A intermediate representation for BSON, a binary encoding for JSON.

Decoding and encoding is (optionally) done as lazily as possible. First, the view function is able to unfold only one level and not the whole structure. Also, serialized versions are cached, so that later serializations of the same object are faster.

Notes:

1. Only JSON compatible BSON documents are supported. BSON extensions are not supported (int32, int64, timestamp, etc.). 2. Arrays in BSON are stored inefficiently. Prefer another binary format if you manipulate lots of arrays. 3. We differ from BSON to allow toplevel immediates. For this, we produce a document with only one byte indicating the kind of immediate followed by the immediate. The byte is 0x80 lor (the corresponding BSON field kind). 4. We differ from BSON to allow unambiguous toplevel arrays. As with (3), the subdocument to be decoded as an array is preceded with a 0x84.

Use the conforming flag to deactivates the extension from notes (3) and (4). In this case, the toplevel value must be an object.

A view over the bson representation.

Pre-instanciated Json_encoding.Make.

Pre-instanciated Json_encoding.Make.

Serializes the intermediate BSON representation to actual BSON.

By default, conforming is false, so that any value can be serialized, including immediates (see bson).

By default, cache is true, so a future serialization of the same data will be faster. The resulting bytes are stored in the value. You may want to turn this off if these values have a long lifespan, and that you care more about memory consumption than serialization speed.

Will raise Invalid_argument "Json_repr.bson_to_bytes" when conforming and trying to serialize a toplevel array or immediate.

Bson decoding error, with a message, the BSON and an offset.

Creates a lazily unfolded representation for some BSON. Because of the mutability of bytes and this laziness, set the copy parameter to true if you are not sure that the bytes will not be mutated in the future.

By default, conforming is false, so that any value can be serialized, including immediates (see bson).

By default, cache is true, so a future serialization of the same data will be faster. The input bytes are stored in the value. You may want to turn this off if these values have a long lifespan, and that you care more about memory consumption than serialization speed.

By default, laziness is true. If the data is a serialized object, it means that only the field names are read, the field values are eluded, and will be deserialized on demand when calling Repr.view. This implies that Bson_decoding_error may be raised later. If set to false, the whole structure is decoded upfront, so any decoding error will happen at this point. This may be preferable mostly when reading from untusted sources.

May raise Bson_decoding_error.