Drydock
Sometimes, external dependencies can be as unreliable and unpredictable as the weather. That's why we built Drydock.
Drydock makes it simple to build stateful, configurable mocks. When finished, your mocks will be able to reproduce any and all expected interactions between your client application and its backend. This drastically simplifies end-to-end testing and offline UI development.
Stop waiting for the seas to calm - use Drydock!
Table of Contents
- Quick start
- Introduction: Get the most out of your mocks
- Creating a mock service
- Controlling your mock service
- Consuming your mock service
- Contributing
Quick start
For established projects
If you already have a working application (both server and client), we provide drydock-scaffold
to get you up and running quickly. This tool will take network traffic and transform it into mocks that are ready to use. It doesn't make any assumptions about state, so you can sprinkle that in as needed.
To get started, make a new directory and install the package:
mkdir mock && cd mock
npm init
npm install drydock-scaffold
While it is possible to install the tool globally, we recommend a non-saved npm install
whenever you start a new mock. This will ensure that the mock that is generated is compatible with the latest version of Drydock.
Once it is installed, it can operate in one of two modes (see below). Once your mocks have been written to disk, you can run them like node ./mock.js
. It is likely that you will want to make modifications, often to add statefulness.
NOTE: Proxies are unable to intercept and decode SSL-encrypted traffic, due to the fundamental constraints of the technology. Because of this, drydock-scaffold
will not be able to record and generate mock data for existing HTTPS endpoints. If possible, it is recommended to temporarily set HTTP endpoints for the initial data capture, if they are available. Otherwise, you will have to resort to other methods to capture the data.
Import mode
Import mode is the appropriate choice if your client is a single-page application.
First, run the Chrome web browser and open up the dev tools, selecting the Network tab. Then, enter the URL of your single-page application and proceed to use the application as your normally would.
After you have completed the full set of typical actions in your application, right-click any of the transactions in the Network tab, and select Save as HAR with Content
.
Now, run drydock-scaffold
like so:
./node_modules/.bin/drydock-scaffold import ./path/to/saved-file.har
drydock-scaffold
will take the captured network data and save several files to the current working directory, including a mock.js
and captured fixture data.
Proxy mode
Proxy mode is the appropriate choice if your client is not a single-page application. To use, run drydock-scaffold
from your new mock directory:
./node_modules/.bin/drydock-scaffold --ip 0.0.0.0 --port 1337 --destination ./
Now, configure your application to use an HTTP proxy at IP 0.0.0.0
and port 1337
. Proceed to use your application as you normally would. You will see the tool indicate in the terminal the HTTP transactions that it has recorded.
When you have completed the full set of typical actions in your application, press CTRL-C
once (or twice to quit immediately), and several files will be saved to the current working directory. This will include a mock.js
and captured fixture data.
For new projects
If you are just starting development of a new project, there won't be any existing services that drydock-scaffold
will be able to mimic. To get a quick start, you can either author the mock using API documentation, or take a look at the example project in the example/
directory of this repo.
Introduction: Get the most out of your mocks
There are a lot of solutions out there for standing up mock APIs, and each one takes a slightly different tack.
The simplest mocks return static responses for a given endpoint, with no variation. These mocks are easy to create but too rigid for dynamic applications. The most complex mocks duplicate the logic of an actual server application. These are as flexible as you need them to be, but they typically require a lot of work for a moderate gain.
Drydock is optimized for a middle approach. In practice, this approach has resulted in mocks that are flexible enough to back some very complex frontend applications. But they've been reasonably quick to create and maintain.
More specifically, Drydock mocks generally don't really care what the client sends to them. This may seem counterintuitive. However, in most cases, the goal for our mock service is to deterministically influence the behavior of the client. If that is the case, all we should care about is what data is sent back to the client in a response. Variations in those responses are enumerated as simple handlers, and can be selected and manipulated through Drydock's control panel or the JavaScript API.
For illustration purposes, lets look at a simple sign-in service. In the simplest case, the frontend application sends a username and password that the user has entered, and the server responds with a session ID. This is very straightforward, and could be handled by a static mock.
However, in our hypothetical, there are a number of ways that a sign-in attempt could fail. The user could have entered a wrong password. Or perhaps the username doesn't exist. Or, maybe the database is having an unexpected problem. Or the account may be temporarily locked.
To simulate these conditions, we could select a response based on the content of the request - garfield@a.cat
would result in a 200
, odie@a.cat
would result in a 400
, and so on. Drydock does support this strategy.
However, these special values can be hard to remember and hard to keep in-sync. Furthermore, the user who is interacting with the client (this could be a real user, or an automated test) does not always have direct control over the content of a request. And often, the schema for a given request type may not have enough variations to accommodate the number of possible responses (the request payload could only include a single boolean value, for example).
Instead, Drydock allows you to enumerate all the possible responses that a request to a particular endpoint could result in. Then, using the control panel or the JavaScript API, the response can be precisely controlled. For cases where it is desired that the result of one request impacts a later request, Drydock exposes a shared-state object to all of the handlers that are invoked. This should not be overused, but can result in a more natural experience for your user.
Creating a mock service
High-level
The process for writing and consuming drydock mocks will, of course, vary in each situation. But the general pattern described will usually look something like this:
- You determine the scope of your mocks by enumerating all of your endpoints. For example, if the user of your application is able to sign-in somehow, there will likely be something akin to a
POST /api/sign-in
endpoint. You can always add more endpoints later, so just aim for the minimum required for your application's "happy path". - Next, determine some or all of the variations for each of those endpoints. In the case of our login example, a
POST
might result in:200 successful login
,400 wrong password
,400 user not found
, etc. - Create a new
.js
file containing the Drydock boilerplate (provided below). - Then, using the documentation provided, author endpoints with their variations.
- Configure your application to point to the mock server. There are some tips for how to go about this at the end of the README.
Boilerplate
The following is the bare minimum you will need to stand up a new Drydock mock. Of course, it won't be of much use to you until you add endpoints!
var Drydock = ;var drydock = initialState: // Anything you put here will be accessible as // `this.state` within your handlers. ; // Define your routes here. drydockstart;
Options
When creating a new Drydock
instance, the following options can be provided. There are no manditory options, as each has a default value.
port
integer - the port that your mock server should listen on. Defaults to1337
if not provided.ip
string - the IP that you mock server should bind to. Defaults to0.0.0.0
(all interfaces) if not provided.verbose
boolean - indicates whether the mock server should output HTTP transaction data and additional logging to the console. Defaults tofalse
if not provided.cors
boolean - indicates whether the mock server should respond to CORS requests. Defaults tofalse
if not provided.cookieEncoding
string - indicates the transformation between raw cookie data and how it is transferred to the client. See theencoding
in Hapi's state management documentation to learn about the available options. Defaults tonone
if not provided.proxyUndefined
boolean - indicates whether requests for unknown routes should be forwarded to the host specified in the HTTP request. This requires that the client is configured to use the Drydocks server as a proxy.caseInsensitive
boolean - indicates whether or not routes will be case sensitive or not. Defaults to false.
Authoring endpoints
Once you have your Drydock
instance, you'll need to define some routes.
JSON routes
If your endpoint accepts and/or replys with JSON payloads, you'll want to define a JSON route. It will look something like this:
drydock;
Each handler function takes a request object. This object will have the following properties:
payload
- the body of the HTTP request.params
- the parameters in the requested URL (see below).query
- the dictionary object equivalent of the query string.
Path values support parameters. In our example above, if the user whose name was to be set had a unique user ID, the path might have been /api/user/{userId}/name
. When processing a request for a URL that matches this pattern, the params
property of the request
object would look something like this:
Drydock routes return status code 200 by default, but it does support specifying what status code the response should have. See an example in the HTML route below. Currently, the headers
option only support status codes.
HTML routes
Routes that return HTML will function similarly to JSON routes, and can be defined using the Drydock.prototype.htmlRoute
method. Here's an example:
drydock;
Proxy routes
If you'd like to define routes that should be forwarded to a specific URL on an existing server, you can define a proxyRoute
. This is useful if the API you are mocking relies on another service, or if you'd like to enumerate all routes that you plan to mock, but implement those mocks incrementally.
Here's an example:
drydock;
If your mock server is running on localhost
port 1337
, opening http://localhost:1337/google
in your browser will result in the HTML from www.google.com
.
forwardTo
can alternatively take a function to dynamically determine the url.
For example:
drydock;
State
The above example shows us now to make our mocks stateful. When invoked, each handler will have access to this.state
, a simple object that is share across your mock instance. In this way, you can persist data from across endpoints and requests.
To define an initial state, you can pass an initialState
object as an option to the Drydock
constructor.
Configurability via Handler Options
The above example also demonstrates how to define multiple handlers for a single endpoint. In the original use-case that brought about Drydock, there were close to hundred possible responses (many of them errors) for some of our endpoints. In this way, multiple handlers can mimic the full breadth of an endpoints behavior.
However, sometimes even with unlimited handlers, they can still be too static. What is needed is not breadth, but depth of configurability for particular handlers.
This need is met through the use of handler options. Here's the same example route as the one above, except more configurability is added to its handler through options.
drydock;
Although a bit pedantic, this example illustrates how adding configurability to a handler can provide your mock increased configurability without bloating your endpoint with hundreds of handlers.
Select one
A handler's options comes in two varieties. The first is selectOne
. In this mode, whatever option has been selected will be passed as a second parameter to the invoked handler.
To use, the following properties should be set on your handler:
optionsHelperText
- This text will be displayed in the control panel, and will help users understand what they're configuring.optionsType
- This should be set toselectOne
.options
- This is object where the keys are the options displayed in the control panel, and the values are the objects that will be passed to the handler when the option is selected.
In the example above, if the user selects Mr.
in the control panel when prompted with What title should precede the name?, then the string "Mr. "
will be passed to the handler when the route is hit. Similarly, if they select None
, the string ""
will be passed to the handler when the route is hit.
The values in the options
object need not be strings - they can be an object, even functions (which provides a significant avenue for handler flexibility).
Select many
The second variety is selectMany
. This functions identically to selectOne
, with a couple of exceptions. The first may be obvious; optionsType
must be set to selectMany
.
The other exception is that, when the handler is invoked, it will receive as its second argument not a single selected value, but an array of selected values.
For example, we might want to select suffixes for a person's name. It is possible for a person to be both a Jr.
and an M.D.
. This is a time we would use selectMany
instead of selectOne
. When invoked, the handler would receive as its second argument something like [" Jr.", " M.D"]
.
Cookies
You can write or read to the cookie state via this.cookies
in your handler.
Earlier, we saw this when this.cookies.nameIsSet
was set to "true"
. You'll notice that it was not set to a boolean, but to a string. This is because cookies take the form of strings. If you need to read or write more complex data to a cookie, you'll need to (de)serialize your data using JSON.stringify
and JSON.parse
.
Headers
It is also possible to directly set a response header. To do so, set a property of this.headers
, like so:
thisheadersMY_HEADER = "SOME STRING VALUE";
Serving files
Drydock is capable of serving up assets to a frontend application.
Single file
To serve a single file, use the Drydock.prototype.staticFile
method. You'll need to provide both a filePath
and a urlPath
. The filePath
must be an absolute path to an existing file on the file system. The urlPath
should look something like /path/to/my/file.jpg
.
Directory
Serving a directory works similarly. filePath
must be an absolute path to an existing directory on the file system. urlPath
should look something like /path/to/my/dir/
.
Controlling your mock service
Via the control panel
After you've created some endpoints, you'll need some way to configure them. Once running, a Drydock mock has two primary interfaces: the browser control panel and the JavaScript API.
The control panel can be accessed via the /drydock/
route. For example, if your mock is running on 127.0.0.1
and port 1337
, the control panel can be accessed via http://127.0.0.1:1337/drydock/.
In the control panel, you will see each route that was defined. If you click on a route, you will see all the possible ways that route can behave - these are the handlers that defined earlier. Some handlers will be configurable, if you have defined options for that handler.
Via the JavaScript API
TODO
Consuming your mock service
It isn't enough to create a mock API, you application has to consume it instead of the real service.
Single-page application
If your client is a SPA, here are a few ways you might get it to point to your new mock.
- Have Drydock render the HTML that bootstraps your SPA. The template for this HTML can take a value for the desired API endpoint.
- Configure your SPA to point to the mock API URL when a particular URL query string is present for your SPA.
- Set your browser's proxy to point to the mock server. This will cause problems with browsing if you don't revert the change.
- Use a browser plugin to switch your proxy, like Proxy SwitchyOmega, and only enable the proxy when you are testing your application.
Node
Generally, the easiest way to go about this is to check environment variables for the run environment. Configure your application to use the mock URL when it is provided as an environment variable. For example, you might execute your node application like so:
API_BASE=http://127.0.0.1:1337 node ./path/to/my/app.js
Java
Java is a little trickier, but you can define proxy settings in the JAVA_OPTS
environment variable. It might look something like this:
export JAVA_OPTS="-XX:MaxPermSize=1024m -Dhttp.proxyHost=127.0.0.1 -Dhttp.proxyPort=1337 -Dhttps.proxyHost=127.0.0.1 -Dhttps.proxyPort=1337"
Contributing
If you'd like to see something changed or added to Drydock, please open an issue or PR. Make sure npm run test
passes before you submit your PR.
Note: If you're making changes and testing Drydock on your local machine, you'll need to run npm run build
so that the ES6 source is compiled to ES5 in lib/
.