A modern and extensible serialization framework for Nim
Go to file
Zahary Karadjov d8ca3daf3a
raises annotations for some generated procs
2020-05-30 22:37:23 +03:00
serialization raises annotations for some generated procs 2020-05-30 22:37:23 +03:00
tests CI: update script path 2019-08-24 05:14:20 +02:00
.appveyor.yml CI: update script path 2019-08-24 05:14:20 +02:00
.travis.yml CI: update script path 2019-08-24 05:14:20 +02:00
README.md add AppVeyor badge 2019-08-24 05:22:46 +02:00
serialization.nim Flush happens automatically on close 2020-05-13 11:40:04 +03:00
serialization.nimble nim 1.2 2020-04-24 08:56:54 +02:00

README.md

nim-serialization

Build Status Build status License: Apache License: MIT

Introduction

The serialization package aims to provide a common generic and efficient interface for marshaling Nim values to and from various serialized formats. Individual formats are implemented in separated packages such as json_serialization while this package provides the common interfaces shared between all of them and the means to customize your Nim types for the purposes of serialization.

The internal mechanisms of the library allow for implementing the required marshaling logic in highly efficient way that goes from bytes to Nim values and vice versa without allocating any intermediate structures.

Defining serialization formats

A serialization format is implemented through defining a Reader and Writer type for the format and then by providing the following type declaration:

serializationFormat Json,                          # This is the name of the format.
                                                   # Most APIs provided by the library will accept
                                                   # this identifier as a required parameter.

                    Reader = JsonReader,           # The associated Reader type.
                    Writer = JsonWriter,           # The associated Writer type.

                    PreferedOutput = string,       # APIs such as `Json.encode` will return this type.
                                                   # The type must support the following operations:
                                                   #   proc initWithCapacity(_: type T, n: int)
                                                   #   proc add(v: var T, bytes: openarray[byte])
                                                   # By default, the PreferedOutput is `seq[byte]`.

                    mimeType = "application/json", # Mime type associated with the format (Optional).
                    fileExt = "json"               # File extension associated with the format (Optional).

Common API

Most of the time, you'll be using the following high-level APIs when encoding and decoding values:

Format.encode(value: auto, params: varargs): Format.PreferedOutput

Encodes a value in the specified format returning the preferred output type for the format (usually string or seq[byte]). All extra params will be forwarded without modification to the constructor of the used Writer type.

Example:

assert Json.encode(@[1, 2, 3], pretty = false) == "[1, 2, 3]"

Format.decode(input: openarray[byte]|string, RecordType: type, params: varargs): RecordType

Decodes and returns a value of the specified RecordType. All params will be forwarded without modification to the used Reader type. A Format-specific descendant of SerializationError may be thrown in case of error.

Format.saveFile(filename: string, params: varargs)

Similar to encode, but saves the result in a file.

Format.loadFile

Similar to decode, but treats the contents of a file as an input.

reader.readValue(RecordType: type): RecordType

Reads a single value of the designated type from the stream associated with a particular reader.

writer.writeValue(value: auto)

Encodes a single value and writes it to the output stream of a paticular writer.

Custom serialization of user-defined types

By default, record types will have all of their fields serialized. You can alter this behavior by attaching the dontSerialize(varargs[Format]) pragma to black-list fields or the serialize(varargs[Format]) to white-list fields. If no formats are listed in the pragmas above, they will affect all formats. The pragma serializedFieldName(name: string) can be used to modify the name of the field in formats such as Json and XML.

Alternatively, if you are not able to modify the definition of a particular Nim type, you can use the serializedFields macro to achive the same in less instrusive way.

The following two definitions can be consired equivalent:

type
  Foo = object
    a: string
    b {.dontSerialize.}: int
    c {.dontSerialize(JSON).}: float
    d {.serializedFieldName("z").}: int

serializedFields Foo:
  a
  c [-JSON]
  # The `d` field is renamed to `z`: 
  d -> z

As you can see, serializedFields accepts a block where each serialzied field is listed on a separate line. The -> operator is used to indicate field renames, while the optional [] modifiers can be used to while-list or black-list specific formats where the field should appear (using + and - respectively).

customSerialization(RecordType: type, spec)

Format.customSerialization(RecordType, spec)

customSerialization(RecordType.field, spec)

Format.customSerialization(Record.field, spec)

Record.totalSerializedFields(Format)

Returns the number of serialized fields in the specified format.

Implementing Readers

Implementing Writers

Contributing

When submitting pull requests, please add test cases for any new features or fixes and make sure nimble test is still able to execute the entire test suite successfully.

License

Licensed and distributed under either of

or

at your option. These files may not be copied, modified, or distributed except according to those terms.