Asynchronous Programming

Introduction

JavaScript is an asynchronous programming language in Node and in the browser. In many languages such as Java, C#, Python, etc. they block the thread for I/O. What this means is when you make an HTTP/ajax call or read a text file for example, the runtime will pause on that line of code until it is successful or failure.

JavaScript does the opposite. Using callbacks or Promises, you basically leave a phone number to call when those operations are done, while the rest of the synchronous code keeps going. In this article we’ll talk about why, give examples from JavaScript and compare against a blocking language, and show you some tips to help.

Why?

The V8 engine that runs in the Chrome web browser and Node, ensures a good user experience. This means, whenever you request a URL, load an image, etc. it doesn’t block the UI thread. The user can still interact with the web page while those asynchronous operations are going on. If it worked like blocking languages, the web page would lock up as things load, and buttons would only occasionally be clickable, text fields would stop working while you type and they run validation code, etc.

When you’re writing imperative code, it’s important you understand how this works so you don’t create hard to debug race conditions.

Real World Examples

Here’s an example of querying an Oracle database in Python:


connection = cx_Oracle.connect("hr", "welcome", "localhost/orclpdb")
cursor = connection.cursor()
cursor.execute("""
    SELECT first_name, last_name
    FROM employees
    WHERE department_id = :did AND employee_id > :eid""",
    did = 50,
    eid = 190)
for fname, lname in cursor:
  print("Values:", fname, lname)

Whether the first line that has the connect function takes 1 nanosecond or 1 minute, no other lines of code will run. The program will chill out until connect returns, and if successfully connected, then it’ll run line 2 to get the cursor from the connection. Connect, get cursor, execute the query, then print the results.

Here’s a broken example written in JavaScript:


var connection;
oracledb.getConnection({user:"hr", password:"welcome", connectString:"localhost/orclpdb"}, (err, conn) {
    if (err) {
      throw err;
      return;
    }
    connection = conn;
}); 
connection.execute(
    `FROM employees
    WHERE department_id = :did AND employee_id > :eid`,
    [50, 190],
    (err, result) {
      if(err) {
        throw err;
        return;
      }
      console.log("Values:", result.rows);
    });
});

We attempt to connect, and once the callback is called, we set the connection variable on line 7 so we can re-use later to execute the query.

Programmers from blocking languages might think the program will wait for the connection callback to fire before proceeding to the next. That’s not how it works. The worst part is that some implementations of callbacks, if they are fast enough, could make this work. Horrible race conditions that are hard to debug follow.

How Does It Work?

Let’s use a simple example in Python that prints out strings, in order, the code runs:


from __future__ import print_function
from time import sleep

def fast(): print('1')
def slow(callback):
    sleep(1)
    print('2')
    callback()

fast()
slow(lambda: print('3'))
print('4')

Python will print out 1, 2, 3, 4. The fast function is pretty straightforward in most programming. Slow uses the sleep function to pause. It works like all other I/O functions in that it takes awhile and pauses the execution of your code on that line. Then it executes the callback which is a lambda function below. Finally 4. Pretty imperative; line by line.

Here’s the same example in JavaScript:


fast = () => console.log('1');
slow = callback =>
    setTimeout(()=> {
        console.log('2');
        callback();
    }, 100);

fast();
slow(()=> console.log('3'));
console.log('4');

Node prints out 1, 4, 2, 3. Wat!?

The fast function makes sense. However, slow operates differently than Python. JavaScript never wait (alert doesn’t count). Any asynchronous calls, JavaScript will assume a callback will be passed in, or a Promise will be returned. When whatever long running task is done, that callback or Promise will be called. So the console.log('4') gets run immediately after slow.

Promises work in a similiar manner:


fast = () => console.log('1');
slow = () => new Promise(success => {
    console.log('2')
    success();
});

fast();
slow().then(()=> console.log('3'));
console.log('4');

It’ll print out 1, 4, 2, 3, same as before. The then or catch will fire whenever the long running process is done.

Callback vs. Promises

If you wish to skip nerdy nuances, head on down to the Conclusions.

There is one nuance of race conditions that callbacks can cause that Promises avoid, yet still have problems of their own.

Callback Race Condition

Let’s revisit our callback example above:

fast = () => console.log('1');
slow = callback =>
    setTimeout(()=> {
        console.log('2');
        callback();
    }, 100);

fast();
slow(()=> console.log('3'));
console.log('4');

It prints out 1, 4, 2, 3. Let’s remove the setTimeout and have the callback immediately invoke:

fast = () => console.log('1');
slow = callback => {
    console.log('2');
    callback();
};

fast();
slow(()=> console.log('3'));
console.log('4');

Oh boy, it prints out 1, 2, 3, 4. Depending on “how fast” your code runs, especially when unit testing with mocks, you now have a race condition. This is why a lot of implementations that use callbacks will often use defer functions internally to ensure the code always runs in the same order (i.e. waiting till the current call stack clears).

Promise Race Condition

Let’s revisit our Promise implementation:

fast = () => console.log('1');
slow = () => new Promise(success => {
    setTimeout(()=> {
        console.log('2');
        success();
    }, 100);
});

fast();
slow().then(()=> console.log('3'));
console.log('4');

It prints out 1, 4, 2, 3. Let’s remove the setTimeout and see what it what happens:

fast = () => console.log('1');
slow = () => new Promise(success => {
        console.log('2');
        success();
    });

fast();
slow().then(()=> console.log('3'));
console.log('4');

It prints out 1, 2, 4, 3. So bad news and good news. The good news, the asynchronous response still fires after your main code has run (console.log('4')). The bad news is the contents of the Promise itself will run. It looks like when you call the success, like the callback, it should be immediately calling your success which would fire the then, but it’s not. Like the defer, the Promise internally will wait until the current call stack has cleared, then running all pending successes or failures. More strange things for your brain to keep track of.

This can get imperative coders in trouble.

Async / Await

Despite the challenges with Promises, there are ways to make them operate more imperatively, and help prevent the race conditions that can occur for programmers who either haven’t spent enough time with Promises, and/or are from other languages where they already have this functionality such as C#.

Modifying our example above:

fast = () => console.log('1');
slow = () => new Promise(success => {
    setTimeout(()=> {
        console.log('2');
        success('slow is done');
    }, 100);
});

const go = async () => {
    fast();
    result = await slow();
    console.log('3, result:', result);
    console.log('4');
};

go();

It prints out 1, 2, 3, result: slow is done, 4. So you can see the code DID appear to “stop and wait” on the result = await slow() line. This can make your code more readable for async code you want to write in a synchronous manner.

Sadly, the error handling is gone, and you must wrap those blocks in try/catch. It’s a trade off. You can ensure this is less of a problem by ensure all Promises you create never fire catch, and the success method returns an Either.

Conclusions

As you can see, JavaScript is an asynchronous language by default. While many things still use callbacks, it’s better to use Promises. Even doing so, you can still create race conditions. Ensuring you chain them together, and/or utilize new functionality like async/await can help mitigate it. This is also why using const as opposed to var and let can force you to ensure your variables created at the right time.

To help remember, think “I’ll call you back later”. For Promise, think “I promise call then or catch later”.

4 Replies to “Asynchronous Programming”

    1. “You can ensure this is less of a problem by ensure all Promises you create never fire catch, and the success method returns an Either.”

      I was trying to google the example of solution with Either but all I see is examples in Scala, not any in RamdaJS. Can you please update the article with the code example of how to implement error handling in async/await with JavaScript Either?

      1. const maybeCow = cow =>
          new Promise( (success, failure) =>
            cow === 'cow' ? success('it is a cow')
            : failure(`it's not a cow, it is: ${cow}`));

        You pass in a String, and if it doesn’t equal ‘cow’, it’ll throw. That’s bad because once you go to using the async/await syntax, you now are forcing users to use a try/catch.

        Instead, you either create a Promise that always succeeds:

        const maybeCow = cow =>
        new Promise( success =>
          cow === 'cow' ? success({ok: true, data: 'it is a cow'})
          : success({ok: false, error: `it's not a cow, it is: ${cow}`}));

        When used as a Promise, there is no need to ever use the .catch, because it always returns a resolved Promise in the .catch:

        maybeCow('chicken')
        .then(result => console.log("result:", result))

        It’s more useful in the async/await, that’d look like:

        const {ok, error, data} = await maybeCow('chicken');

        Now your code looks more like Go or Lua in error handling.

        The other way is to just create a simple wrapper so “Look homey, I don’t know where I got this Promise, but uh, can you make sure it doesn’t throw and just let me know if it worked or not?”

        const sureThing = promise =>
          promise
          .then(result => ({ok: true, result}))
          .catch(error => Promise.resolve({ok: false, error}));

        And to use it:

        const go = async () => {
            const {ok, error, result} = await sureThing(maybeCow('chicken'));
            console.log("ok:", ok);
            console.log("error:", error);
            console.log("result:", result);
        };
        
        go();
        // ok: false
        // error: it's not a cow, it is: chicken
        // result: undefined
  1. Wow now I get it- these are very interesting techniques! Thank you for great examples and a fast reply too!

Comments are closed.