It's game time: Using Google Docs as a CMS
The whistle sounds, the kick is up and, just like that, football season is upon us. Most newspapers throughout the years produced some kind of high school football preview, which pretty perfectly meets the sweet spot of subscriber interest coupled with advertising dollars. Moving that over to the digital realm has been a bit more difficult, at least for us.
Our (corporately homegrown) CMS doesn't really do well with one-off tabs short of creating a brand-new section, so previously the only items making the jump from print to digital were the tab stories, as stories. Last year we changed that trend with an iPad-only app we produced using Adobe's Digital Publishing Suite.
With help from a corporate deal, we wanted to explore the ways that an app could help us present our content. At the time of creation, there were options for more device-agnostic profiles, but the way the DPS deal was set up we could produce the iPad app for free; anything else incurred a per-download charge (being a free download, we weren't ready to lose money on the basis of popularity). We were all pretty happy with the way the product turned out, but were disappointed by the limitations. The iPad-only specification severely limited its potential audience, and the fact that none of it was indexable or easily importable made it feel more like producing an interactive PDF than a true digital product. Though we were satisfied with the app, we determined in the future we'd likely steer clear of the app-only route.
When we decided we wanted to do the preview again for this year, everyone was in favor of going with a responsive design — it allowed for the maximum possible audience as well as the smallest amount of work to hit said audience. The only problem was that our CMS doesn't support responsive design, so we'd have to go around it.
This problem was compounded when we decided on the scope of the project. Our high school football coverage is run by GameTimePA, which consists of the sports journalists from the York Daily Record, Hanover Evening Sun, Chambersburg Public Opinion and Lebanon Daily News. The four newsrooms are considered a "cluster," which means that we're relatively close geographically and tend to work together. Since the last preview, however, GameTimePA had expanded to include our corporate siblings in the Philadelphia area, meaning we now encompassed something like 10 newsrooms stretching from Central Pennsylvania to the New Jersey border.
And we're all on different CMSes.
One of the few
commonalities we do share are Google corporate accounts. Though our
corporate policy does not allow for publishing to the web or sharing
publicly (another rant for another time), it at least gives us an
authentication system to work with.
now, there's a fairly defined set of content that goes into the tab.
There are league-specific items (preview, review, players to watch) and
team-specific ones (story, photo, writeup, etc.). S
tarting to look like a data table to you yet? By the time we finished, we actually ended up with some fairly robust sheets/tables for things that would generally fall under the category of "administration." But the content was only half the problem. Translating it into the final product still loomed ahead of us. Because we only have one server we can use, ever (thanks, zero dollars to spend on tech!), it couldn't be too resource-intensive — I honestly worried that even using PHP includes to power that many pageviews would overtax it.
Since the site is a preview, it's not going to be updated that often, negating the primary downside of a flat-file build system (longer time to publish). I've mentioned before that we've previously built off of Bootstrap, but the limitations we kept hitting in terms of templating (many elements require specific, one-off classes and styles to work right on all devices) drove us looking in another direction.
The framework that seemed most complete and contained the elements we were looking for was Zurb's Foundation.
Though it was not without its own headaches (Foundation 5 is built off
an old version of SASS, which can play hell with your compiler — the
solution is to replace the deprecated global variables, specifically
!global; and replacing
if === false statements with
if not statements, as outlined in an answer here. Zurb says they're rewriting the SASS for F6), it ultimately worked out for us.
The basic method for extracting data from the Google Docs turned out both easier and more difficult than expected. The original plan was to query the two main admin sheets (that described the league structures as well as the league pages) and go from there.
That much was easy — I wrote a Google Apps script that I granted access to my Docs that outputs some customized JSON based on which pages are queried.
A PHP build script (which can be set to rebuild the whole thing, a whole league, or a league's teams or league pages) grabs that info, then goes back and grabs the data for the queried pages. It's a lot of calls (hence why each update is referred to as a "build," so that the content desk would understand that this is not a WordPress post they're updating), but the most important thing was to keep the content creation and updating as easy possible — I can convince editors to go back and edit their typos in a Google Doc, whereas it's much more difficult to convince them to dig into an HTML file to find their errors without creating more problems. The PHP script outputs partial templates based on the type of page — again, in the interest of not wanting to have to rebuild the whole app every time a small change is made, I didn't want to rely on the PHP scripts to build everything — they're strictly for extracting data in a sensible manner.
The PHP script outputs a combination of JSON and .kit files. .kit is the file extension for CodeKit 2's .kit language (I heartily recommend CodeKit2 for web devs, by the way), which is essentially PHP includes for HTML. This worked perfectly for our plans, since it allowed the major parts of the templates to be kept in a single location without having to literally regenerate the whole site (the PHP build script takes, on average, about 3-5 minutes to output the site — the .kit compile takes about 20 seconds). Dropping the .kit files into the build folder automatically generates the static HTML files in a different directory, and the site is ready to go.
Aside from the obvious challenges of just getting things to work, the biggest challenge was extracting the text from the Google Docs with formatting intact. There are methods using the getAttributes method of the text class, but I could not get it to work reliably. (Of course, when I went to Google the partial answers I saw before I found a Markdown converter script that can email you the document that could be easily adapted. Damnit.)
We did not even look at, much less open the can of worms that is embedded images.
We're beyond happy in our decision to forego the app route in favor of responsive design — we had more visitors to the site in the first hour of its going live than we did downloads of the app to that point (more than a year later). The larger potential audience, the ability to deep-link into the site and the ease of access (get it wherever you are!) combined to make it a much bigger success. There are still a few updates we're going to get in before the start of the season, though — more teams, full rosters and some videos are still to come.