Twitter Changing Their API Again and Parsing Twitter Tweet Responses

Twitter Changing Their API Again and Parsing Twitter Tweet Responses

If you haven't noticed by now you should do the next time you try and use Twitter on the twitter.com website.

Not only are they re-working their API they are cracking down on the sending of links in DM messages.

I personally have had an account suspended within the last week then re-activated 3 days later and I still have had no explanation why this happened from their support team.

Also this happened directly after I was in contact with them about the problems they were having falsely identifying double encoded links as phishing or spam and directing users of my account to a "Twitter thinks this page might be dangerous" page!

Even though their development team claimed the links I was using were okay and safe if you clicked them you would be redirected to this page by mistake.

I think the problem was due to the fact that to get a link into 140 chars you have to first encode it with a link shortener like bit.ly and then Twitter always encode ALL links with their own t.co shortener so they can track the clicks.

I reckon they were having issues identifying the final destination URL and were therefore flagging double encoded links as spam/phishing/hacks etc - despite their claims they weren't.

However now it seems Twitter has changed their DM API. Not only have I found that it constantly refreshes as I am typing causing me to lose my message but if I receive a message as I type it also refreshes causing the same lost wording.

It seems more like a badly written interactive Messenger service now with an AJAX timer to reload new content than the good old DM service it used to be. They still maybe working on it so hopefully these bugs will get fixed soon.

A few people have complained to me they cannot now send links in D or M messages (Direct Messages) and I have found the same problem. It seems Twitter are cracking down on companies that log people who follow and unfollow you as well as treating the majority of links in DM's as spam, whether they are or not.

Anyway bitching time is now over.

What I wanted to write about was that I had to release new versions of my WordPress plugins Strictly AutoTags and Strictly TweetBot the other day.

I was finding on sites with a lot of posts that the Tweets from the TweetBot (which are hooked into the onpublish action/event) were going out before tagging could be completed. Therefore only the default #HashTags were being used.

Therefore if you wanted to use categories or post tags as your #HashTags in the Tweet they were not available at the point of Tweeting.

Therefore I changed the Strictly AutoTags plugin to fire an event once tagging had been completed in the SaveAutoTags method which enabled Strictly TweetBot to run then instead of on publish. I also pass in the Post ID so the event listener knows which post to send Tweets for e.g:

do_action('finished_doing_tagging', $object->ID);

The code in the TweetBot automatically checks for the existence of the AutoTag plugin by looking for a function that I use within the plugin that will only exist if the plugin is active.

If this is found the TweetBot hooks the listener to fire off Tweets into this event instead of the standard on publish event.

As I was explaining to a WordPress developer the other day, I prefer the use of the terms events and listeners due to my event driven coding in JS or C# etc whereas hooks and actions are more WordPress driven.

However as I was re-writing this code I also noticed that in the admin panel of Strictly TweetBots which shows a history of the last Tweets sent and any error messages returned from Twitter that I was not getting back my usual "Duplicate Tweet" error when I tried sending the same tweet multiple times.

By the looks of things the JSON response from Twitter when using OAuth to send a Tweet has changed slightly (when I don't know) and now I have to parse the object and containing array for any error messages rather than checking just for a string as I used to be able to do.

If anyone is interested, or works with OAuth to send Twitter messages, the code to parse the JSON response to collect error messages is below.

Please notice the ShowTweetBotDebug debug function which outputs arrays, objects and strings to the page if needed.

/*
 * My Show Debug Function
 *
 * @param $msg string, object or array
 */
function ShowTweetBotDebug($msg)
{ 
 if(!empty($msg))
 {
  if(is_array($msg)){
   print_r($msg);
   echo "<br />";
  }else if(is_object($msg)){
   var_dump($msg);
   echo "<br />";
  }else if(is_string($msg)){
   echo htmlspecialchars($msg) . "<br>";
  }
 } 
}



$twitterpost = "Some tweet about something";

ShowTweetBotDebug("post this exact tweet = $twitterpost");

// Post our tweet - you should have set up your oauth object before hand using the standard classes
$res = $oauth->post(
 'statuses/update', 
 array(
  'status' => $twitterpost,
  'source' => 'Strictly Tweetbot'
 )
);



ShowTweetBotDebug("response from Twitter is below");

//output errors using vardump to show whole object
ShowTweetBotDebug($res);

// parse error messages from Twitter
if($res){
 // do we have any errors?
 if(isset($res->errors)){
  
  ShowTweetBotDebug("we have errors");

  // if we do then it will be inside an array
  if(is_array($res->errors)){

   ShowTweetBotDebug("we have an array of errors to parse");

   foreach($res->errors as $err){     
    
    ShowTweetBotDebug("Error = " . $err->message);        
   }

  }
 }else{
  ShowTweetBotDebug("tweet sent ok");  

 }
}else{
 ShowTweetBotDebug("Could not obtain a response from Twitter");
}

This code might come in handy to other people who need to parse Twitter responses and just remember - Twitter is ALWAYS changing their API which is a real pain the ass to any developers using their API or trying to make a business model from it.

It seems a lot of companies have had to shut down due to the recent changes to Direct Messages so who knows what they will do in the future?

Handle jQuery requests when you want to reference code that hasn't loaded yet

Handle jQuery requests when you want to reference code that hasn't loaded yet

As you should be aware it is best practise to load your JavaScripts at the bottom of your HTML for performance and to stop blocking or slow load times.

However sometimes you may want to reference an object that is not yet loaded higher up in the page.

If you are using a CMS or code that you cannot change then you may not be able to add your event handlers below any scripts that maybe needed to use them. This can cause errors such as:

Uncaught ReferenceError: $ is not defined 
Uncaught ReferenceError: jQuery is not defined

If you cannot move your code below where the script is loaded then you can make use of a little PageLoader function that you can pass any functions to and which will hold them until jQuery (or any other object) is loaded before running the functions.

A simple implementation of this would involve a setTimeout call that constantly polls the check function until your script has loaded.

For example:

PageLoader = { 
 
 // holds the callback function to run once jQuery has loaded if you are loading jQuery in the footer and your code is above
 jQueryOnLoad : function(){}, 

 // call this function with your onload function as the parameter
 CheckJQuery : function(func){
  var f = false;

  // has jQuery loaded yet?
  if(window.jQuery){
   f=true;
  }
  // if not we store the function if first time in loop otherwise and set a timeout
  if(!f){
   // if we have a function store it until jQuery has loaded
   if(typeof(func)=="function"){    
    PageLoader.jQueryOnLoad = func;
   }
   // keep looping until jQuery is in the DOM
   setTimeout(PageLoader.CheckJQuery,200);
  }else{
   // jQuery has loaded so call the function
   PageLoader.jQueryOnLoad.call();    
  }
 }
}

As you can see the object just holds the function passed to it in memory until jQuery has been loaded in the DOM. This will be apparent because window.jQuery will be true.

If the object isn't in the DOM yet then it just uses a setTimeout call to poll the function until it has loaded.

You could increase the length of time between the polls or even have a maximum limit so that it doesn't poll for ever and instead after 10 loops returns an error to the console. However this is just a simple example.

You would call the function by passing your jQuery referencing function to the CheckJQuery function like so.

<script>
PageLoader.CheckJQuery(function(){
 $("#myelement").bind('click', function(e) { 
  // some code
  alert("hello");
 });
});
</script>

It's just a simple way to overcome a common problem where you cannot move your code about due to system limitations but require access to an object that will be loaded later on.