Batch get and empty sets fix

This is a bugfix release that resolves an issue where BatchGet would return ErrNotFound early upon encountering an empty set of 100 results (#161). Thanks @Dave-Dohmeier!

Expression wrapping fix

This is a bugfix release that fixes #158, in which expressions like (A) OR (B) were not getting properly wrapped to ((A) OR (B)) when combined with multiple .If calls, etc.
Also includes #157 which fixes ListTables's context methods not passing the context through to the AWS SDK.

Update.Set auto-omit fix

This is a bugfix release that fixes Update.Set's behavior when given an empty typed string value (#151). Instead of removing the empty value, it was encoding an invalid request resulting in a SerializationException. Thanks to @finnbear for discovering and triaging the issue.

Number sets fix

This is a minor release that fixes an issue with decoding number sets (#150). Thanks to @aoldershaw for the contribution.

New Update features and AWSEncoding fixes

This release adds Update.DeleteFromSet which can be useful for deleting values from sets not covered by the pre-existing helper methods (#146). Also, Update.OnlyUpdatedValue and Update.OnlyUpdatedOldValue have been added, which are the equivalents of UPDATED_NEW and UPDATED_OLD and encode only the values that have been changed from an update (#142).

AWSEncoding support has been improved (#147). You can now use it with methods that operate on slices, such as:

var list []myAWSEncodedStuff
table.Get("UserID", 123).All(dynamo.AWSEncoding(&list))

This release also bumps up aws-sdk-go to the latest version.
Thanks to everyone who reported issues and sent PRs, and to everyone who uses this library.

Encoding improvements

This release adds support for embedding struct pointers (#139) and avoids panicking when trying to unmarshal to private fields. Thanks @gir for the contribution.

Improved support for null and empty values

This release improves and extends many areas of the encoder, improving support for empty and null values and fixing some important issues.
We can now take advantage of DynamoDB's support for empty strings and binary attributes!

• Added support for empty/null values in DynamoDB lists (slices and arrays) and sets (slices or maps with the dynamo:",set" struct tag option). By default, empty strings and binary values are encoded as-is, and nil pointers are set to a DynamoDB NULL type. Previously these cases were a SerializationException. Fixes #102 and #137.
  • Added a new struct tag option dynamo:",omitemptyelem" that omits these values instead. Note that this will mess up your array indexes.
• Added a new struct tag option dynamo:",allowempty" to disable auto-omit behavior for empty string or binary field values. Keep in mind that primary keys can't be empty.
• Added a new struct tag option dynamo:",allowemptyelem" to disable auto-omit behavior for empty string, empty binary, and nil values in maps.
• Added a new struct tag option dynamo:",null" that will marshal nil or empty values to DynamoDB NULL types. When allowempty is also set, allowempty will take precedence.
• Added Update.SetNullable which is Update.Set without automatic empty/nil removal.
• Added support for empty/nil arguments in If and Filter.
• Added support for custom empty structs in sets: #133, thanks @finnbear.
• Docs improvements: #123, #136. Thanks @ZacharyGroff and @d-tsuji.

This release should have total backwards compatibility with previous versions. New support for empty/nil values was added in areas that were previously serialization errors. Auto-omit behavior was kept the same, but now you can use allowempty and allowemptyelem to disable it.

Example:

var widget = struct {
	HashKey string
	Desc    string `dynamo:",allowempty"`
	Words   []string
	Ptr     *int `dynamo:",null"`
}{
	HashKey: "abc",
	Desc:    "",
	Words:   []string{"hello", "world", ""},
	Ptr:     nil,
}

will be marshaled like so

{
	HashKey: {S: "abc"},
	Desc: {S: ""},
	Words: {L: [{S: "hello"}, {S: "world"} {S: ""}]},
	Ptr: {NULL: true},
}

Big thanks to everyone who submitted issues and PRs, and all the users of this library.

Custom item marshaling

This release adds ItemMarshaler and ItemUnmarshaler, which are interfaces allowing for total control of item marshaling. This is useful for handling dynamic items whose structure depends on its content. Thanks to @luma for submitting this in PR #126!

// ItemMarshaler is the interface implemented by objects that
// can marshal themselves into an Item (a map of strings
// to AttributeValues).
type ItemMarshaler interface {
	MarshalDynamoItem() (map[string]*dynamodb.AttributeValue, error)
}

// ItemUnmarshaler is the interface implemented by
// objects that can unmarshal an Item (a map of strings
//to AttributeValues) into themselves.
type ItemUnmarshaler interface {
	UnmarshalDynamoItem(item map[string]*dynamodb.AttributeValue) error
}

AWSEncoding fixes

This release fixes a bug where AWSEncoding items were not being (un)marshaled properly in Put, Get, etc (#125).
The module dependency on aws-sdk-go was also updated to the latest version.

Improved array handling ②

This release fixes unmarshaling to return an error when unmarshaling a DynamoDB list into an array that is too small.
See notes for v1.7.1, which includes a similar fix for DynamoDB binary data.