<!doctype html>
<html>
<head>
<title>CodeMirror: Internals</title>
<link rel="stylesheet" type="text/css" href="http://fonts.googleapis.com/css?family=Droid+Sans|Droid+Sans:bold"/>
<link rel="stylesheet" type="text/css" href="css/docs.css"/>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>
<style>dl dl {margin: 0;}</style>
</head>
<body>
<h1><span class="logo-braces">{ }</span> <a href="http://codemirror.net/">CodeMirror</a></h1>
<pre class="grey">
<img src="css/baboon.png" class="logo" alt="logo"/>/* (Re-) Implementing A Syntax-
Highlighting Editor in JavaScript */
</pre>
<div class="clear"><div class="leftbig blk">
<p style="font-size: 85%" id="intro">
<strong>Topic:</strong> JavaScript, code editor implementation<br>
<strong>Author:</strong> Marijn Haverbeke<br>
<strong>Date:</strong> March 2nd 2011
</p>
<p>This is a followup to
my <a href="http://codemirror.net/story.html">Brutal Odyssey to the
Dark Side of the DOM Tree</a> story. That one describes the
mind-bending process of implementing (what would become) CodeMirror 1.
This one describes the internals of CodeMirror 2, a complete rewrite
and rethink of the old code base. I wanted to give this piece another
Hunter Thompson copycat subtitle, but somehow that would be out of
place—the process this time around was one of straightforward
engineering, requiring no serious mind-bending whatsoever.</p>
<p>So, what is wrong with CodeMirror 1? I'd estimate, by mailing list
activity and general search-engine presence, that it has been
integrated into about a thousand systems by now. The most prominent
one, since a few weeks,
being <a href="http://googlecode.blogspot.com/2011/01/make-quick-fixes-quicker-on-google.html">Google
code's project hosting</a>. It works, and it's being used widely.</a>
<p>Still, I did not start replacing it because I was bored. CodeMirror
1 was heavily reliant on <code>designMode</code>
or <code>contentEditable</code> (depending on the browser). Neither of
these are well specified (HTML5 tries
to <a href="http://www.w3.org/TR/html5/editing.html#contenteditable">specify</a>
their basics), and, more importantly, they tend to be one of the more
obscure and buggy areas of browser functionality—CodeMirror, by using
this functionality in a non-typical way, was constantly running up
against browser bugs. WebKit wouldn't show an empty line at the end of
the document, and in some releases would suddenly get unbearably slow.
Firefox would show the cursor in the wrong place. Internet Explorer
would insist on linkifying everything that looked like a URL or email
address, a behaviour that can't be turned off. Some bugs I managed to
work around (which was often a frustrating, painful process), others,
such as the Firefox cursor placement, I gave up on, and had to tell
user after user that they were known problems, but not something I
could help.</p>
<p>Also, there is the fact that <code>designMode</code> (which seemed
to be less buggy than <code>contentEditable</code> in Webkit and
Firefox, and was thus used by CodeMirror 1 in those browsers) requires
a frame. Frames are another tricky area. It takes some effort to
prevent getting tripped up by domain restrictions, they don't
initialize synchronously, behave strangely in response to the back
button, and, on several browsers, can't be moved around the DOM
without having them re-initialize. They did provide a very nice way to
namespace the library, though—CodeMirror 1 could freely pollute the
namespace inside the frame.</p>
<p>Finally, working with an editable document means working with
selection in arbitrary DOM structures. Internet Explorer (8 and
before) has an utterly different (and awkward) selection API than all
of the other browsers, and even among the different implementations of
<code>document.selection</code>, details about how exactly a selection
is represented vary quite a bit. Add to that the fact that Opera's
selection support tended to be very buggy until recently, and you can
imagine why CodeMirror 1 contains 700 lines of selection-handling
code.</p>
<p>And that brings us to the main issue with the CodeMirror 1
code base: The proportion of browser-bug-workarounds to real
application code was getting dangerously high. By building on top of a
few dodgy features, I put the system in a vulnerable position—any
incompatibility and bugginess in these features, I had to paper over
with my own code. Not only did I have to do some serious stunt-work to
get it to work on older browsers (as detailed in the
previous <a href="http://codemirror.net/story.html">story</a>), things
also kept breaking in newly released versions, requiring me to come up
with <em>new</em> scary hacks in order to keep up. This was starting
to lose its appeal.</p>
<h2 id="approach">General Approach</h2>
<p>What CodeMirror 2 does is try to sidestep most of the hairy hacks
that came up in version 1. I owe a lot to the
<a href="http://ace.ajax.org">ACE</a> editor for inspiration on how to
approach this.</p>
<p>I absolutely did not want to be completely reliant on key events to
generate my input. Every JavaScript programmer knows that key event
information is horrible and incomplete. Some people (most awesomely
Mihai Bazon with <a href="http://ymacs.org">Ymacs</a>) have been able
to build more or less functioning editors by directly reading key
events, but it takes a lot of work (the kind of never-ending, fragile
work I described earlier), and will never be able to properly support
things like multi-keystoke international character input.</p>
<p>So what I do is focus a hidden textarea, and let the browser
believe that the user is typing into that. What we show to the user is
a DOM structure we built to represent his document. If this is updated
quickly enough, and shows some kind of believable cursor, it feels
like a real text-input control.</p>
<p>Another big win is that this DOM representation does not have to
span the whole document. Some CodeMirror 1 users insisted that they
needed to put a 30 thousand line XML document into CodeMirror. Putting
all that into the DOM takes a while, especially since, for some
reason, an editable DOM tree is slower than a normal one on most
browsers. If we have full control over what we show, we must only
ensure that the visible part of the document has been added, and can
do the rest only when needed. (Fortunately, the <code>onscroll</code>
event works almost the same on all browsers, and lends itself well to
displaying things only as they are scrolled into view.)</p>
<h2 id="input">Input</h2>
<p>ACE uses its hidden textarea only as a text input shim, and does
all cursor movement and things like text deletion itself by directly
handling key events. CodeMirror's way is to let the browser do its
thing as much as possible, and not, for example, define its own set of
key bindings. One way to do this would have been to have the whole
document inside the hidden textarea, and after each key event update
the display DOM to reflect what's in that textarea.</p>
<p>That'd be simple, but it is not realistic. For even medium-sized
document the editor would be constantly munging huge strings, and get
terribly slow. What CodeMirror 2 does is put the current selection,
along with an extra line on the top and on the bottom, into the
textarea.</p>
<p>This means that the arrow keys (and their ctrl-variations), home,
end, etcetera, do not have to be handled specially. We just read the
cursor position in the textarea, and update our cursor to match it.
Also, copy and paste work pretty much for free, and people get their
native key bindings, without any special work on my part. For example,
I have emacs key bindings configured for Chrome and Firefox. There is
no way for a script to detect this.</p>
<p>Of course, since only a small part of the document sits in the
textarea, keys like page up and ctrl-end won't do the right thing.
CodeMirror is catching those events and handling them itself.</p>
<h2 id="selection">Selection</h2>
<p>Getting and setting the selection range of a textarea in modern
browsers is trivial—you just use the <code>selectionStart</code>
and <code>selectionEnd</code> properties. On IE you have to do some
insane stuff with temporary ranges and compensating for the fact that
moving the selection by a 'character' will treat \r\n as a single
character, but even there it is possible to build functions that
reliably set and get the selection range.</p>
<p>But consider this typical case: When I'm somewhere in my document,
press shift, and press the up arrow, something gets selected. Then, if
I, still holding shift, press the up arrow again, the top of my
selection is adjusted. The selection remembers where its <em>head</em>
and its <em>anchor</em> are, and moves the head when we shift-move.
This is a generally accepted property of selections, and done right by
every editing component built in the past twenty years.</p>
<p>But not something that the browser selection APIs expose.</p>
<p>Great. So when someone creates an 'upside-down' selection, the next
time CodeMirror has to update the textarea, it'll re-create the
selection as an 'upside-up' selection, with the anchor at the top, and
the next cursor motion will behave in an unexpected way—our second
up-arrow press in the example above will not do anything, since it is
interpreted in exactly the same way as the first.</p>
<p>No problem. We'll just, ehm, detect that the selection is
upside-down (you can tell by the way it was created), and then, when
an upside-down selection is present, and a cursor-moving key is
pressed in combination with shift, we quickly collapse the selection
in the textarea to its start, allow the key to take effect, and then
combine its new head with its old anchor to get the <em>real</em>
selection.</p>
<p>In short, scary hacks could not be avoided entirely in CodeMirror
2.</p>
<p>And, the observant reader might ask, how do you even know that a
key combo is a cursor-moving combo, if you claim you support any
native key bindings? Well, we don't, but we can learn. The editor
keeps a set known cursor-movement combos (initialized to the
predictable defaults), and updates this set when it observes that
pressing a certain key had (only) the effect of moving the cursor.
This, of course, doesn't work if the first time the key is used was
for extending an inverted selection, but it works most of the
time.</p>
<h2 id="update">Intelligent Updating</h2>
<p>One thing that always comes up when you have a complicated internal
state that's reflected in some user-visible external representation
(in this case, the displayed code and the textarea's content) is
keeping the two in sync. The naive way is to just update the display
every time you change your state, but this is not only error prone
(you'll forget), it also easily leads to duplicate work on big,
composite operations. Then you start passing around flags indicating
whether the display should be updated in an attempt to be efficient
again and, well, at that point you might as well give up completely.</p>
<p>I did go down that road, but then switched to a much simpler model:
simply keep track of all the things that have been changed during an
action, and then, only at the end, use this information to update the
user-visible display.</p>
<p>CodeMirror uses a concept of <em>operations</em>, which start by
calling a specific set-up function that clears the state and end by
calling another function that reads this state and does the required
updating. Most event handlers, and all the user-visible methods that
change state are wrapped like this. There's a method
called <code>operation</code> that accepts a function, and returns
another function that wraps the given function as an operation.</p>
<p>It's trivial to extend this (as CodeMirror does) to detect nesting,
and, when an operation is started inside an operation, simply
increment the nesting count, and only do the updating when this count
reaches zero again.</p>
<p>If we have a set of changed ranges and know the currently shown
range, we can (with some awkward code to deal with the fact that
changes can add and remove lines, so we're dealing with a changing
coordinate system) construct a map of the ranges that were left
intact. We can then compare this map with the part of the document
that's currently visible (based on scroll offset and editor height) to
determine whether something needs to be updated.</p>
<p>CodeMirror uses two update algorithms—a full refresh, where it just
discards the whole part of the DOM that contains the edited text and
rebuilds it, and a patch algorithm, where it uses the information
about changed and intact ranges to update only the out-of-date parts
of the DOM. When more than 30 percent (which is the current heuristic,
might change) of the lines need to be updated, the full refresh is
chosen (since it's faster to do than painstakingly finding and
updating all the changed lines), in the other case it does the
patching (so that, if you scroll a line or select another character,
the whole screen doesn't have to be re-rendered).</p>
<p>All updating uses <code>innerHTML</code> rather than direct DOM
manipulation, since that still seems to be by far the fastest way to
build documents. There's a per-line function that combines the
highlighting, <a href="manual.html#markText">marking</a>, and
selection info for that line into a snippet of HTML. The patch updater
uses this to reset individual lines, the refresh updater builds an
HTML chunk for the whole visible document at once, and then uses a
single <code>innerHTML</code> update to do the refresh.</p>
<h2 id="parse">Parsers can be Simple</h2>
<p>When I wrote CodeMirror 1, I
thought <a href="http://codemirror.net/story.html#parser">interruptable
parsers</a> were a hugely scary and complicated thing, and I used a
bunch of heavyweight abstractions to keep this supposed complexity
under control: parsers
were <a href="http://bob.pythonmac.org/archives/2005/07/06/iteration-in-javascript/">iterators</a>
that consumed input from another iterator, and used funny
closure-resetting tricks to copy and resume themselves.</p>
<p>This made for a rather nice system, in that parsers formed strictly
separate modules, and could be composed in predictable ways.
Unfortunately, it was quite slow (stacking three or four iterators on
top of each other), and extremely intimidating to people not used to a
functional programming style.</p>
<p>With a few small changes, however, we can keep all those
advantages, but simplify the API and make the whole thing less
indirect and inefficient. CodeMirror
2's <a href="manual.html#modeapi">mode API</a> uses explicit state
objects, and makes the parser/tokenizer a function that simply takes a
state and a character stream abstraction, advances the stream one
token, and returns the way the token should be styled. This state may
be copied, optionally in a mode-defined way, in order to be able to
continue a parse at a given point. Even someone who's never touched a
lambda in his life can understand this approach. Additionally, far
fewer objects are allocated in the course of parsing now.</p>
<p>The biggest speedup comes from the fact that the parsing no longer
has to touch the DOM though. In CodeMirror 1, on an older browser, you
could <em>see</em> the parser work its way through the document,
managing some twenty lines in each 50-millisecond time slice it got. It
was reading its input from the DOM, and updating the DOM as it went
along, which any experienced JavaScript programmer will immediately
spot as a recipe for slowness. In CodeMirror 2, the parser usually
finishes the whole document in a single 100-millisecond time slice—it
manages some 1500 lines during that time on Chrome. All it has to do
is munge strings, so there is no real reason for it to be slow
anymore.</p>
<h2 id="summary">What Gives?</h2>
<p>Given all this, what can you expect from CodeMirror 2? First, the
good:</p>
<ul>
<li><strong>Small.</strong> the base library is some 32k when minified
now, 12k when gzipped. It's smaller than its own logo.</li>
<li><strong>Lightweight.</strong> CodeMirror 2 initializes very
quickly, and does almost no work when it is not focused. This means
you can treat it almost like a textarea, have multiple instances on a
page without trouble.</li>
<li><strong>Huge document support.</strong> Since highlighting is
really fast, and no DOM structure is being built for non-visible
content, you don't have to worry about locking up your browser when a
user enters a megabyte-sized document.</li>
<li><strong>Extended API.</strong> Some things kept coming up in the
mailing list, such as marking pieces of text or lines, which were
extremely hard to do with CodeMirror 1. The new version has proper
support for these built in.</li>
<li><strong>Tab support.</strong> Tabs inside editable documents were,
for some reason, a no-go. At least six different people announced they
were going to add tab support to CodeMirror 1, none survived (I mean,
none delivered a working version). CodeMirror 2 no longer removes tabs
from your document.</li>
<li><strong>Sane styling.</strong> <code>iframe</code> nodes aren't
really known for respecting document flow. Now that an editor instance
is a plain <code>div</code> element, it is much easier to size it to
fit the surrounding elements. You don't even have to make it scroll if
you do not <a href="demo/resize.html">want to</a>.</li>
</ul>
<p>Then, the bad:</p>
<ul>
<li><strong>No line-wrapping.</strong> I'd have liked to get
line-wrapping to work, but it doesn't match the model I'm using very
well. It is important that cursor movement in the textarea matches
what you see on the screen, and it seems to be impossible to have the
lines wrapped the same in the textarea and the normal DOM.</li>
<li><strong>Some cursor flakiness.</strong> The textarea hack does not
really do justice to the complexity of cursor handling—a selection is
typically more than just an offset into a string. For example, if you
use the up and down arrow keys to move to a shorter line and then
back, you'll end up in your old position in most editor controls, but
CodeMirror 2 currently doesn't remember the 'real' cursor column in
this case. These can be worked around on a case-by-case basis, but
I haven't put much energy into that yet.</li>
<li><strong>Limited interaction with the editable panel.</strong>
Since the element you're looking at is not a real editable panel,
native browser behaviour for editable controls doesn't work
automatically. Through a lot of event glue code, I've managed to make
drag and drop work pretty well, have context menus work on most
browsers (except Opera). Middle-click paste on Firefox in Linux is
broken until someone finds a way to intercept it.</li>
</ul>
</div><div class="rightsmall blk">
<h2>Contents</h2>
<ul>
<li><a href="#intro">Introduction</a></li>
<li><a href="#approach">General Approach</a></li>
<li><a href="#input">Input</a></li>
<li><a href="#selection">Selection</a></li>
<li><a href="#update">Intelligent Updating</a></li>
<li><a href="#parse">Parsing</a></li>
<li><a href="#summary">What Gives?</a></li>
</ul>
</div></div>
<div style="height: 2em"> </div>
</body></html>