We give here a non-comprehensive list of related de/serialization projects. We indicate how they differ from Mapry and explain why we took pains to develop (and maintain!) our own tool instead of using an existing one.
Standard JSON libraries all support object trees, but not object graphs. Moreover, they do not support data based on a schema. While this is handy when the structure of your data is unknown at runtime, it makes code unnecessarily more difficult to maintain when the structure is indeed known in advance.
There is a large ecosystem around structured objects and their serialization based on property annotations (e.g., Rapidschema (C++), encoding/json (Go) or Jackson (Java)). While some of them support handling object graphs (usually through custom logic), we found the lack of polyglot support (and resulting maintenance effort required by synchronization of custom de/serialization rules across languages) a high barrier-to-usage.
Standard or widely used serialization libraries such as Boost.Serialization (C++), Gob (Go) or Pickle (Python) serialize object graphs out-of-the-box and handle impedance mismatch well. However, the representation of the serialized data is barely human-readable and difficult to get right in a polyglot setting due to a lack of common poly-language libraries (e.g., reading pickled data structures in C++). We deemed it a Herculean task to maintain the corresponding de/serializations accross different languages.
Popular serializers such as Protocol Buffers or Cap’n Proto support only object trees. If you need to work with cross-references in the serialized message, you need to dereference instances yourself. We found manual dereferencing in code to be error prone and lead to a substantial code bloat.
Flatbuffers handle object graphs natively, but exhibit a great deal of impedance mismatch through lack of maps and sophisticated data types such as date/time, duration etc.
JSON for Linking Data and JSON Graph are conventions to provide a systematic approach to modeling the object graphs in JSON. While these conventions look promising, we found the existing libraries lacking for production-ready code. On a marginal note, the JSON representations seem unnecessarily verbose when representing references.
JVM serializers presents a report on different object serializers running on top of Java Virtual Machine. The serializers are evaluated based on their run time and size.