Record and Play Back your API
This records API responses so you can play them back later, quickly and reliably.
If you're a front-end developer this means you can develop without an API server--even without the internet.
You can also write tests for front end components that have some API dependencies but shouldn't break if the API server has hiccups.
It's good for:
- Testing (Responses are always the same, and fast)
- Nomads (Work on JS apps without internet)
- Unstable APIs (the VCR smooths out API downtimes)
- Throttled APIs (don't get shut down for querying too much while developing)
- Coordinating with frequent back end changes that require 30 minutes of API downtime to rebuild and deploy (ahem).
- Other solutions are focused on testing. That's great and valid, but I want to develop against something fast, deterministic, and reliable, too.
- This is written in Node, so it's easy to fit in with a front end developer's workflow.
- I store the responses as plain text JSON files so you can modify them whenever you want, and see very clearly where I create bugs or mismatch your expectations on API paths.
- You can fake out your own API by making a tree of
jsonfiles. No recording necessary. That could be pretty useful, huh?
- Supports multiple API servers. You just run multiple instances and it stores data in a folder tree by hostname and port.
This module is a dev tool with a command line interface, so it is installed globally.
npm install -g api-vcr
I'd actually prefer it to be local to a project, but there's no way to provide a nice CLI API in a locally installed node module.
You'd have to type
./node_modules/.bin/api-vcr all the time (plus args), and I don't want to do that to you.
This is actually why grunt-cli exists, I now understand.
Quick Start Guide
First let's launch in record mode so we can pass through requests and record the responses.
To run in record mode:
api-vcr http://jsonplaceholder.typicode.com/ --record
Now lets hit some endpoints so we can get a nice set of data.
In a browser, you can drop in these urls, or any from the jsonplaceholder api:
For each request, the VCR will create a single JSON file, using the request information to name the file.
For example, the request:
Is mapped to the file:
After making requests to localhost with the recorder running, you should see the log saying that files are being written, and you should be able to explore a tree of easy to see, editable JSON files. You can change these manually if you want. Be aware the recorder will overwrite them if you run it again later.
Go ahead and kill the server now.
Armed with recorded data, you're ready to run in offline mode.
Now any request you've previously recorded returns instantly from disk:
Similar requests you haven't recorded yet return their best guess:
Totally new requests 404:
You can specify a port for the vcr server to listen on. This is useful for running more than once instance. The port of the API server is used by default (unless that port is 80), this makes keeping proxies straight a little easier if you have remote APIs identified only by port number.
Set the port:
api-vcr http://myapi.com:8080 --port=1337
By default, data is stored wherever you run the command. If you always run the command from your project dir, this works well.
You can also set a hard-coded data path:
api-vcr http://myapi.com:8080 --data=~/sites/myApp/testData
By default the vcr looks for a sibling if it can't find a requested object. If you ask for
user/7, for example, it will return
user/42 if it has one.
This is awesome if you just want to keep working and it doesn't matter too much that it's the wrong user/product/sprocket/whatever.
Not everyone wants this behavior though. To turn it off:
api-vcr http://myapi.com:8080 --noSiblings
Data is all in the
./api-vcr-data folder local to where you run the command.
This is configurable, so you can store your data files with your project if you want to.
Folder and file names are used to determine the api path the VCR playback should respond to.
If you'd like the server to respond to
users/1, create this file:
You can record and modify, or seed from scratch, a complex tree of test data and distribute it in git with your project for consistent testing across machines.
Running the local version is the same except instead of
api-vcr, you use
./cli.js from this folder.
You can bump the version with npm's awesome
npm version patch && npm publish.
Start the app a
Logs all requests to console
Pass requests on to the recorded server
Create a directory structure that matches requests (namespace by server and port to support multiple APIs)
Store request responses as JSON files
Support missing objects (eg: if you have recorded
surfboard/3and they request
Print version on startup
Switch to a global install for easier run API by packages that depend on this one.
Bug: encode colons to
%3Aor something else in filenames
Bug: returning similar siblings returns deep nested JSON if there is a child folder (eg: posts/777 returns posts/1/comments.json)
Bug: returning siblings shouldn't return list objects for id requests and visa-versa
Support a config file, maybe in
./api-vcr-config.json(is there an option parsing library that supports fall-through from args to json file to defaults?)
Print the fully resolved path name for data, otherwise it's unclear where
Have a simple index page with list of all routes we know about, for easy debugging/transparency
Support query params
Option to trim large array responses to 20-100 max (default true)
Support response types other than JSON gracefully
Support POST, PUT, DELETE (at least don't error, and return a sensible object)