Speak Instapaper Posts — Part 2

In the last part we built a useful workflow that would open a given number of unread article from your Instapaper feed. But we stopped short of the goal, to convert the text of the articles to speech files.

If you look into the library of Automator actions there is one with the promising name “Get Text from Webpage.” However this will extract all the text, usually including all the menus, ads and all the other detritus that clutters webpages these days. The latest version of Safari (( Safari 5, as I write this )) has a functionality called “Reader,” which removes all this clutter and allows the user to focus on just the text. Unfortunately, the “Reader” functionality in Safari is not scriptable.

But before Safari had “Reader” there was the Readability javascriptlet from Arclab90 which does very much the same thing. Since Safari’s AppleScript dictionary allows us to execute arbitrary JavaScript against a webpage, we can use that to extract the relevant text from the article. That saves us from having to recreate the logic of the Readabilty scriptlet in AppleScript.

Do the following with the workflow we built in Part 1:

  • duplicate the Workflow file and name the copy: Speak Instapaper Articles to iTunes
  • remove the last action “New Safari Documents” from the workflow (( there is a bug in Safari’s AppleScript implementation where document references from freshly created web documents will go stale once the page is loaded. This also affects the “New Safari Documents” action. We will work around this bug in our AppleScript))
  • add a new empty “Run AppleScript” action at the end of the workflow and enter the following code:
on run {input, parameters}
	
	-- uses the 'Readability' javascript from
	-- http://lab.arc90.com/experiments/readability/
	
	set readabilityScript to "javascript:(function(){readConvertLinksToFootnotes=false;readStyle='style-newspaper';readSize='size-medium';readMargin='margin-medium';_readability_script=document.createElement('script');_readability_script.type='text/javascript';_readability_script.src='http://lab.arc90.com/experiments/readability/js/readability.js?x='+(Math.random());document.documentElement.appendChild(_readability_script);_readability_css=document.createElement('link');_readability_css.rel='stylesheet';_readability_css.href='http://lab.arc90.com/experiments/readability/css/readability.css';_readability_css.type='text/css';_readability_css.media='all';document.documentElement.appendChild(_readability_css);_readability_print_css=document.createElement('link');_readability_print_css.rel='stylesheet';_readability_print_css.href='http://lab.arc90.com/experiments/readability/css/readability-print.css';_readability_print_css.media='print';_readability_print_css.type='text/css';document.getElementsByTagName('head')[0].appendChild(_readability_print_css);})();"
	
	set output to {}
	tell application "Safari"
		repeat with x in input
			set theURL to contents of x
			make new document with properties {URL:theURL}
			delay 0.5
			
			repeat until ( (do JavaScript "document.readyState;" in document of window 1) is equal to "complete")
				delay 0.5
			end repeat
			set d to document of window 1
			
			do JavaScript readabilityScript in d
			delay 3
			repeat until ( (do JavaScript "document.readyState;" in d) is equal to "complete")
				delay 1
			end repeat
			
			set thetext to text of d
			-- remove first three and last four paragraphs since these are Readability links
			set AppleScript's text item delimiters to return
			set thetext to (paragraphs 4 through -5 of thetext) as text
			close d
			
			set output to output & {thetext}
		end repeat
	end tell
	
	return output
end run

Let’s slowly go through this code:

  • first we setup variable to store the Readabilty javascript code.
  • then we initialize a list output to store the results.
  • then we loop through the items that were passed into the action in the input variable. In this case the items are the URLs of the Instapaper posts.
  • set theURL to contents of x
    this de-references the iterator variable. Due to some oddities of the AppleScript language this is usually a wise thing to do in a repeat loop.
  • make new document with properties {URL:theURL}
    delay 0.5

    we tell Safari to open a new document with the given URL and pause for a while to let Safari start loading

  • repeat until ( (do JavaScript "document.readyState;" in document of window 1) is equal to "complete")
        delay 0.5
    end repeat
    set d to document of window 1

    We have to wait until the page is completely loaded before we can apply the Readability script against the page. Unfortunately Safari does not expose the state of the page (loading or complete) to AppleScript. This is however exposed to the JavaScript DOM within the page and we can access DOM information from AppleScript with the do Javascript event. So we poll the document.readyState attribute in Javascript until it reports complete. Then we remember a reference to this document in a variable. ((Safari has a bug where a AppleScript reference to document will change while it is loading, resulting in broken references. All this is a clumsy, but effective workaround.))

  • now we can execute the Readability script against the page:
    do JavaScript readabilityScript in d
    delay 3
    repeat until ( (do JavaScript "document.readyState;" in d) is equal to "complete")
    	delay 1
    end repeat

    We use the same DOM trick to wait until Safari is done.

  • Now the text property of the document contains the cleaned up text of the article. We can extract that, remove some extra lines that Readabilty inserts, close the Safari window and append the text as its own element to the output list.
    set thetext to text of d
    -- remove first three and last four paragraphs since these are Readability links
    set AppleScript's text item delimiters to return
    set thetext to (paragraphs 4 through -5 of thetext) as text
    close d
    			
    set output to output & {thetext}

This would be a good time to save the workflow, and do a test run. You can show the results of the workflow in Automator to see if the text is extracted properly. Readability is not perfect and does not work on all pages, but the success rate is quite high.

The remaining work of converting the text into audio is very straightforward. Add the following workflow actions:

  • Text to Audio File
  • Import File into iTunes
  • Add Songs to Playlist ((You want to create a specific playlist for these files in iTunes))

And then you are done. You can also download the complete Workflow.

Speak Instapaper Posts — Part 1

I saw this in my Twitter stream the other day:

You know what I want? A text-to-speech plugin for @instapaper so, while commuting to/from work, I can listen to the stuff I find at work.

— @seankaiser

That shouldn’t be too hard, shouldn’t it?

First we have to get the unread articles from Instapaper. If you go instapaper, log in, and go to your unread articles, you can see the RSS button in the URL field in Safari. To get to the RSS feed in Automator, do the following:

  • Open Automator, create a new Workflow
  • Open www.instapaper.com/u in Safari and drag the link from the URL bar to the empty Automator Workflow window. Automator will create an new “Get Specified URLs” action with the Instapaper URL unread in it
  • Add a “Get Feeds from URLs” action next
  • Add a “Get Link URLs from Articles” action. Unselect the “only in the same domain” option
  • Finally add the “New Safari Documents” action.
  • The workflow already does something useful. Save as “Open unread Instapaper articles”

If you ares anything like me this workflow will open quite a large number of pages. I think Instapaper limits the RSS feed to 25. That’s still a lot of new Safari tabs/windows you are opening there. We want to add an action that restricts the number of items passed through it. Surprisingly there is none in the default actions, but this is fairly easy to add. Insert a new “Run AppleScript” action before the “New Safari Documents” action and replace the default code with the following:

on run {input, parameters}
	
	set maxNum to 3
	-- filters all but the first maxNum items from the articles, change as appropriate
	-- enter '-1' or remove this action entirely to get all urls
	if (count of input) > maxNum then
		set output to items 1 through maxNum of input
	else
		set output to input
	end if
	
	return output
end run

This will only pass through the first maxNum of items passed into it, regardless of type. You can change maxNum to fit your taste and/or needs. You can also set maxNum to -1 if you want to pass all items without removing the AppleScript action.

Save again and try running it. The next step will be to filter the actual text out of the web page which will be a little tougher and the main topic of Part 2.