Implementing an internal DSL: regular expressions

Overview

In this assignment, you'll use the techniques we've learned in class to implement an internal DSL. This internal DSL allows users to easily check whether a given string matches a pattern. For example, imagine a user wants to check whether a string s takes the form of a letter, followed by a number (e.g., c3). In your DSL, the user might write:

val pattern = letter ~ number
pattern matches s

Patterns like this are called regular expressions. Before you start implementing a DSL for the domain of regular expressions, you should first read the next section, which describes everything you'll need to know about this domain.

Background: regular expressions

A regular expression is a pattern—a shorthand for describing sets of strings. Every string that matches a regular expression is said to be a member of the set described by that language. As an informal example, the pattern watch(es) describes a set of two strings: {watch, watches}. Similarly, the pattern colo(u)r also contains two strings: {color, colour}. Neither of these informal patterns is a regular expression in the formal sense, but they give a sense of how a pattern can stand in for a set of strings.

The most precise definition of regular expressions is a formal one, i.e., one that's based on math. That's the definition we're most interested in because it's easy to translate it into code. Here we go...

We start with an alphabet, which is a set of characters. For this assignment, our alphabet will be any character that can appear in a Scala string, i.e., anything that the Scala compiler will let us put between two quotes like so: 'a'.

A string s is a sequence of characters c1⋅⋅⋅cn; this string has length n. There is a special string called the null string, spelled ε; this word has length 0.

A language L is a set of strings.

Given an alphabet, the regular languages “over that alphabet” can be defined as follows:

• The empty language ∅ is regular. This language contains 0 strings.
• The language {ε} that contains one string—the null string—is regular.
• If c is a character, then the language {c} that contains one string is regular.
• If L1 and L2 are regular languages (i.e., they're each sets of strings), then the language L1 ∪ L2 (i.e., the union of the two languages) is regular.
• If L1 and L2 are regular, then the language L1 ⋅ L2 = {s1 ⋅ s2 | s1 ∈ L1 ∧ s2 ∈ L2} is regular. Here, the symbol ⋅ means "concatenation". So s1 ⋅ s2 is all the characters in s1, followed by all the characters in s2.
• If L is regular, then L* is regular. Here, the symbol * means "0 or more repetitions of any string in L". So, if L is the language that contains the single string ab, then L* is {ε, ab, abab, ababab, …}.

A regular expression is a short-hand description for a regular language. For example, (ab)* is a regular expression that describes the infinitely large set of strings {ab}*. The regular expression 42 describes the language {42}, which contains a single string. The regular expression 4 || 2 describes the language {4, 2}, which contains two strings.

Given a regular expression and a candidate string, there are many ways to determine whether the string matches the expression, i.e., whether the string is in the set described by the expression. In this assignment, you don't have to worry about implementing a matching algorithm—it's been provided for you. Your job is to implement the syntax of regular expressions: an easy way for users to describe them.

Implementing a DSL for regular expressions

Note: Be sure to read the entire assignment before you start implementing!

This repository contains some starter code. The most important pieces are:

RegularExpression.scala This file defines data structures that correspond to the different kinds of regular languages described above. You'll modify this file to write your DSL.

Program.scala A program that uses the data structures in RegularExpresion.scala to build descriptions of regular expressions, then match strings against those expressions. This program compiles and runs correctly as-is, but it's clunky: it uses the data structures from the regular expression library but it's not very DSL-y. Your job is to modify this file to use your DSL instead.

There are a couple of other pieces of code provided:

RegexMatcher.scala This file defines an algorithm for matching a string against a regular expression. You won't need to modify this file.

src/test/scala/dsls/regex Some test files for the matching algorithm. You won't need to modify these files.

The syntax for your DSL

The user of your DSL eventually should be able to write the following pattern:

"42" || ( ('a'*) ~ ('b'+) ~ ('c'{3}))

which matches either:

• the literal string "42" or
• any string that has zero or more 'a's, followed by one or more 'b's, followed by exactly three 'c's

To do so, you'll need to:

• implement literal extension, so that Scala strings can be treated as regular expressions
• implement literal extension, so that Scala characters can be treated as regular expressions
• implement the binary operator ||, which corresponds to the union operation
• implement the binary operator ~, which corresponds to the concatenation operation
• implement the postfix operator *, which corresponds to the Kleene star operation
• implement the postfix operator +, which means "one or more repetitions of the preceding pattern"
• implement the repetition operator {n} which means "n repetitions of the preceding pattern"

Reflecting on design / implementation

In the file reflection.md, you'll write about:

• Your process of implementing the DSL
• Your thoughts about the design of this DSL

I highly recommend that you work on this part as you go. In other words, treat it as a running diary. Take a moment now to look at the questions in that file, then update the file as you work on the assignment. Although you might treat the file as a running diary, the version you submit should not be a stream of consciousness. Make your final submission clear and concise.

Grading

Good responses (i.e., responses that receive a 3) will:

• modify RegularExpression.scala to implement the internal DSL
• transform Program.scala to work with the DSL
• respond to the questions in reflection.md

Great responses (i.e., responses that receive a 4) will additionally:

• provide very well-documented code in RegularExpression.scala.
• deeply engage in the reflection process, e.g., by discussing the benefits and drawbacks of every operation in the language, from both the implementer's and user's perspective

Critiques

After the submission deadline, read over and comment on your critique partners' work. Be sure to comment on their code and on their reflection.

Here are some suggestions for critiques:

• What do you like about their implementation?
• Is there anything that could be done to improve their implementation?
• Did they implement things in the same or in different ways as you?
• With which aspects of their reflection do you agree? With which aspects do you disagree?

Tasks

• Read the assignment.
• Implement literal extension for characters.
• Implement literal extension for strings.
• Implement the binary operator ||.
• Implement the binary operator ~.
• Implement the postfix operator *.
• Implement the postfix operator +.
• Implement the repetition operator {n}.
• Write responses in reflection.md.
• Submit your work.
• Comment on your critique partners' work.