We're working hard to make *bitcore* the most powerful JavaScript library for working with bitcoin. Our goal is to have *bitcore* be a library that can be used by anyone interested in bitcoin, and to level expertise differences with great design and documentation.
In particular, you may notice that some names in this library are quite long for the average JavaScript user. That's because we prefer a long but comprehensible name than an abbreviation that might confuse new users.
Write a test for all your code. We encourage Test Driven Development so we know when our code is right. We have increased test coverage from 80% to around 95% and are targeting 100% as we move towards our 1.0 release.
Interfaces should accept as many types of arguments as possible, so there's no mental tax on using them: we want to avoid questions such as "should I use a string here or a buffer?", "what happens if I'm not sure if the type of this variable is an Address instance or a string with it encoded in base-58?" or "what kind of object will I receive after calling this function?".
Accept a wide variety of use cases and arguments, always return an internal form of an object. For example, the class `PublicKey` can accept strings or buffers with a DER encoded public key (either compressed or uncompressed), another PublicKey, a PrivateKey, or a Point, an instance of the `elliptic.js` library with the point in bitcoin's elliptic curve that represents the public key.
The design guidelines have quite a high abstraction level. These style guidelines are more concrete and easier to apply, and also more opinionated. The design guidelines mentioned above are the way we think about general software development and we believe they should be present in any software project.
A special case for [G2](#g2---general-internal-objects-should-be-instances) all network references must be `Network` instances (see `lib/network.js`), but when returned to the user, its `.name` property should be used.
`JSON.stringify` and `JSON.parse` are expected to be handled outside of the scope of Bitcore methods. For example, calling `JSON.stringify` on an Bitcore object will behave as expected and call `transaction.toJSON()` and then stringify it:
```javascript
var transactionString = JSON.stringify(transaction);
```
Likewise to instantiate a transaction from that string:
* Whenever a new class is created, add a generic error for that class in `lib/errors/spec.js`.
* Specific errors for that class should subclass that error. Take a look at the structure in `lib/errors/spec.js`, it should be clear how subclasses are generated from that file.
In order to deal with JavaScript's weak typing and confusing errors, we ask our code to fail as soon as possible when an unexpected input was provided.
There's a module called `util/preconditions`, loosely based on `preconditions.js`, based on `guava`, that we use for state and argument checking. It should be trivial to use. We recommend using it on all methods, in order to improve robustness and consistency.
Most classes have static methods named `fromBuffer`, `fromString`, `fromJSON`. Whenever one of those methods is provided, the constructor for that class should also be able to detect the type of the arguments and call the appropriate method.
Don't write long tests, write helper functions to make them be as short and concise as possible (they should take just a few lines each), and use good variable names.
This helps to make tests more useful as examples, and more independent of where they are placed. This also helps prevent forgetting to include all submodules in the bitcore object.
If possible, data for tests should be included in a JSON file in the `test/data` directory. This improves interoperability with other libraries and keeps tests cleaner.
All modules should include a developer guide and API reference. The API reference documentation is generated using JSDOC. Each function that exposes a public API should include a description, @return and @param, as appropriate. The general documentation guide for the module should be located in the `docs/guide` directory and is written in GitHub Flavored Markdown.
Our workflow is based on GitHub's pull requests. We use feature branches, prepended with: `test`, `feature`, `fix`, `refactor`, or `remove` according to the change the branch introduces. Some examples for such branches are:
Finally go to [github.com/bitpay/bitcore](https://github.com/bitpay/bitcore) in your web browser and issue a new pull request.
Main contributors will review your code and possibly ask for changes before your code is pulled in to the main repository. We'll check that all tests pass, review the coding style, and check for general code correctness. If everything is OK, we'll merge your pull request and your code will be part of bitcore.