如何根据特定条件锁定代码块？

Question

EDIT: I've added a table example (see google sheets link) and how the resulting apple object should look like.

I've programmed a multi-threaded web scraper using Jsoup, which extracts information from a website and saves it into a map. The main thing which i can't get to work is that the program does NOT connect to the website if it already scraped a certain information.

Information about the program

It extracts information from a table on a website and starts a thread for every word in the table.

So the threads get started with a certain word as class member. Every thread also has the same ConcurrentHashMap object. My plan was to check if the word already exists in the map as key.
If NOT, it should connect to a website to get information about the word, add some data to it and put it in the map afterwards.
If the map already contains the word, the thread should get the value from the map and only add the data to it.

So the main goal is NOT to connect to the website twice for the same word.

Here are the relevant code snippets:

Main class
Starting a thread for every word in the table. "element" contains the word and an url for more information about the word.

for (Element element : allRelevantTableElements) {
    executorService.execute(new Worker(element, data, concurrentMap));
}

Worker class
1. Check if word is already in map.
2a. If it is in map, just add data to it.
2b. If it is not in map, scrape information from website and then add data to it.

public class Worker implements Runnable {

MyWebScraper scraper;
Element element;    
String data;
ConcurrentMap<String, Fruit> concurrentMap;

public Worker(Element element, String data, ConcurrentMap<String, Fruit> concurrentMap) {
    this.element = element;
    this.data = data;
    this.concurrentMap = concurrentMap;
}

@Override
public void run() {

    Fruit fruit;

    if (concurrentMap.containsKey(element.text())) { 
        fruit = concurrentMap.get(element.text());
        fruit.addData(data)
    } else {            
        scraper = new WebScraper("http://fruitinformation.com" + element.attr("href"));
        scraper.connect();
        fruit = scraper.getInformation();
        fruit.addData(data)
    }

    concurrentMap.put(element.text(), fruit);
}
}

Example
Lets say the table looks like this:

https://docs.google.com/spreadsheets/d/1JF8sh8Sp9y0SV3Xb5mlISgcJp5s_DhaSp3KbnQLa248/edit?usp=sharing

The main class will start 3 threads:
Thread 1: Element contains "Apple" and the suburl "/apple",
Data contains "1,20€"
Thread 2: Element contains "Orange" and the suburl "/orange",
Data contains "2,40€"
Thread 3: Element contains "Apple" and the suburl "/apple",
Data contains "1,50€"

The problem is that all threads run almost simultaneously, so thread 1 and 3 will both check if "apple" is already in the map and BOTH will get false as result. So they BOTH connect to the website fruitinformation.com/apple and get the basic information about apples which i only need once. Then BOTH will add their data to the returned object and put it in the map, but thread 1 will do that first with "1,20€" and then thread 2 overrides the "1,20€" apple with his "1,50€ apple as the value.

However the goal is that only ONE apple thread connects to the website and adds his data(for example 1,20€) and then the other one realizes that an apple object already exists in the map and only adds his data(1,50€) to the existing apple. The fruit objects have Lists for that.
So the resulting map entry should look like this:
Key=Apple , Value= Fruit["Apple", basicInformationFromWebsite, List["1,20€"; "1,50€"]]

The other thread (orange) should run totally unaffected by all this. So all different fruits should run simutaneously but elements with the same fruit have to respect each other somehow. Is there a type of synchronization which only blocks instances with the same fruit names, but doesnt block any other instances?

---

I've read a lot about synchronization, locks, etc but can't find a solution for my problem.
It would be nice if someone can help me, thanks in advance!

Answer 1

XY problem. Synchronisation won't fix this. Even assuming you could implement it, the second thread would just be blocked by the first and then proceed to do the unwanted crawl.

You could add a Set of words that have begun to be processed, or add a dummy element into the map that shows it is already being processed although not complete.

Answer 2

If you get the total list of words first , then just pre-populate the map with placeholder values. then you only need to start threads for each of the keys in your map.

Answer 3

Not sure my answer is in line with how you've structured your app, but what follows is the "correct" way of handling your type of problem which is quite common in parallel applications.

It is certainly doable to obtain what you want and avoid "double" computation. I suggest you read java concurrency in practice and more specifically chapter 5 I think it is, where they have to do memoization of calculations (huge computations) and also have to avoid two threads calculating the same number.

Some tricks you can apply are to use putIfAbsent (method for only putting an item into a map if it does not already exist). More to the point however I suggest you store Futures in your map instead. https://docs.oracle.com/javase/7/docs/api/java/util/concurrent/Future.html They represent the result of a computation and then you will both have the computation ongoing and be certain that it will not be computed twice yet you will still get the result for both threads as you just call future.get() which blocks until a result is received. I will not go into much detail as it is in fact shown very nicely in the java concurrency book.

So something like (pseudocode)

if !map.containsKey(word) {
    Future f = new Future(word)
    map.putIfAbsent(word, future<curWord>)
    f.get()
} else {
    Future f = map.get(word)
    f.get()
}