Javascript: Parsing and tokenising a markup language with regular expressions

Previously we did some tokenising and parsing for a simple markup lanuage (

But the problem was the token were delimited by spaces, therefore _this could not be used_.

We can get around this problem by doing more complex tokenisation. And we can use regexs for this.

So for hello _i am a token_ and *i am listening to james brown* and /get up/ we will split the tokens based on if they are delimited by _, *, /.

Building the regex

We need to know how regexs work, but let's look at \*.*\* brefiely:

  1. This match a * (we must prefix it with \)
  2. ...then any character (represented by .)
  3. ..multiple times (respresented by a *)
  4. until we get a * (again, prefixed with \)

Therefore "hello *i am james* again".match(/\*.*\*/g) gives us ["*i am james*"]

But we have a problem since "hello *bold1* and *bold2*".match(/\*.*\*/g) gives us ["*bold1* and *bold2*"].

This is because we "greedily" match the biggest * ... * we can. We can tell the regex not be greedy with *? instead of just *. So "hello *bold1* and *bold2*".match(/\*.*?\*/g) gives us ["*bold1*", "*bold2*"]

We can also match _this_ and /this/ by using anor separator (represented by |) in the regex.

The other two regexs we'll use for those are: _.*?_ and \/.*?\/ (the / must be prefixed with a \).

So "hello *i am james* again /italic/ and _underlined_ yes".match(/\*.*\*|_.*?_|\/.*?\//g) gives us ["*i am james*", "/italic/", "_underlined_"].

But we're still missing out all the text that's not italic, bold or underlined. We can grab this by saying "start with a character that's not *, / or _ and continue looking at such a thing until you find a character that is *, / or _.

In regex land we use [^b] to say match anything that's not b. So our regex will look like this: [^*_/][^*_/]*[^*_/]. Notice we're not using ?. This is because we now want to match as many of these as we can.

Our regex looks like this: "hello *i am james* again /italic/ and _underlined_ yes".match(/\*.*\*|_.*?_|\/.*?\/|[^*_/][^*_/]*[^*_/]/g) which gives us ["hello ", "*i am james*", " again ", "/italic/", " and ", "_underlined_", " yes"].

Our final problem is that this doesn't work across multiple lines hello *i am\nbold* yeah will not work. That's because . in regex world doesn't work over multiple lines. We can make it do that by replacing it with a dotall operator: [\s\S].

Our regex is now huge and unmaintainable. So let's split each match into array element, and pass that into the Regex object:

var dot = "[\s\S]"
var tokens = [`_${dot}*?_`, 
"hello *i am james* again /italic/ and _underlined_ yes".match(new RegExp(tokens.join("|"), "g"))

But we have an extra problem with string interpretation since we're no longer using /reg_ex_here/. The \ won't be shown. And we need it. So we need to put an extra \ in front of it. But this will make our regex even uglier. There's a better solution though. Things in [...] don't need to be quoted, so we can use:

var dot = "[\\s\\S]"
var tokens = [`_${dot}*?_`, 
"hello *i am james* again /italic/ and _underlined_ yes".match(new RegExp(tokens.join("|"), "g"))

Plugging the regex into our function

One difference between the last version is that we're no longer dealing with spaces and newlines at the start of the function. We replace a \n and a double space at the end of the function with the string replace function.

The code for the full function is:

function quickText(txt) {
  var replaceSpaceAndNL = t => t.replace(/  /g, "&nbsp;").replace(/\n/g, ' <br> ')
  var dot = "[\\s\\S]"
  var tokens = [`_${dot}*?_`, 
  return txt.match(new RegExp(tokens.join("|"), "g"))
    .map(t => {
      if(t.startsWith("/")) return ["italic", t.slice(1, -1)]
      else if (t.startsWith("_")) return ["underline", t.slice(1, -1)]
      else if (t.startsWith("*")) return ["bold", t.slice(1, -1)]
      else return ["normal", t]
    .map(token => {
      if(token[0] == "italic") return `<i>${replaceSpaceAndNL(token[1])}</i>`
      else if (token[0] == "underline") return `<u>${replaceSpaceAndNL(token[1])}</u>`
      else if (token[0] == "bold") return `<b>${replaceSpaceAndNL(token[1])}</b>`
      else return replaceSpaceAndNL(token[1])

You can play with the result here:

It doesn't deal with _hello *there* again_. But that can come in a later post.


Edit on github
comments powered by Disqus
Click me