Gemini/GeminiFormat/GeminiParser.swift

//
//  GeminiParser.swift
//  GeminiFormat
//
//  Created by Shadowfacts on 7/12/20.
//

import Foundation

public struct GeminiParser {
    
    private init() {}
    
    public static func parse(text: String, baseURL: URL) -> Document {
        var doc = Document(url: baseURL)
        
        var preformattingState = PreformattingState.off
        text.enumerateLines { (line, stop) in
            if line.starts(with: "```") {
                switch preformattingState {
                case .off:
                    let alt: String?
                    if line.count > 3 {
                        alt = String(line[line.index(line.startIndex, offsetBy: 3)...])
                    } else {
                        alt = nil
                    }
                    preformattingState = .on(alt)
                case .on(_):
                    preformattingState = .off
                }
                if case .off = preformattingState {
                    
                }
            } else if case let .on(alt) = preformattingState {
                doc.lines.append(.preformattedText(line, alt: alt))
            } else if line.starts(with: "=>") {
                // Link line
                let urlStart = line.firstNonWhitespaceIndex(after: line.index(line.startIndex, offsetBy: 2))
                let urlEnd = line.firstWhitespaceIndex(after: urlStart)
                let textStart = line.firstNonWhitespaceIndex(after: urlEnd)
                
                let urlString = String(line[urlStart..<urlEnd])
                // todo: if the URL initializer fails, should there be a .link line with a nil URL?
                let url = URL(string: urlString, relativeTo: baseURL)!.absoluteURL
                
                let text: String?
                if textStart < line.endIndex {
                    text = String(line[textStart..<line.endIndex])
                } else {
                    text = nil
                }
                
                doc.lines.append(.link(url, text: text))
            } else if line.starts(with: "#") {
                let level: Document.HeadingLevel
                if line.starts(with: "###") {
                    level = .h3
                } else if line.starts(with: "##") {
                    level = .h2
                } else {
                    level = .h1
                }
                let headingStart = line.firstNonWhitespaceIndex(after: line.index(line.startIndex, offsetBy: level.rawValue))
                let headingText = String(line[headingStart...])
                doc.lines.append(.heading(headingText, level: level))
            } else if line.starts(with: "* ") {
                let listItemStart = line.firstNonWhitespaceIndex(after: line.index(line.startIndex, offsetBy: 2))
                let listItemText = String(line[listItemStart...])
                doc.lines.append(.unorderedListItem(listItemText))
            } else if line.starts(with: ">") {
                let quoteStartIndex = line.firstNonWhitespaceIndex(after: line.index(after: line.startIndex))
                let quoteText = String(line[quoteStartIndex...])
                doc.lines.append(.quote(quoteText))
            } else {
                doc.lines.append(.text(line))
            }
        }
        
        return doc
    }
    
}

fileprivate extension GeminiParser {
    enum PreformattingState {
        case off
        case on(_ alt: String?)
    }
}

fileprivate extension String {
    func firstNonWhitespaceIndex(after index: String.Index) -> String.Index {
        var index = index
        // using .unicodeScalars.first should be fine, since all whitespace characters are single scalars
        while index < self.endIndex, CharacterSet.whitespaces.contains(self[index].unicodeScalars.first!) {
            index = self.index(after: index)
        }
        return index
    }
    
    func firstWhitespaceIndex(after index: String.Index) -> String.Index {
        var index = index
        // todo: could the first unicode scalar of a character be whitespace even though the whole character is not?
        while index < self.endIndex, !CharacterSet.whitespaces.contains(self[index].unicodeScalars.first!) {
            index = self.index(after: index)
        }
        return index
    }
}