An implementation of Canonical JSON.
The "canonical JSON" format is designed to provide repeatable hashes of JSON values. It is designed for applications that need to hash, sign or authenitcate JSON data structures., including embedded signatures.
Canonical JSON is parsable with any full JSON parser, and it allows white space for pretty-printed human readable presentation, but it can be put into a canonical form which then has a stable serialised representation and thus a stable hash.
The basic concept is that a file in the canonical JSON format can be read
using parseCanonicalJSON
. Note that this input file does not itself need
to be in canonical form, it just needs to be in the canonical JSON format.
Then the renderCanonicalJSON
function is used to render into the canonical
form. This is then the form that can be hashed or signed etc.
The prettyCanonicalJSON
is for convenience to render in a human readable
style, since the canoncal form eliminates unnecessary white space which
makes the output hard to read. This style is again suitable to read using
'parseCanonicalJSON'. So this is suitable to use for producing output that
has to be later hashed or otherwise checked.
See the API docs on Hackage.
This package has been extracted from the hackage-security package where
canonical JSON is used for all the signed TUF files, such as the
root keys file, etc. As you can see from that, canoncal JSON allows keeping
JSON files in a human readable pretty-printed form, and still allows verifying
signatures. In particular this demonstrates the use of embedded signatures,
where the root.json
both contains a body value and multiple signatures of
that body all within the same file. This is because canoncal JSON is about
hashes for JSON values, not serialised JSON text.
- Decoding/encoding Unicode code-points beyond
U+00ff
is currently broken