DataKit offers a modern, intuitive and declarative interface for reading and writing binary formatted data in Swift.
As an introduction into how this library can be used to make working with binary formatted data easier, let me first introduce you to the type, we are going to read/write. Let's assume we are building a weather station and we are using the following type(s) to give updates about the currently measured values:
struct WeatherStationFeatures: OptionSet, ReadWritable {
var rawValue: UInt8
static var hasTemperature = Self(rawValue: 1 << 0)
static var hasHumidity = Self(rawValue: 1 << 1)
static var usesMetricUnits = Self(rawValue: 1 << 2)
}
struct WeatherStationUpdate {
var features: WeatherStationFeatures
var temperature: Measurement<UnitTemperature>
var humidity: Double
}
The encoded format should be:
- Each message starts with a byte with the value 0x02.
- The following byte contains multiple feature flags:
- bit 0 is set: Using Β°C instead of Β°F for the temperature
- bit 1 is set: The message contains temperature information
- bit 2 is set: The message contains humidity information
- Temperature as a big-endian 32-bit floating-point number
- Relative Humidity as UInt8 in the range of [0, 100]
- CRC-32 with the default polynomial for the whole message (incl. 0x02 prefix).
You have two options for converting the above type WeatherStationUpdate
into data: A DataBuilder
or the Writable
protocol. If you intend to both read and write the data - make sure to read the Reading & Writing data
section
A DataBuilder
provides you with a very simple and limited interface. Using the power of result builders, you can simply state the values to be written in a given order and DataBuilder
will take over all the work to encode the values and append the individual bytes to form a Data
object. DataBuilder
is always expected to return a Data
object without throwing errors, which is why conversion are not supported here - You might want to have a look at the Writable
protocol then!
extension WeatherStationUpdate {
@DataBuilder var data: Data {
UInt8(0x02)
features
if features.contains(.hasTemperature) {
Float(temperature.converted(to: features.contains(.usesMetricUnits) ? .celsius : .fahrenheit).value)
}
if features.contains(.hasHumidity) {
UInt8(humidity * 100)
}
CRC32.default
}
}
With this addition, you can easily get the data of this object using its data
property.
With the power of keyPaths and result builders, you can also write your objects into Data
using the Writable
protocol and its writeFormat
property. Simply state out individual fixed values (e.g. byte prefixes), keyPaths with Writable
values or other constructs that are further explained in the Extras
section of this document.
extension WeatherStationUpdate: Writable {
static var writeFormat: WriteFormat {
Scope {
UInt8(0x02)
\.features
Using(\.features) { features in
if features.contains(.hasTemperature) {
let unit: UnitTemperature =
features.contains(.usesMetricUnits) ? .celsius : .fahrenheit
Convert(\.temperature) {
$0.converted(to: unit).cast(Float.self)
}
}
if features.contains(.hasHumidity) {
Convert(\.humidity) {
Double($0) / 100
} writing: {
UInt8($0 * 100)
}
}
}
CRC32.default
}
.endianness(.big)
}
}
By conforming to the Writable
protocol, you are now able to simply call its write
function to write its data out:
let message: WeatherStationUpdate = ...
let messageData = try message.write() // You can also inject a custom environment here, if needed
Supporting reading of data into objects is slightly more complicated. Conforming to the Readable
protocol will require you to implement both an initializer to create an object from a given ReadContext
and a static readFormat
property describing how data is aligned.
A ReadContext
provides you with the values that have been read using the readFormat
. Make sure to use the same keyPaths in the initializer and readFormat
to ensure smooth reading of values.
extension WeatherStationUpdate: Readable {
init(from context: ReadContext<WeatherStationUpdate>) throws {
features = try context.read(for: \.features)
temperature = try context.readIfPresent(for: \.temperature) ?? .init(value: .nan, unit: .kelvin)
humidity = try context.readIfPresent(for: \.humidity) ?? .nan
}
static var readFormat: ReadFormat {
Scope {
UInt8(0x02)
\.features
Using(\.features) { features in
if features.contains(.hasTemperature) {
let unit: UnitTemperature =
features.contains(.usesMetricUnits) ? .celsius : .fahrenheit
Convert(\.temperature) {
$0.converted(to: unit).cast(Float.self)
}
}
if features.contains(.hasHumidity) {
Convert(\.humidity) {
Double($0) / 100
} writing: {
UInt8($0 * 100)
}
}
}
CRC32.default
}
.endianness(.big)
}
}
By implementing all these requirements of the Readable
protocol, you now gain another initializer init(_: Data) throws
to read objects from Data
objects:
let data: Data = ...
let message = try WeatherStationUpdate(data) // You can also inject a custom environment here, if needed
To make a type both Readable
and Writable
, you can conform your type to the ReadWritable
protocol. Instead of providing a separate format for reading and writing, you can define a Format
property that is used for both reading and writing. For our example type, we can simply merge the two formats into one and provide the initializer for creating an object from a given ReadContext
.
extension WeatherStationUpdate: ReadWritable {
init(from context: ReadContext<WeatherStationUpdate>) throws {
features = try context.read(for: \.features)
temperature = try context.readIfPresent(for: \.temperature) ?? .init(value: .nan, unit: .kelvin)
humidity = try context.readIfPresent(for: \.humidity) ?? .nan
}
static var format: Format {
Scope {
UInt8(0x02)
\.features
Using(\.features) { features in
if features.contains(.hasTemperature) {
let unit: UnitTemperature =
features.contains(.usesMetricUnits) ? .celsius : .fahrenheit
Convert(\.temperature) {
$0.converted(to: unit).cast(Float.self)
}
}
if features.contains(.hasHumidity) {
Convert(\.humidity) {
Double($0) / 100
} writing: {
UInt8($0 * 100)
}
}
}
CRC32.default
}
.endianness(.big)
}
}
Hooray, you can now read and write your objects! π
Reading/Writing data is often quite complicated and different format pose different challenges to minimize payloads, reduce bandwidth, improve performance, etc. To make it easy to handle different common scenarios, DataKit
provides a couple of extra features to handle the most common challenges.
In some special cases, you might need more control over how data is read/written. For these cases, the wrappers Custom
, Convert
and Property
might be of interest.
Property
makes it easy to wrap a keyPath, if the Root type may not be recognized by the compiler. You can further use functions on it to map aProperty
to either aCustom
orConvert
wrapper.Convert
allows you to convert a keyPath's value before reading/writing it. Oftentimes, this is very usefuly for sequence values with variable size. You can either provide custom conversion methods directly or use a pre-existingConversion
/ReversibleConversion
value.Custom
allows you to access the raw reading/writing functionality with direct access to theReadContainer
/WriteContainer
and respective context values. If you need the read/write behavior more than once in your codebase, you might want to have a look at conversions though.
For some types, there is not a single "correct" format (e.g. thinking about Pascal vs C strings), which is why DataKit
uses so called Conversion
values to allow for conversion to be defined once and then used multiple times. Especially helpful is the ReversibleConversion
type that allows for conversion to be provided in both directions at the same time.
Assuming our type has a \.string
keyPath with a String
value, you could either use a suffix 0-byte to encode the string using UTF8 (similar to C strings):
Convert(\.string) { // UTF8 string with a suffix 0-byte
$0.encoded(.utf8).dynamicCount
}
.suffix(0 as UInt8)
Or you encode the string with a prefix byte containing the byte count (similar to Pascal ShortString):
Convert(\.string) { // Ascii string with a prefix count byte
$0.encoded(.ascii).prefixCount(UInt8.self)
}
There are many more conversion available, e.g. for converting between integer/floating-point types, making it easy to convert directly to your preferred types without the need of converting yourself.
With a Using
construct, you can access values from the ReadContext
or the value to be written. Using
can be very helpful, if values in the data itself depend on each other or how individual values need to be read or written.
Similar to SwiftUI's environment, you can also modify the behavior of individual components in DataKit
using the Environment
.
You can modify the environment when starting the reading/writing process or using modifiers inside the readFormat/writeFormat/format-properties.
To access environment values, you may want to have a look at the Environment
type to be used in one of the format builders - or for more direct access both ReadContainer
and WriteContainer
have a environment
property.
Here are some modifiers, you might want to use:
endianness
: By default,DataKit
reads & writes values in the endianness of the current machine (except for CRCs, where big-endian is used). If your protocol requires a different endianness, make sure to specify a concrete one.skipChecksumVerification
: If you want to have a CRC to be created when writing out data, but ignore an incorrect checksum value when reading, you can set this property totrue
- by defaultfalse
is used. Read theChecksums
section for more information.suffix
: For values with dynamic count (e.g. a sequence of values with a 0-suffix-byte), you can specify a.dynamicCount
conversion on a given property. The specified value will stop the reading process of a given value and the value will be written out after the given sequence's end is encountered.
Feel free to add your own environment values using the EnvironnentKey
protocol and an extension to the EnvironmentValues
struct (very similar to SwiftUI).
Some constructs (e.g. CRC checksums) make assumptions about the data as a whole and not only the part where the specific value is read/written. By using Scope
, you can limit that data context to where it is individually placed.
As an example, let's assume our WeatherStationUpdate
is supposed to ignore the prefix 0x02
byte for the checksum calculation. We could simply exclude it from the scope:
static var format: Format {
UInt8(0x02)
Scope {
\.features
Using(\.features) { features in
if features.contains(.hasTemperature) {
let unit: UnitTemperature =
features.contains(.usesMetricUnits) ? .celsius : .fahrenheit
Convert(\.temperature) {
$0.converted(to: unit).cast(Float.self)
}
}
if features.contains(.hasHumidity) {
Convert(\.humidity) {
Double($0) / 100
} writing: {
UInt8($0 * 100)
}
}
}
CRC32.default
}
.endianness(.big)
}
With this change in place, the CRC will only be verified on the Scope itself and the prefix byte is ignored!
DataKit
's dependency crc-swift
provides CRC checksums and and easy-to-conform protocol for your own custom checksum values.
You may simply specify the checksum value itself inside one of the format builders. Alternatively, use a ChecksumProperty
with a keyPath, so that you can store checksums in properties and write custom checksums from properties. In combination to the skipChecksumVerification
environment value, you can also verify the checksum at a later stage for example.
DataKit
currently only supports Swift package manager.
See this WWDC presentation about more information how to adopt Swift packages in your app.
Specify https://github.com/QuickBirdEng/DataKit.git
as the package link.
If you prefer not to use a dependency manager, you can integrate DataKit into your project manually by downloading the source code and placing the files in your project directory.
DataKit is created with β€οΈ by QuickBird.
Feel free to open issues for help, found bugs or to discuss new feature requests. Happy to help! Open a pull request, if you want to propose changes to DataKit.
DataKit is released under an MIT license. See License.md for more information.