A regular expressions library

Related Projects
,
1

I've created a regular expressions library for swift! Check it out here. I would appreciate some feedback. Here are some examples:

var inputText = "name: Chris Lattner"

// create the regular expression object
let regex = try Regex(
    pattern: "name: ([a-z]+) ([a-z]+)",
    regexOptions: [.caseInsensitive]
)
 
if let match = try inputText.regexMatch(regex) {
    print("full match: '\(match.fullMatch)'")
    print("first capture group: '\(match.groups[0]!.match)'")
    print("second capture group: '\(match.groups[1]!.match)'")
    
    // perform a replacement on the first capture group
    inputText.replaceSubrange(
        match.groups[0]!.range, with: "Steven"
    )
    
    print("after replacing text: '\(inputText)'")
}

// full match: 'name: Chris Lattner'
// first capture group: 'Chris'
// second capture group: 'Lattner'
// after replacing text: 'name: Steven Lattner'

let name = "Charles Darwin"
let reversedName = try name.regexSub(
    #"(\w+) (\w+)"#,
    with: "$2 $1"
    // $1 and $2 represent the
    // first and second capture group, respectively.
    // $0 represents the entire match.
)
// reversedName = "Darwin Charles"

My library also supports naming the capture groups!. For example:

var inputText = "season 8, EPISODE 5; season 5, episode 20"

// create the regular expression object
let regex = try Regex(
    pattern: #"season (\d+), Episode (\d+)"#,
    regexOptions: [.caseInsensitive],
    groupNames: ["season number", "episode number"]
    // the names of the capture groups
)
        
let results = try inputText.regexFindAll(regex)
for result in results {
    print("fullMatch: '\(result.fullMatch)'")
    print("capture groups:")
    for captureGroup in result.groups {
        print("    \(captureGroup!.name!): '\(captureGroup!.match)'")
    }
    print()
}
let firstResult = results[0]
// perform a replacement on the first full match
inputText.replaceSubrange(
    firstResult.range, with: "new value"
)
print("after replacing text: '\(inputText)'")

// fullMatch: 'season 8, EPISODE 5'
// capture groups:
//     'season number': '8'
//     'episode number': '5'
//
// fullMatch: 'season 5, episode 20'
// capture groups:
//     'season number': '5'
//     'episode number': '20'
//
// after replacing text: 'new value; season 5, episode 20'
2 Likes
2

great work! have u thought about removing the NSRegularExpression dependency and leveraging swift’s Optional<T> and Array<T> types to model ? and *, respectively? string-based regular expressions are so limiting after all, and most of the time you want to do more with capture groups than just match a substring. i have a working sketch of this idea here but of course, it still needs a lot more design work to refine it into a usable framework.

3

Thank you for your response.

Are you suggesting that I implement my own regex engine? I wouldn't even know where to begin doing something like that. And If I did, I would want it to have all of the same features as NSegularExpression, such as lookahead and lookbehind. That sounds extremely complicated and beyond my skillset.

I'm not sure what a non-string-based regular expression pattern would look like. Could you give me an example? And how does swift's Optional type fit into this?

4

it’s not as hard as it sounds, if you start with the basic features and work your way up to the more advanced functionality :)

let’s say you want to parse an tv episode descriptor like your last example. to make things interesting, you might want to allow a user to specify both a single episode and a range of episodes. but if a show only has one season, a user might only specify the episode number. so a string-based regex for that might look like this:

/\s*(season\s+(\d+)\s*,\s*)?episode\s+(\d+)\s*(-\s*(\d+)\s*)?/
     ^~~~~~~~~~~~~~~~~~~~~            ^~~~~   ^~~~~~~~~~~~~~~
              ^~~~~              episode start     ^~~~~
              season                             episode end 
     optional capture group 0               optional capture group 2

which would produce

((season:Substring)?, episodeStart:Substring, (episodeEnd:Substring)?)

let episodeStart = capture.episodeStart
guard let season = capture.0?.season, let episodeEnd = capture.2?.episodeEnd
...

and accept all of the following strings:

"season 1, episode 1"
" season 1 ,episode  1"
"  episode  20 "
" episode  9- 2"
"season 5 , episode 2-4"
"season 5, episode 2 - 4"

but it would be a lot easier to use if we could express this regex pattern using swift, instead of a string literal using its own niche syntax. what if we could do the following?

struct Digit:Parseable 
{
    // regex for [\d]
}
struct Space:Parseable 
{
    // regex for [\s]
}

struct Season:Parseable.Terminal 
{
    static 
    let token:String = "season"
}
struct Episode:Parseable.Terminal 
{
    static 
    let token:String = "episode"
}
struct Comma:Parseable.Terminal 
{
    static 
    let token:String = ","
}
struct Hyphen:Parseable.Terminal 
{
    static 
    let token:String = "-"
}

// parser for /\d+/
struct Integer:Parseable 
{
    let value:Int 
    
    static 
    func parse(_ context:ParsingInput) throws -> Self 
    {
        let head:Digit      = try .parse(&context), 
            body:[Digit]    =     .parse(&context)
        // pretend we have an init that takes a digit sequence
        return .init([head] + body) 
    }
}
// parser for /\s+/
struct Whitespace:Parseable 
{
    static 
    func parse(_ context:ParsingInput) throws -> Self 
    {
        let _:Space         = try .parse(&context), 
            _:[Space]       =     .parse(&context)
        return .init() 
    }
}
struct Title:Parseable 
{
    let season:Int, 
        episodes:Range<Int>
    
    static 
    func parse(_ context:ParsingInput) throws -> Self 
    {
        let _:Whitespace?   =     .parse(&context), 
            season:
                List<Season, 
                List<Whitespace, 
                List<Integer, 
                List<Whitespace?, 
                List<Comma, Whitespace?>>>>>? = 
                                  .parse(&context), 
            _:Episode       = try .parse(&context), 
            _:Whitespace    = try .parse(&context), 
            start:Integer   = try .parse(&context), 
            _:Whitespace?   =     .parse(&context), 
            end:
                List<Hyphen, 
                List<Whitespace?, 
                List<Integer, Whitespace?>>>? = 
                                  .parse(&context)
        return .init(season: (season?.body.body.head.value ?? 1) - 1, 
            episodes: start.value - 1 ..< 
                end?.body.body.head.value ?? start.value) 
    }
}

of course, it would be the job of your library to define the protocols Parseable and Parseable.Terminal, implement List<T, U>, and to conform Optional<T> and Array<T> to Parseable. (hint, you can use do-catch to implement the requirement for Optional<T>, and use your Optional<T> implementation to implement Array<T>)

5

And how would I use these structs to parse data from an input string?

6

your implementation for the ParsingInput type in the example above would store the input, and the current buffer read position. so you might define a convenience method in an extension to Parseable that sets up the input buffer from a String argument.

7

I don't understand how any of this code works. Could you give me an example in which you use these structs to parse a string containing a season and episode number?