scalacheck-gen-regexp

A library for creating scalacheck generators from regular expressions

Cross-built for Scala 2.12/2.13/3.1/3.2

Installation

In your build.sbt

libraryDependencies += "io.github.wolfendale" %% "scalacheck-gen-regexp" % "[VERSION]"

Quick start

import wolfendale.scalacheck.regexp.RegexpGen

val generator: Gen[String] = RegexpGen.from("[1-9]\\d?(,\\d{3})+")

Supported syntax

Feature	Example	Notes
Literals	a, \\w, 7	Literals are transformed into constant generators
Character Classes	[abc], [^abc], [a-zA-Z0-9]	Character classes are transformed with Gen.oneOf
Default Classes	\w, \d, \S, .	These are treated as predefined character classes
Quantifiers	a?, b+, c*, d{3}, e{4,5}, f{5,}	These use Gen.listOfN to create sized lists of the preceding term
Groups	(abc), (?:def)	Backreferences are not supported, groups can only be used for grouping terms
Alternates	a|b|c, a(b|c)d	Alternates are also transformed with Gen.oneOf
Boundaries	^, $, \b	Although these will be parsed they do not modify the generator output

Unsupported syntax

Feature	Example	Notes
Backreferences	([ab]\1)	With the current implementation there's no simple way to do this, definitely in consideration for a future release
Octal / Hex / Special Literals	\012, \xF1, \p{Lower}	Most of these should be simple to implement but I wanted to get an initial release created first
Character Class Intersection	[a&&[b]], [a[b]]	Difficult to implement currently but not impossible, definite consideration for a future release

Tips for usage

• In order to represent any character, RegexpGen#from takes an implicit Arbitrary[Char]. There is a default instance provided by scalacheck however for most uses you probably want to provide your own.

• If you use the + or * quantifiers you'll end up getting huge variance in string sizes. If this isn't what you want, consider bounding the lengths of certain string segments with the {min,max} quantifier.

• Negated character classes / default classes are implemented by generating an arbitrary Char within certain bounds via suchThat, because of this you can end up throwing away a lot of cases and in certain circumstances your tests may fail. Try to refactor out negative cases.

• In character classes each option is given equal weighting. If you'd prefer to weight a particular entry you can add it multiple times, this is made easier with string interpolation: s"[${"a-z"*5}\s]". In the example case the generator is 5 times more likely to generate an alpha character than a space.