Warning: The documentation is undergoing a rewrite.
Define an API once in Stone. Use backends, i.e. code generators, to translate your specification into objects and functions in the programming languages of your choice.
- Introduction
- Language Reference
- Builtin Backends
- Managing Specs
- Evolving a Spec
- Backend Reference
- JSON Serializer
- Network Protocol
Stone is made up of several components:
A language for writing API specifications, "specs" for short.
The CLI (stone) generates code based on the provided specs, backend,
and additional arguments.
There are builtin backends that come with Stone: Javascript, Python, Obj-C, Swift, and Typescript.
There are other backends we've written that aren't part of the Stone package because they aren't sufficiently general, and can't realistically be re-used for non-Dropbox APIs: Go and Java.
Stone includes a Python interface for defining new backends based on its intermediate representation of specs. This gives you the freedom to generate to any target.
Stone defines a JSON-compatible serialization scheme.
Stone was birthed at Dropbox at a time when it was becoming clear that API development needed to be scaled beyond a single team. The company was undergoing a large expansion in the number of product groups, and it wasn't scalable for the API team, which traditionally dealt with core file operations, to learn the intricacies of each product and build corresponding APIs.
Stone's chief goal is to decentralize API development and ownership at Dropbox. To be successful, it needed to do several things:
Decouple APIs from SDKS: Dropbox has first-party clients for our mobile apps, desktop client, and website. Each of these is implemented in a different language. Moreover, we wanted to continue providing SDKs to third-parties, over half of whom use our SDKs. It's untenable to ask product groups that build APIs to also implement these endpoints in a half-dozen different language-specific SDKs. Without decoupling, as was the case in our v1 API, the SDKs will inevitably fall behind. Our solution is to have our SDKs automatically generated.
Improve Visibility into our APIs: These days, APIs aren't just in the domain of engineering. Product managers, product specialists, partnerships, sales, and services groups all need to have clear and accurate specifications of our APIs. After all, APIs define Dropbox's data models and functionality. Before Stone, API design documents obseleted by changes during implementation were the source of truth.
Consistency and Predictability: Consistency ranging from documentation tense to API patterns are important for making an API predictable and therefore easier to use. We needed an easy way to make and enforce patterns.
JSON: To make consumption easier for third parties, we wanted our data types to map to JSON. For cases where serialization efficiency (space and time) are important, you can try using msgpack (alpha support available in the Python generator). It's possible also to define your own serialization scheme, but at that point, you may consider using something like Protobuf.
Stone is in active use for the Dropbox v2 API.
Stone makes no assumptions about the protocol layer being used to make API requests and return responses; its first use case is the Dropbox v2 API which operates over HTTP. Stone does not come with nor enforce any particular RPC framework.
Stone makes some assumptions about the data types supported in target programming languages. It's assumed that there is a capacity for representing dictionaries (unordered string keys -> value), lists, numeric types, and strings.
Stone assumes that a route (or API endpoint) can have its argument and result types defined without relation to each other. In other words, the type of response does not change based on the input to the endpoint. An exception to this rule is afforded for error responses.
Download or clone Stone, and run the following in its root directory:
$ sudo python setup.py install
This will install a script stone to your PATH that can be run from the
command line:
$ stone -h
If you choose not to install stone using the method above, you will need
to ensure that you have the Python packages ply and six, which can be
installed through pip:
$ pip install "ply>=3.4" "six>=1.3.0" "typing>=3.5.2"
If the stone package is in your PYTHONPATH, you can replace stone
with python -m stone.cli as follows:
$ python -m stone.cli -h
If you have the stone package on your machine, but did not install it or
add its location to your PYTHONPATH, you can use the following:
$ PYTHONPATH=path/to/stone python -m stone.cli -h