
split := strings.Split(s, "\n")
if len(split) < 1 {
return s
}
escNL := false // Keep track of the last non-blank line was escaped.
for i, line := range split {
line = strings.TrimRight(line, " \t\r")
if len(line) == 0 || line[len(line)-1] != '\\' {
split[i] = strings.TrimRight(split[i], "\r")
if !escNL && i != len(split)-1 {
split[i] += "\n"
}
continue
}
escBS := true
for j := len(line) - 1; j >= 0 && line[j] == '\\'; j-- {
escBS = !escBS
}
if escNL {
line = strings.TrimLeft(line, " \t\r")
}
escNL = !escBS
if escBS {
split[i] += "\n"
continue
}
split[i] = line[:len(line)-1] // Remove \
if len(split)-1 > i {
split[i+1] = strings.TrimLeft(split[i+1], " \t\r")
}
}
return strings.Join(esc, "")
return strings.Join(split, "")
}
func (p *parser) replaceEscapes(str string) string {
@ -533,6 +682,9 @@ func (p *parser) replaceEscapes(str string) string {
default:
p.bug("Expected valid escape code after \\, but got %q.", s[r])
return ""
case ' ', '\t':
p.panicf("invalid escape: '\\%c'", s[r])
return ""
case 'b':
replaced = append(replaced, rune(0x0008))
r += 1
@ -585,8 +737,3 @@ func (p *parser) asciiEscapeToUnicode(bs []byte) rune {
}
return rune(hex)
}
func isStringType(ty itemType) bool {
return ty == itemString || ty == itemMultilineString ||
ty == itemRawString || ty == itemRawMultilineString
}

View file

@ -1 +0,0 @@
au BufWritePost *.go silent!make tags > /dev/null 2>&1

View file

@ -68,24 +68,3 @@ func (p *parser) typeOfPrimitive(lexItem item) tomlType {
p.bug("Cannot infer primitive type of lex item '%s'.", lexItem)
panic("unreachable")
}
// typeOfArray returns a tomlType for an array given a list of types of its
// values.
//
// In the current spec, if an array is homogeneous, then its type is always
// "Array". If the array is not homogeneous, an error is generated.
func (p *parser) typeOfArray(types []tomlType) tomlType {
// Empty arrays are cool.
if len(types) == 0 {
return tomlArray
}
theType := types[0]
for _, t := range types[1:] {
if !typeEqual(theType, t) {
p.panicf("Array contains values of type '%s' and '%s', but "+
"arrays must be homogeneous.", theType, t)
}
}
return tomlArray
}