Reactor Flux：如何使用 header 解析文件

Question

Recently I started using project reactor 3.3 and I don't know what is the best way to handle flux of lines, with first line as column names, then use those column names to process/convert all other lines. Right now I'm doing this way:

Flux<String> lines = ....;
Mono<String[]> columns = Mono.from(lines.take(1).map(header -> header.split(";"))); //getting first line
Flux<SomeDto> objectFlux = lines.skip(1) //skip first line
  .flatMapIterable(row -> //iterating over lines
      columns.map(cols -> convert(cols, row)));  //convert line into SomeDto object

So is it the right way?

Answer 1

So is it the right way?

There's always more than one way to cook an egg - but the code you have there seems odd / suboptimal for two main reasons:

• I'd assume it's one line per record / DTO you want to extract, so it's a bit odd you're using flatMapIterable() rather than flatMap()
• You're going to resubscribe to lines once for each line, when you re-evaluate that Mono . That's almost certainly not what you want to do. (Caching the Mono helps, but you'd still resubscribe at least twice.)

Instead you may want to look at using switchOnFirst() , which will enable you to dynamically transform the Flux based on the first element (the header in your case.) This means you can do something like so:

lines
        .switchOnFirst((signal, flux) -> flux.zipWith(Flux.<String[]>just(signal.get().split(";")).repeat()))
        .map(row -> convert(row.getT1(), row.getT2()))

Note this is a bear-bones example, in real-world use you'll need to check whether the signal actually has a value as per the docs:

Note that the source might complete or error immediately instead of emitting, in which case the Signal would be onComplete or onError. It is NOT necessarily an onNext Signal, and must be checked accordingly.