Kotlin是否从文件加载类的集合？

Question

In C++ I can load a collection of a class from a file using a friend operator>> overload:

friend std::istream& operator>>(std::istream& is_a, College& college_a) {
    return is_a >> college_a.id_,
        is_a.get(),
        std::getline(is_a,college_a.name_);
}

std::set<College> colleges {};
std::copy(std::istream_iterator<College> {std::cin},
    std::istream_iterator<College> {},
    std::inserter(colleges, colleges.begin())); 

from a file like this:

0707 Rowan Technical College
0980 University of Saskatchewan
1058 Belmont University
1072 Belmont Technical College

How would this be done with Kotlin?

Answer 1

There are many ways to parse a file and split the lines in Kotlin / Java. The next example uses a Regex to split the line into an id and the name for the college:

data class College(val id: String, val name: String)

val colleges = File("input.txt").useLines { lines ->
    val regex = Regex("(\\d+) (.+)")
    lines.mapNotNull { line ->
        regex.matchEntire(line)?.let {
            College(it.groupValues[1], it.groupValues[2])
        }
    }.toList()
}

If you need to read from the standard input, then replace the third line with:

val colleges = InputStreamReader(System.`in`).useLines { lines ->

Update: Hide details

There is no general way to parse text files into object in Java / Kotlin. For XML and Json files exists some standard APIs and you can register Serializer/Deserializer for each type.

If you need such an abstraction for simple line oriented text files you have to build it yourself. The next examples may give you an idea:

fun <T : Any> File.parseLines(lineParser: (String) -> T?): List<T> =
    useLines { it.mapNotNull(lineParser).toList() }

The function extend the File class with a parseLines method.

The implementation of the line parser for class College would be like this:

val collegeLineParser: (String) -> College? = { line ->
    val regex = Regex("(\\d+) (.+)")
    regex.matchEntire(line)?.let {
        College(it.groupValues[1], it.groupValues[2])
    }
}

Or if you like to cache the RegEx :

val collegeLineParserCachedRegex = object : (String) -> College? {
    val regex = Regex("(\\d+) (.+)")
    override fun invoke(line: String): College? =
        regex.matchEntire(line)?.let {
            College(it.groupValues[1], it.groupValues[2])
        }
}

Now you can invoke the method parseLines on a File :

val colleges = File("input.txt").parseLines(collegeLineParser)

Update: Registry for parsers

If you do not like to provide the lineParser for every call, you can create a registry:

object LineParserRegistry {
    val parsers = ConcurrentHashMap<KClass<*>, (String) -> Any?>()

    inline fun <reified T> register(noinline parser : (String) -> T?) {
        parsers[T::class] = parser
    }

    inline fun <reified T> get(): (String) -> T? {
        // force companion initializer
        Class.forName(T::class.java.name)
        return parsers[T::class] as (String) -> T??
    }
}

inline fun <reified T : Any> File.parseLines(): List<T> =
    useLines { it.mapNotNull(LineParserRegistry.get<T>()).toList() }

If you like to register the parsers automatically you need to add an companion object with an init method:

data class College(val id: String, val name: String) {
    companion object {
        init {
            val collegeLineParser: (String) -> College? = { line ->
                val regex = Regex("(\\d+) (.+)")
                regex.matchEntire(line)?.let {
                    College(it.groupValues[1], it.groupValues[2])
                }
            }
            LineParserRegistry.register(collegeLineParser)
        }
    }
}

This init method is invoked the first time the College class is used. But reflection does not count, therefore we had to add Class.forName(T::class.java.name) into the get method of the registry. This call forces the initialization of the companion object.

Now you can call the parser without any preparations:

val colleges = File("input.txt").parseLines<College>()