Currently, the marshalling code encodes the embedded structs as sub-tables.
This is a bit unexpected, as it differs from what encoding/json does in
that case: https://play.golang.org/p/KDPaGtrijV1
Unmarshalling code handles this scenario gracefully.
This PR adapts the encoder to behave like encoding/json.
Fields in an embedded struct are promoted to the top level table.
In case the embedded struct is named in the tag, it will still
encode as a sub-table.
The added PromoteAnonymous option on the Encoder allows configuring
the old behavior, where anonymous structs are encoded as sub-tables.
On duplicate keys, the behavior of encoding/json is mimicked:
Fields from anonymous structs are shadowed by regular fields.
An example is added to show the affects of setting PromoteAnonymous.
* add test for unexported field preservation
* merge struct values instead of replacing them
* use struct merging on nested value structs
* unmarshalling merges nested struct pointers when non-nil
Previously, this would fail with:
```
panic: reflect.Value.SetMapIndex: value of type string is not assignable to type toml.letter [recovered]
panic: reflect.Value.SetMapIndex: value of type string is not assignable to type toml.letter
```
Now this only panics when the key type cannot be converted from a
string.
When a struct is unmarshalled, go-toml now looks at the `default` tag to
provide a default value in case the key is not present in the TOML
document. This is only implemented for string, bool, int, int64,
float64. Additional types can be further implemented on a request-basis.
The name for each field in a struct is used to look up a key in the TOML
tree. A few different (case-sensitive) forms of this name are tried.
Previously, the current, lower-cased, and title-cased versions of the
name are tried. This precludes camelCased keys from mapping back to
fields in structs. This change adds camelCase to the set of keys to
try.
For example, the following TOML:
fooBar = 10
Would previously *not* map to the following struct:
type Foo struct {
FooBar int
}
This change corrects this.
Decoder: allow to customize default field name tag "toml" on decoding.
Example:
```
type doc struct {
title `file:"header"`
}
```
Encoder: allow to customize tags for encoding struct to toml.
Example:
```
type doc struct {
title `file:"header" description:"document title"`
}
```
Fixes#238
A new Encoder option emits arrays with more than one line on multiple lines.
This is off by default and toggled with `ArraysWithOneElementPerLine`.
For example:
```
A = [1,2,3]
```
Becomes:
```
A = [
1,
2,
3
]
```
Fixes#200
Usage is similar to the stdlibs JSON encoder/decoder but I tried to
leave the general structure of the code the same.
Main motivation was to support encoding/decoding options to allow
encoding string-type map keys as quoted TOML keys.
This was implemented on the Encoder with QuoteMapKeys(bool).
> The TOML spec supports using UTF-8 strings as keys.
> https://github.com/toml-lang/toml/blob/master/versions/en/toml-v0.4.0.md#table
Oncilla