twittergo

This project implements a Go client library for the Twitter APIs. This library supports version 1.1 of Twitter's API and application-only auth.

The goal of this project is to provide a thin veneer over my oauth1a library in order to simplify access to Twitter's APIs from Go. Where possible, I've tried to defer to native objects (use of http requests for example). Additionally, responses could be parsed directly as JSON, but some wrapper types have been defined in order to provide some convenience methods for accessing data.

Installing

Run

go get github.com/kurrik/twittergo

Include in your source:

import "github.com/kurrik/twittergo"

Godoc

See http://godoc.org/github.com/kurrik/twittergo

Using

I have moved all of the examples to the https://github.com/kurrik/twittergo-examples project in order to make this library easier to import. Please reference that project for ways to address specific use cases.

The simplest example in the twittergo-examples project is probably verify_credentials. This calls an endpoint which will return the current user if the request is signed correctly.

The example starts by loading credentials, which can be done in many ways. The example implements a LoadCredentials which looks for the CREDENTIALS file mentioned above:

var (
    err    error
    client *twittergo.Client
    req    *http.Request
    resp   *twittergo.APIResponse
    user   *twittergo.User
)
client, err = LoadCredentials()
if err != nil {
    fmt.Printf("Could not parse CREDENTIALS file: %v\n", err)
    os.Exit(1)
}

Then, a standard http request is created to a /1.1/ endpoint:

req, err = http.NewRequest("GET", "/1.1/account/verify_credentials.json", nil)
if err != nil {
    fmt.Printf("Could not parse request: %v\n", err)
    os.Exit(1)
}

The client object handles sending the request:

resp, err = client.SendRequest(req)
if err != nil {
    fmt.Printf("Could not send request: %v\n", err)
    os.Exit(1)
}

The response object has some convenience methods for checking rate limits, etc:

if resp.HasRateLimit() {
    fmt.Printf("Rate limit:           %v\n", resp.RateLimit())
    fmt.Printf("Rate limit remaining: %v\n", resp.RateLimitRemaining())
    fmt.Printf("Rate limit reset:     %v\n", resp.RateLimitReset())
} else {
    fmt.Printf("Could not parse rate limit from response.\n")
}

Finally, if the response format is known, the library provides some standard objects which make parsing response data easier:

user = &twittergo.User{}
err = resp.Parse(user)
if err != nil {
    fmt.Printf("Problem parsing response: %v\n", err)
    os.Exit(1)
}
fmt.Printf("ID:                   %v\n", user.Id())
fmt.Printf("Name:                 %v\n", user.Name())

Error handling

Errors are returned by most methods as is Golang convention. However, these errors may sometimes be cast into twittergo.ResponseError, twittergo.Errors or twittergo.RateLimitError structs which will provide additional information.

To check for rate limiting or other types of server errors, attempt to cast any errors returned by the APIResponse.Parse method.

resp, err = client.SendRequest(req)
if err != nil {
    fmt.Printf("Could not send request: %v\n", err)
    os.Exit(1)
}
tweet = &twittergo.Tweet{}
err = resp.Parse(tweet)
if err != nil {
    if rle, ok := err.(twittergo.RateLimitError); ok {
        fmt.Printf("Rate limited, reset at %v\n", rle.Reset)
    } else if errs, ok := err.(twittergo.Errors); ok {
        for i, val := range errs.Errors() {
            fmt.Printf("Error #%v - ", i + 1)
            fmt.Printf("Code: %v ", val.Code())
            fmt.Printf("Msg: %v\n", val.Message())
        }
    } else {
        fmt.Printf("Problem parsing response: %v\n", err)
    }
    os.Exit(1)
}

The previous snippet would print the following if a user attempted to Tweet the same text twice in a row:

Error #1 - Code: 187 Msg: Status is a duplicate

Rate limit errors are pretty easy to use. They're a simple struct containing what the limit for the request was, how many were remaining (should be 0) and when the limiting resets:

type RateLimitError struct {
    Limit     uint32
    Remaining uint32
    Reset     time.Time
}

The Errors type is a little more complicated, as it may return one or more server side errors. It is possible to cast one to a string using the standard Error method, but if you need to handle individual errors, iterate over the slice returned by Errors (plural) instead:

for i, val := range errs.Errors() {
    fmt.Printf("Error #%v - ", i + 1)
    fmt.Printf("Code: %v ", val.Code())
    fmt.Printf("Msg: %v\n", val.Message())
}

Each of those errors has a Code and a Message method, which return values and strings corresponding to those listed in the "Error codes" section of this page: https://dev.twitter.com/docs/error-codes-responses

Application-only auth

If no user credentials are set, then the library falls back to attempting to authenticate with application-only auth, as described here: https://dev.twitter.com/docs/auth/application-only-auth

If you want to obtain an access token for later use, create a client with no user credentials.

config := &oauth1a.ClientConfig{
    ConsumerKey:    "consumer_key",
    ConsumerSecret: "consumer_secret",
}
client = twittergo.NewClient(config, nil)
if err := c.FetchAppToken(); err != nil {
    // Handle error ...
}
token := c.GetAppToken()
// ... Save token in data store

To restore a previously obtained token, just call SetAppToken():

// Get token from data store ...
c.SetAppToken(token)

Saving and restoring the token isn't necessary if you keep the client in memory, though. If you just create a client without any user credentials, calls to SendRequest will automatically fetch and persist the app token in memory. See search_app_auth/main.go for an example of this.

Google App Engine

This library works with Google App Engine's Go runtime but requires slight modification to fall back on the urlfetch package for http transport.

After creating a Client, replace its HttpClient with an instance of urlfetch.Client:

var (
    r      *http.Request
    config *oauth1a.ClientConfig
    user   *oauth1a.UserConfig
)
...
ctx = appengine.NewContext(r)
c = twittergo.NewClient(config, user)
c.HttpClient = urlfetch.Client(ctx)

For a comprehensive example, see user_timeline_appengine

Custom models

The twittergo library comes with some standard models for structures like Tweets and Timelines, but you are not required to use them. Pass any struct which will deserialize from a Twitter API response to APIResponse.Parse:

type CustomTweet struct {
    CustomID   string `json:"id_str"`
    CustomText string `json:"text"`
}
...

req, err = http.NewRequest("GET", url, nil)
...
resp, err = client.SendRequest(req)
customTweet = &CustomTweet{}
err = resp.Parse(customTweet)
...

For complete code, see the custom_struct example.

Debugging

To see what requests are being issued by the library, set up an HTTP proxy such as Charles Proxy and then set the following environment variable:

export HTTP_PROXY=http://localhost:8888

Because Go will reject HTTPS requests through a proxy, you'll need to set the following for any HTTPS endpoints:

export TLS_INSECURE=1

Make sure not to use this in production!