mirror of
https://github.com/42wim/matterbridge.git
synced 2024-11-29 22:22:01 -08:00
300 lines
7.6 KiB
Go
300 lines
7.6 KiB
Go
// Copyright 2012 The Gorilla Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package schema
|
|
|
|
import (
|
|
"encoding"
|
|
"errors"
|
|
"fmt"
|
|
"reflect"
|
|
"strings"
|
|
)
|
|
|
|
// NewDecoder returns a new Decoder.
|
|
func NewDecoder() *Decoder {
|
|
return &Decoder{cache: newCache()}
|
|
}
|
|
|
|
// Decoder decodes values from a map[string][]string to a struct.
|
|
type Decoder struct {
|
|
cache *cache
|
|
zeroEmpty bool
|
|
ignoreUnknownKeys bool
|
|
}
|
|
|
|
// SetAliasTag changes the tag used to locate custom field aliases.
|
|
// The default tag is "schema".
|
|
func (d *Decoder) SetAliasTag(tag string) {
|
|
d.cache.tag = tag
|
|
}
|
|
|
|
// ZeroEmpty controls the behaviour when the decoder encounters empty values
|
|
// in a map.
|
|
// If z is true and a key in the map has the empty string as a value
|
|
// then the corresponding struct field is set to the zero value.
|
|
// If z is false then empty strings are ignored.
|
|
//
|
|
// The default value is false, that is empty values do not change
|
|
// the value of the struct field.
|
|
func (d *Decoder) ZeroEmpty(z bool) {
|
|
d.zeroEmpty = z
|
|
}
|
|
|
|
// IgnoreUnknownKeys controls the behaviour when the decoder encounters unknown
|
|
// keys in the map.
|
|
// If i is true and an unknown field is encountered, it is ignored. This is
|
|
// similar to how unknown keys are handled by encoding/json.
|
|
// If i is false then Decode will return an error. Note that any valid keys
|
|
// will still be decoded in to the target struct.
|
|
//
|
|
// To preserve backwards compatibility, the default value is false.
|
|
func (d *Decoder) IgnoreUnknownKeys(i bool) {
|
|
d.ignoreUnknownKeys = i
|
|
}
|
|
|
|
// RegisterConverter registers a converter function for a custom type.
|
|
func (d *Decoder) RegisterConverter(value interface{}, converterFunc Converter) {
|
|
d.cache.regconv[reflect.TypeOf(value)] = converterFunc
|
|
}
|
|
|
|
// Decode decodes a map[string][]string to a struct.
|
|
//
|
|
// The first parameter must be a pointer to a struct.
|
|
//
|
|
// The second parameter is a map, typically url.Values from an HTTP request.
|
|
// Keys are "paths" in dotted notation to the struct fields and nested structs.
|
|
//
|
|
// See the package documentation for a full explanation of the mechanics.
|
|
func (d *Decoder) Decode(dst interface{}, src map[string][]string) error {
|
|
v := reflect.ValueOf(dst)
|
|
if v.Kind() != reflect.Ptr || v.Elem().Kind() != reflect.Struct {
|
|
return errors.New("schema: interface must be a pointer to struct")
|
|
}
|
|
v = v.Elem()
|
|
t := v.Type()
|
|
errors := MultiError{}
|
|
for path, values := range src {
|
|
if parts, err := d.cache.parsePath(path, t); err == nil {
|
|
if err = d.decode(v, path, parts, values); err != nil {
|
|
errors[path] = err
|
|
}
|
|
} else if !d.ignoreUnknownKeys {
|
|
errors[path] = fmt.Errorf("schema: invalid path %q", path)
|
|
}
|
|
}
|
|
if len(errors) > 0 {
|
|
return errors
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// decode fills a struct field using a parsed path.
|
|
func (d *Decoder) decode(v reflect.Value, path string, parts []pathPart, values []string) error {
|
|
// Get the field walking the struct fields by index.
|
|
for _, name := range parts[0].path {
|
|
if v.Type().Kind() == reflect.Ptr {
|
|
if v.IsNil() {
|
|
v.Set(reflect.New(v.Type().Elem()))
|
|
}
|
|
v = v.Elem()
|
|
}
|
|
v = v.FieldByName(name)
|
|
}
|
|
|
|
// Don't even bother for unexported fields.
|
|
if !v.CanSet() {
|
|
return nil
|
|
}
|
|
|
|
// Dereference if needed.
|
|
t := v.Type()
|
|
if t.Kind() == reflect.Ptr {
|
|
t = t.Elem()
|
|
if v.IsNil() {
|
|
v.Set(reflect.New(t))
|
|
}
|
|
v = v.Elem()
|
|
}
|
|
|
|
// Slice of structs. Let's go recursive.
|
|
if len(parts) > 1 {
|
|
idx := parts[0].index
|
|
if v.IsNil() || v.Len() < idx+1 {
|
|
value := reflect.MakeSlice(t, idx+1, idx+1)
|
|
if v.Len() < idx+1 {
|
|
// Resize it.
|
|
reflect.Copy(value, v)
|
|
}
|
|
v.Set(value)
|
|
}
|
|
return d.decode(v.Index(idx), path, parts[1:], values)
|
|
}
|
|
|
|
// Get the converter early in case there is one for a slice type.
|
|
conv := d.cache.converter(t)
|
|
if conv == nil && t.Kind() == reflect.Slice {
|
|
var items []reflect.Value
|
|
elemT := t.Elem()
|
|
isPtrElem := elemT.Kind() == reflect.Ptr
|
|
if isPtrElem {
|
|
elemT = elemT.Elem()
|
|
}
|
|
|
|
// Try to get a converter for the element type.
|
|
conv := d.cache.converter(elemT)
|
|
if conv == nil {
|
|
// As we are not dealing with slice of structs here, we don't need to check if the type
|
|
// implements TextUnmarshaler interface
|
|
return fmt.Errorf("schema: converter not found for %v", elemT)
|
|
}
|
|
|
|
for key, value := range values {
|
|
if value == "" {
|
|
if d.zeroEmpty {
|
|
items = append(items, reflect.Zero(elemT))
|
|
}
|
|
} else if item := conv(value); item.IsValid() {
|
|
if isPtrElem {
|
|
ptr := reflect.New(elemT)
|
|
ptr.Elem().Set(item)
|
|
item = ptr
|
|
}
|
|
if item.Type() != elemT && !isPtrElem {
|
|
item = item.Convert(elemT)
|
|
}
|
|
items = append(items, item)
|
|
} else {
|
|
if strings.Contains(value, ",") {
|
|
values := strings.Split(value, ",")
|
|
for _, value := range values {
|
|
if value == "" {
|
|
if d.zeroEmpty {
|
|
items = append(items, reflect.Zero(elemT))
|
|
}
|
|
} else if item := conv(value); item.IsValid() {
|
|
if isPtrElem {
|
|
ptr := reflect.New(elemT)
|
|
ptr.Elem().Set(item)
|
|
item = ptr
|
|
}
|
|
if item.Type() != elemT && !isPtrElem {
|
|
item = item.Convert(elemT)
|
|
}
|
|
items = append(items, item)
|
|
} else {
|
|
return ConversionError{
|
|
Key: path,
|
|
Type: elemT,
|
|
Index: key,
|
|
}
|
|
}
|
|
}
|
|
} else {
|
|
return ConversionError{
|
|
Key: path,
|
|
Type: elemT,
|
|
Index: key,
|
|
}
|
|
}
|
|
}
|
|
}
|
|
value := reflect.Append(reflect.MakeSlice(t, 0, 0), items...)
|
|
v.Set(value)
|
|
} else {
|
|
val := ""
|
|
// Use the last value provided if any values were provided
|
|
if len(values) > 0 {
|
|
val = values[len(values)-1]
|
|
}
|
|
|
|
if val == "" {
|
|
if d.zeroEmpty {
|
|
v.Set(reflect.Zero(t))
|
|
}
|
|
} else if conv != nil {
|
|
if value := conv(val); value.IsValid() {
|
|
v.Set(value.Convert(t))
|
|
} else {
|
|
return ConversionError{
|
|
Key: path,
|
|
Type: t,
|
|
Index: -1,
|
|
}
|
|
}
|
|
} else {
|
|
// When there's no registered conversion for the custom type, we will check if the type
|
|
// implements the TextUnmarshaler interface. As the UnmarshalText function should be applied
|
|
// to the pointer of the type, we convert the value to pointer.
|
|
if v.CanAddr() {
|
|
v = v.Addr()
|
|
}
|
|
|
|
if u, ok := v.Interface().(encoding.TextUnmarshaler); ok {
|
|
if err := u.UnmarshalText([]byte(val)); err != nil {
|
|
return ConversionError{
|
|
Key: path,
|
|
Type: t,
|
|
Index: -1,
|
|
Err: err,
|
|
}
|
|
}
|
|
|
|
} else {
|
|
return fmt.Errorf("schema: converter not found for %v", t)
|
|
}
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// Errors ---------------------------------------------------------------------
|
|
|
|
// ConversionError stores information about a failed conversion.
|
|
type ConversionError struct {
|
|
Key string // key from the source map.
|
|
Type reflect.Type // expected type of elem
|
|
Index int // index for multi-value fields; -1 for single-value fields.
|
|
Err error // low-level error (when it exists)
|
|
}
|
|
|
|
func (e ConversionError) Error() string {
|
|
var output string
|
|
|
|
if e.Index < 0 {
|
|
output = fmt.Sprintf("schema: error converting value for %q", e.Key)
|
|
} else {
|
|
output = fmt.Sprintf("schema: error converting value for index %d of %q",
|
|
e.Index, e.Key)
|
|
}
|
|
|
|
if e.Err != nil {
|
|
output = fmt.Sprintf("%s. Details: %s", output, e.Err)
|
|
}
|
|
|
|
return output
|
|
}
|
|
|
|
// MultiError stores multiple decoding errors.
|
|
//
|
|
// Borrowed from the App Engine SDK.
|
|
type MultiError map[string]error
|
|
|
|
func (e MultiError) Error() string {
|
|
s := ""
|
|
for _, err := range e {
|
|
s = err.Error()
|
|
break
|
|
}
|
|
switch len(e) {
|
|
case 0:
|
|
return "(0 errors)"
|
|
case 1:
|
|
return s
|
|
case 2:
|
|
return s + " (and 1 other error)"
|
|
}
|
|
return fmt.Sprintf("%s (and %d other errors)", s, len(e)-1)
|
|
}
|