All Marked Up with No Place to Go
NOT ON THE CPAN
Today's article is about code not found on the CPAN.
Talk, talk, talk!
For the past few years, I've given quite a few technical talks at YAPC, OSCON, and other conferences. I really enjoy doing this, and I like to think I've gotten fairly good at producing good slides to illustrate the things I'm discussing. Since the initial release of Apple Keynote, I've been using it for nearly all of my slide construction, and I like it quite a bit, but every once in a while I run into one of its shortcomings.
Although Keynote can be scripted with AppleScript, I've found that its scripting facilities are pretty weak, and they've never been sufficient to let me do what I want. Sometimes, this means I skip doing something, and sometimes it means I find a very convoluted way to achieve my goals using only Keynote's stock features.
One of the things that's left me most frustrated, though, is my inability to automatically syntax highlight source code on slides. When you're displaying fifteen lines of code on a big screen to people sitting in an amphiteater, syntax highlighting can be a huge help for readability. Just compare this plain code:
...to this syntax highlighted code:
Adding syntax highlighting seemed like it would be a huge win, and would be useful to quite a few people, so I decided to undertake the endeavor. I knew that like many OS X apps (at the time, at least), Keynote was using RTF to store the rich text of slides. If I could produce RTF syntax highlighted code, I could get it into Keynote and be done!
Making Stew from the CPAN
In the end, making the RTF -- which I'd assumed would be the most difficult part of the job -- was pretty simple. The thing that made it simple, as with so many things in Perl, was the CPAN. All I needed to do was find the right set of modules and figure out how to tie them together. What follows is a run-through of the code-to-RTF program I produced.
First, we load the libraries we'll need:
Everybody remember Getopt::Long::Descriptive? Good. We're going to use that to get some very basic options:
--filetype option is important, because it tells us what kind of syntax to highlight. After all, I have been known to include non-Perl code in my slides, once in a while.
Here, we use one of my favorite modules. I only use it very occasionally, but it always saves quite a lot of time. While there are a number of syntax highlighting libraries on the CPAN, they often strike me as experimental or awkward to use. Vim, on the other hand, does a pretty good job with many different languages, and Text::VimColor lets us get exactly the coloring that Vim would have used. Even better, we'll be able to get at the syntax highlighted document as a list of substrings of the document, each with a syntax type, something like this:
$VAR1 = [ [ Statement => 'my' ], [ '' => ' ' ], [ Identifier => '$syntax' ], [ '' => ' = ' ], ... ];
With our input primed in
$syn, it's time to prime our output mechanism with RTF::Writer:
We've initialized our RTF::Writer so that as it writes out text, it puts it in an in-memory buffer. We're not going to be doing anything gargantuan, and thinking about temporary files is always a drag. Then we have to futz around with some settings to tweak out output. We prevent the replacement of hyphens with "non-breaking hypens" because Apple's RTF implementation doesn't seem to understand them, and we preserve the location of newlines so that we can skim our output RTF by eye to look for problems.
Now we've set up the RTF prolog, which tells future readers in advance what fonts and colors will be used in the document. We scale up to RTF's weird "half point" font sizing, and get into our starting format by setting up our font size and colors by printing some raw RTF commands.
\fs sets font size,
\f sets font, and
\c sets the color. Where do
color_controls_for come from? We'll get back to those in a bit.
At this point, we've loaded in a stream of syntax-marked text hunks and we've got an RTF output stream ready to go. All we have to do is read, transform, and print -- the Platonic Perl Program:
@$tokens is our input, each entry a pair of "type" and "text" values. For each one, we spit out a color command, terminated by a space, and then the text. We don't need to worry about anything else, because we've picked a monospace font and RTF won't try to mangle our whitespace. (Thanks, RTF!) We can just finalize the RTF document and print out the buffer to which it was written:
So, that's the whole program -- save for color handling -- in about 45 significant lines.
It turns out that almost all of the work was in reading the Vim color scheme file!
We read in the syntax highlighting groups from a given file -- here, my personal Vim color scheme -- and save everything as two entries in a hash. The keys are names like Identifier:fg and Identifier:bg, and the values are numbers -- because in RTF, we always refer to colors by their position in the index of colors, like this:
So this code has to set up the index and make sure all the possible names refer back to index positions. The color index looks like this:
...so we need to get all our colors into RGB values, but in Vim color schemes, we're free to use named colors, like "SlateBlue." We'll resolve these with Graphics::ColorUtils, a weird little module that turns out to be really handy once in a while.
Also note that for some bizarre reason I chose to use
unpack, above. It's actually the only form of unpack that I know how to use.
Finally, we get to the heart of the matter:
all_colors returns a list of RGB arrayrefs to be used as the color index, and
color_controls_for returns an RTF command sequence to switch to the color for the given group.
The Proof of the Stew is in the Eating
I got this code written pretty quickly, and felt pretty good about it. It didn't do anything clever or tricky, it just tied together a few simple CPAN modules to perform a simple text transformation, and it worked! Those are the qualities I like to end up with in my code: simple, straightforward, mostly implemented in terms of other libraries, and working. You can look at some sample output, if you want, too.
From here, things should have been simple. I'd write some AppleScript like this (forgive my pseudo-AppleScript):
Well, it turns out that you can't address the formatting of slides in Keynote, at all. You also can't address anything on the slide that isn't the main body -- and the main body has several behaviors that differ from other shapes with text. I could've resorted to GUI-level automation to do things like:
assume that the current text object is the target
simulate Cmd-A, Cmd-C to select all text and copy it
pbclipto get pasteboard contents
synrtfoutput and put it on the pasteboard
simulate pressing Cmd-V to paste the formatted text
It was pretty clear by this point, though, that all the work lay ahead of me would be boring drudgery, and that I could just quit now and feel good about the RTF bits, maybe to revisit them later, when Keynote's automation improved.
I'm sorry to report that now, three years later, the automation hasn't gotten any better, and it no longer uses RTF. I don't see myself switching away from Keynote, despite that, but maybe
synrtf, or some of the code it shows off, will be useful to someone else the way I'd hoped the original project would be.