konverter/Sources/Konverter/Convert/SubtitleConverter.swift

import Foundation

struct SubtitleCue {
    var start: Double
    var end: Double
    var text: String
}

/// SRT ↔ VTT ↔ TXT, plus the offset tool.
enum SubtitleConverter {
    static func convert(
        source: URL,
        to format: Format,
        at destination: URL,
        cancellation: Cancellation? = nil
    ) throws {
        var cues = try parse(source)
        try cancellation?.check()
        cues = cues.map { cue in
            var copy = cue
            copy.text = SubtitleConverter.normalized(copy.text)
            return copy
        }
        switch format {
        case .srt, .vtt:
            try write(cues, to: destination, format: format)
        case .txt:
            try TextDocument.write(cues.map(\.text).joined(separator: "\n"), to: destination)
        default:
            throw ConversionError.unsupported("Subtitles cannot become \(format.displayName).")
        }
    }

    static func parse(_ url: URL) throws -> [SubtitleCue] {
        guard let raw = try? String(contentsOf: url, encoding: .utf8)
            ?? (try? String(contentsOf: url, encoding: .isoLatin1)) else {
            throw ConversionError.unreadable(url.lastPathComponent)
        }
        let ext = FileKind.canonicalExtension(url.pathExtension)
        switch ext {
        case "vtt":
            return parseBlocks(raw, timeSeparator: ".")
        case "ass", "ssa":
            return parseASS(raw)
        default:
            return parseBlocks(raw, timeSeparator: nil)
        }
    }

    private static func parseBlocks(_ raw: String, timeSeparator: Character?) -> [SubtitleCue] {
        let normalized = raw
            .replacingOccurrences(of: "\r\n", with: "\n")
            .replacingOccurrences(of: "\r", with: "\n")
            .replacingOccurrences(of: "\u{FEFF}", with: "")
        let blocks = normalized.components(separatedBy: "\n\n")
        var cues: [SubtitleCue] = []

        for block in blocks {
            let lines = block.components(separatedBy: "\n").map { $0.trimmed }.filter { !$0.isEmpty }
            guard let timeLineIndex = lines.firstIndex(where: { $0.contains("-->") }) else { continue }
            let timeLine = lines[timeLineIndex]
            let parts = timeLine.components(separatedBy: "-->")
            guard parts.count == 2,
                  let start = seconds(from: parts[0], separator: timeSeparator),
                  let end = seconds(from: parts[1], separator: timeSeparator) else { continue }
            let text = lines.dropFirst(timeLineIndex + 1).joined(separator: "\n")
            cues.append(SubtitleCue(start: start, end: max(end, start), text: text.isEmpty ? " " : text))
        }
        return cues
    }

    private static func parseASS(_ raw: String) -> [SubtitleCue] {
        var cues: [SubtitleCue] = []
        for line in raw.components(separatedBy: "\n") where line.hasPrefix("Dialogue:") {
            let fields = line.dropFirst("Dialogue:".count).components(separatedBy: ",")
            guard fields.count >= 10,
                  let start = seconds(from: fields[1], separator: nil),
                  let end = seconds(from: fields[2], separator: nil) else { continue }
            let text = fields[9...].joined(separator: ",")
                .replacingOccurrences(of: "\\N", with: "\n")
                .replacingOccurrences(of: "\\n", with: "\n")
            cues.append(SubtitleCue(start: start, end: end, text: text))
        }
        return cues
    }

    static func write(_ cues: [SubtitleCue], to url: URL, format: Format) throws {
        var output = format == .vtt ? "WEBVTT\n\n" : ""
        for (index, cue) in cues.enumerated() {
            if format == .srt {
                output += "\(index + 1)\n"
                output += "\(timecode(cue.start, separator: ",")) --> \(timecode(cue.end, separator: ","))\n"
            } else {
                output += "\(timecode(cue.start, separator: ".")) --> \(timecode(cue.end, separator: "."))\n"
            }
            output += cue.text + "\n\n"
        }
        try TextDocument.write(output, to: url)
    }

    static func shift(_ cues: [SubtitleCue], by offset: Double) -> [SubtitleCue] {
        cues.map { cue in
            var copy = cue
            let shift = max(0, cue.start + offset) - cue.start
            copy.start = max(0, cue.start + offset)
            copy.end = max(copy.start, cue.end + shift)
            return copy
        }
    }

    private static func normalized(_ text: String) -> String {
        var value = text
        for tag in ["<i>", "</i>", "<b>", "</b>", "<u>", "</u>", "<font>", "</font>"] {
            value = value.replacingOccurrences(of: tag, with: "")
        }
        return value
    }

    // MARK: - Timecodes

    static func timecode(_ seconds: Double, separator: Character = ",") -> String {
        let clamped = max(0, seconds)
        let hours = Int(clamped) / 3600
        let minutes = (Int(clamped) % 3600) / 60
        let wholeSeconds = Int(clamped) % 60
        let millis = Int(((clamped - Double(Int(clamped))) * 1000).rounded())
        return String(format: "%02d:%02d:%02d%@%03d", hours, minutes, wholeSeconds, String(separator), millis)
    }

    private static func seconds(from text: String, separator: Character?) -> Double? {
        let cleaned = text.trimmingCharacters(in: .whitespaces).replacingOccurrences(of: ",", with: ".")
        let components = cleaned.components(separatedBy: ":")
        var value: Double = 0
        for (index, component) in components.enumerated() {
            guard let number = Double(component) else { return nil }
            switch components.count - index {
            case 1: value += number
            case 2: value += number * 60
            case 3: value += number * 3600
            default: return nil
            }
        }
        _ = separator
        return value
    }
}