zbor - Zig CBOR

Donate using Liberapay

The Concise Binary Object Representation (CBOR) is a data format whose design goals include the possibility of extremely small code size, fairly small message size, and extensibility without the need for version negotiation (RFC8949). It is used in different protocols like the Client to Authenticator Protocol CTAP2 which is a essential part of FIDO2 authenticators/ Passkeys.

I have utilized this library in several projects throughout the previous year, primarily in conjunction with my FIDO2 library. I'd consider it stable. With the introduction of Zig version 0.11.0, this library will remain aligned with the most recent stable release. If you have any problems or want to share some ideas feel free to open an issue or write me a mail, but please be kind.

Getting started

Versions

First add this library as a dependency to your build.zig.zon file:

# Replace <VERSION TAG> with the version you want to use
zig fetch --save https://codeberg.org/r4gus/zbor/archive/<VERSION TAG>.tar.gz
# e.g. zig fetch --save https://codeberg.org/r4gus/zbor/archive/0.20.1.tar.gz

then within you build.zig add the following code:

// First fetch the dependency...constzbor_dep=b.dependency("zbor",.{.target=target,.optimize=optimize,});constzbor_module=zbor_dep.module("zbor");// If you have a module that has zbor as a dependency...constyour_module=b.addModule("your-module",.{.root_source_file=.{.path="src/main.zig"},.imports=&.{.{.name="zbor",.module=zbor_module},},});// Or as a dependency for a executable...exe.root_module.addImport("zbor",zbor_module);

Usage

This library lets you inspect and parse CBOR data without having to allocate additional memory.

Inspect CBOR data

To inspect CBOR data you must first create a new DataItem.

constcbor=@import("zbor");constdi=DataItem.new("\x1b\xff\xff\xff\xff\xff\xff\xff\xff")catch{// handle the case that the given data is malformed};

DataItem.new() will check if the given data is well-formed before returning a DataItem. The data is well formed if it's syntactically correct.

You can also read from a std.Io.Reader using the DataItem.readAlloc function. The function will read a CBOR data item from a Reader and output a DataItem object. If you use the readAlloc method, don't forget to free the allocated memory using the `deinit() ́ method!

To check the type of the given DataItem use the getType() function.

std.debug.assert(di.getType()==.Int);

Possible types include Int (major type 0 and 1) ByteString (major type 2), TextString (major type 3), Array (major type 4), Map (major type 5), Tagged (major type 6) and Float (major type 7).

Based on the given type you can the access the underlying value.

std.debug.assert(di.int().?==18446744073709551615);

All getter functions return either a value or null. You can use a pattern like if (di.int()) |v| v else return error.Oops; to access the value in a safe way. If you've used DataItem.new() and know the type of the data item, you should be safe to just do di.int().?.

The following getter functions are supported:

• int - returns ?i65
• string - returns ?[]const u8
• array - returns ?ArrayIterator
• map - returns ?MapIterator
• simple - returns ?u8
• float - returns ?f64
• tagged - returns ?Tag
• boolean - returns ?bool

Iterators

The functions array and map will return an iterator. Every time you call next() you will either get a DataItem/ Pair or null.

constdi=DataItem.new("\x98\x19\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0a\x0b\x0c\x0d\x0e\x0f\x10\x11\x12\x13\x14\x15\x16\x17\x18\x18\x18\x19");variter=di.array().?;while(iter.next())|value|{_=value;// doe something}

Encoding and decoding

Serialization

You can serialize Zig objects into CBOR using the stringify() function.

constallocator=std.testing.allocator;varstr=std.Io.Writer.Allocating.init(allocator);deferstr.deinit();constInfo=struct{versions:[]const[]constu8,};consti=Info{.versions=&.{"FIDO_2_0"},};trystringify(i,.{},&str.writer);

Note: Compile time floats are always encoded as single precision floats (f32). Please use @floatCast before passing a float to stringify().

The stringify() function is convenient but also adds extra overhead. If you want full control over the serialization process you can use the following functions defined in zbor.build: writeInt, writeByteString, writeTextString, writeTag, writeSimple, writeArray, writeMap. For more details check out the manual serialization example and the corresponding source code.

Stringify Options

You can pass options to the stringify function to influence its behavior. Without passing any options, stringify will behave as follows:

• Enums will be serialized to their textual representation
• u8 slices will be serialized to byte strings
• For structs and unions:
  • null fields are skipped by default
  • fields of type std.mem.Allocator are always skipped.
  • the names of fields are serialized to text strings

You can modify that behavior by changing the default options, e.g.:

constEcdsaP256Key=struct{/// kty:kty:u8=2,/// alg:alg:i8=-7,/// crv:crv:u8=1,/// x-coordinatex:[32]u8,/// y-coordinatey:[32]u8,pubfnnew(k:EcdsaP256.PublicKey)@This(){constxy=k.toUncompressedSec1();return.{.x=xy[1..33].*,.y=xy[33..65].*,};}};//...trystringify(k,.{.field_settings=&.{.{.name="kty",.field_options=.{.alias="1",.serialization_type=.Integer}},.{.name="alg",.field_options=.{.alias="3",.serialization_type=.Integer}},.{.name="crv",.field_options=.{.alias="-1",.serialization_type=.Integer}},.{.name="x",.field_options=.{.alias="-2",.serialization_type=.Integer}},.{.name="y",.field_options=.{.alias="-3",.serialization_type=.Integer}},}},&str.writer);

Here we define a alias for every field of the struct and tell serialize that it should treat those aliases as integers instead of text strings.

See Options and FieldSettings in src/parse.zig for all available options!

Deserialization

You can deserialize CBOR data into Zig objects using the parse() function.

conste=[5]u8{1,2,3,4,5};constdi=DataItem.new("\x85\x01\x02\x03\x04\x05");constx=tryparse([5]u8,di,.{});trystd.testing.expectEqualSlices(u8,e[0..],x[0..]);

Parse Options

You can pass options to the parse function to influence its behaviour.

This includes:

• allocator - The allocator to be used. This is required if your data type has any pointers, slices, etc.
• duplicate_field_behavior - How to handle duplicate fields (.UseFirst, .Error).
  • .UseFirst - Use the first field.
  • .Error - Return an error if there are multiple fields with the same name.
• ignore_unknown_fields - Ignore unknown fields (default is true).
• field_settings - Lets you specify aliases for struct fields. Examples on how to use field_settings can be found in the examples directory and within defined tests.
• ignore_override - Flag to break infinity loops. This has to be set to true if you override the behavior using cborParse or cborStringify.

Builder

You can also dynamically create CBOR data using the Builder.

constallocator=std.testing.allocator;varb=tryBuilder.withType(allocator,.Map);tryb.pushTextString("a");tryb.pushInt(1);tryb.pushTextString("b");tryb.enter(.Array);tryb.pushInt(2);tryb.pushInt(3);//try b.leave(); <-- you can leave out the return at the endconstx=tryb.finish();deferallocator.free(x);// { "a": 1, "b": [2, 3] }trystd.testing.expectEqualSlices(u8,"\xa2\x61\x61\x01\x61\x62\x82\x02\x03",x);

Commands

• The push* functions append a data item
• The enter function takes a container type and pushes it on the builder stack
• The leave function leaves the current container. The container is appended to the wrapping container
• The finish function returns the CBOR data as owned slice

Overriding stringify

You can override the stringify function for structs and tagged unions by implementing cborStringify.

constFoo=struct{x:u32=1234,y:struct{a:[]constu8="public-key",b:u64=0x1122334455667788,},pubfncborStringify(self:*const@This(),options:Options,out:*std.Io.Writer)!void{// First stringify the 'y' structconstallocator=std.testing.allocator;varo=std.Io.Writer.Allocating.init(allocator);defero.deinit();trystringify(self.y,options,&o.writer);// Then use the Builder to alter the CBOR outputvarb=trybuild.Builder.withType(allocator,.Map);tryb.pushTextString("x");tryb.pushInt(self.x);tryb.pushTextString("y");tryb.pushByteString(o.written());constx=tryb.finish();deferallocator.free(x);tryout.writeAll(x);}};

The StringifyOptions can be used to indirectly pass an Allocator to the function.

Please make sure to set ignore_override to true when calling recursively into stringify(self) to prevent infinite loops.

Overriding parse

You can override the parse function for structs and tagged unions by implementing cborParse. This is helpful if you have aliases for your struct members.

constEcdsaP256Key=struct{/// kty:kty:u8=2,/// alg:alg:i8=-7,/// crv:crv:u8=1,/// x-coordinatex:[32]u8,/// y-coordinatey:[32]u8,pubfncborParse(item:DataItem,options:Options)!@This(){_=options;returntryparse(@This(),item,.{.ignore_override=true,// prevent infinite loops.field_settings=&.{.{.name="kty",.field_options=.{.alias="1"}},.{.name="alg",.field_options=.{.alias="3"}},.{.name="crv",.field_options=.{.alias="-1"}},.{.name="x",.field_options=.{.alias="-2"}},.{.name="y",.field_options=.{.alias="-3"}},},});}};

The Options can be used to indirectly pass an Allocator to the function.

Please make sure to set ignore_override to true when calling recursively into parse(self) to prevent infinite loops.

Structs with fields of type std.mem.Allocator

If you have a struct with a field of type std.mem.Allocator you have to override the stringify funcation for that struct, e.g.:

pubfncborStringify(self:*const@This(),options:cbor.StringifyOptions,out:*std.Io.Writer)!void{_=options;trycbor.stringify(self,.{.ignore_override=true,.field_settings=&.{.{.name="allocator",.options=.{.skip=true}},},},out);}

When using parse make sure you pass a allocator to the function. The passed allocator will be assigned to the field of type std.mem.Allocator.

Indefinite-length Data Items

CBOR supports the serialization of many container types in two formats, definite and indefinite. For definite-length data items, the length is directly encoded into the data-items header. In contrast, indefinite-length data items are terminated by a break-byte 0xff.

Zbor currently supports indefinite-length encoding for both arrays and maps. The default serialization type for both types remains definite to support backwards compatibility. One can control the serialization type for arrays and maps via the serialization options. The two fields in question are array_serialization_type and map_serialization_type.

Indefinite-length Arrays

This is an example for serializing a array as indefinite-length map:

constarray=[_]u16{500,2};vararr=std.Io.Writer.Allocating.init(allocator);deferarr.deinit();trystringify(array,.{.allocator=allocator,.array_serialization_type=.ArrayIndefinite,},&arr.writer,);

For the de-serialization of indefinite-length arrays you don't have to do anything special. The parse function will automatically detect the encoding type for you.

Indefinite-length Maps

This is an example for serializing a struct as indefinite-length map:

constallocator=std.testing.allocator;constS=struct{Fun:bool,Amt:i16,};consts=S{.Fun=true,.Amt=-2,};vararr=std.Io.Writer.Allocating.init(allocator);deferarr.deinit();trystringify(s,.{.allocator=allocator,.map_serialization_type=.MapIndefinite,},&arr.writer,);

For the de-serialization of indefinite-length maps you don't have to do anything special. The parse function will automatically detect the encoding type for you.

ArrayBackedSlice

This library offers a convenient function named ArrayBackedSlice, which enables you to create a wrapper for an array of any size and type. This wrapper implements the cborStringify and cborParse methods, allowing it to seamlessly replace slices (e.g., []const u8) with an array.

test"ArrayBackedSlice test"{constallocator=std.testing.allocator;constS64B=ArrayBackedSlice(64,u8,.Byte);varx=S64B{};tryx.set("\x01\x02\x03\x04");varstr=std.Io.Writer.Allocating.init(allocator);deferstr.deinit();trystringify(x,.{},&str.writer);trystd.testing.expectEqualSlices(u8,"\x44\x01\x02\x03\x04",str.written());constdi=tryDataItem.new(str.written());consty=tryparse(S64B,di,.{});trystd.testing.expectEqualSlices(u8,"\x01\x02\x03\x04",y.get());}