8 Tricks When Using the Fetch() API

Boss, and Scott

Welcome to Syntech. On this Monday, hasty treat. We're gonna be talking about 8 it. Awesome, fun, neat tricks to using the fetch API in JavaScript.

Wes has been doing a lot of fun stuff with fetch and, he wanted to share some of his hot tips on the fetch. Before we get going, syntax is presented by Century at century.io.

Use the coupon code tasty treat, all all lowercase, all one word, and get 2 months for free of it. The best error tracking, performance tracking, and everything application in the entire world. So let's get going here on 8 tricks For using the fetch API,

what you what you got? Ever since I did I did a talk on, like, standard server JavaScript, we did the episode six forty eight on standardized server JavaScript, you know, workers bun, working with Hano JS. We had Ryan on to talk about remix.

All of those are like, yeah, it's all fetch. It's all based on fetch, Specifically, request and response in streams. So we are like, I've been diving a little bit deeper Into that 1st part of fetch, meaning that, like, you know, like, fetch has the double. If you're, like, fetching some JSON, You have to do dot then and return dot JSON.

A lot of times, we're like, that's annoying. Like, just give me the data straight away.

And it turns out that that first response has a lot of handy dandy stuff in it. Wait. What? And we're not we're just supposed Supposed to ignore it? We're supposed to Yeah. Actually learn about what that is? Okay. Yeah. There's there's some kind of cool stuff in there. And as we get into streaming and Error collection and whatnot. So these these are 8 different tips that I've I've come up with. So the first one is, you can stream the result, meaning that When you fire off a fetch request, what is returned to you is a promise, but that promise resolves To what's called a response, and in that response is not necessarily the data yet. So if you fire off a fetch request for, like, MP 3 of syntax, and you wanna download the entire thing. This is something I was doing with the transcript results because I needed to download the syntax m p 3 as well as 2 other pieces of audio that I needed to tag on to the end.

And if you fire off a fetch request for an MP 3, you almost immediately get the result back.

But You don't have the entire MP 3 just yet, and that's because it's still downloading. So that first response is sort of it has a bunch of things. It has some headers in it, but it also has the stream, which is handy if you the next tip is going to be how to actually handle progress. So If you are streaming in some data, you can decode that stream and start to view it. So if it is text That's being streamed in. You can start to stream it stream in the data.

Could be progress, could be any number of things that is being sent from the server.

So you can use different stream APIs. Go back to the episode we had there to be able to do different things with them. You can even Zip them up with the gzip stream, and and encode them that way. But that brings me to the next one, which is before we do that. So you're saying that you could have a request somewhere.

And without even looking into that request, you could stream it directly to a ZIP That is being downloaded Yeah. Exactly. And and zip it as it's being streamed in. That's wild. Yeah. And that's that's how the

GitHub, is What they download works? Zip works, right? But you could that's how a lot of things work is as it is being streamed in, You can pipe it into a zip file. So if you are maybe recording yourself in the browser with, like, a video API, You could be taking the output of that and putting it into a zip file. So by the time you hit end, You immediately have the zip file ready to download. You don't have to sit there and wait for the thing to zip itself up because it's been streaming into a zip file The entire time. Wow. So that's kinda the benefits of obviously, you wouldn't do that with a fetch request. But, if you were Doing that with a fetch request, you could pipe that into any number of different types of streams, be it decode the text so you can view it Or you like another example of what I wanted to do is, I have a button on the News Syntax website for Fetching the transcript, and that will take 30, 60 seconds. And it that should probably be thrown into a queue, and it probably will. But as I was just sort of debugging it, I wanted to, like, see the result of how it's going In the browser rather than just sit there and look at a loading indicator. And right now, what I have to do is I have to go to the dev tools and look at the console logs of the server. Right? Wouldn't it be nice if you could stream from the server saying, I'm okay. Now I've started encoding the file. Now I'm done Encoding the file. Now I've sent it off to the transcription.

Now I'm back from the transcription. Now I've sent it off to Chat GPT. Now I'm back from chat. You know, like, you can you can give it updates as to how it's going rather than just sitting there And waiting for a loading indicator to finish.

That's awesome. Wow. Next one we have here is download progress.

This is something I was Just sitting in bed last night thinking about it, and I I woke up this morning and coded it.

So basically, like I said, when you're downloading an m p three or a large File.

The first response will give you response dot body, and the body may not be done downloading yet because it's a larger file. So what you can do is, first of all, you can read the how big the file is you are downloading in in total. That's with a header. So you just say headers dot get content length.

Then you can loop over each what's called a chunk, a piece of data that has come in.

You can use the 4 await loop, which is, not something I've used a whole lot, and I've I've been starting to use it a lot more as I've been Getting into streams, but the for await loop will basically just say, like, wait for the next chunk.

And when it when you're done downloading the next chunk, Run the next iteration of this loop.

In turn. And what you can do is you can say, alright. Well, the first time I got a chunk, I got This many bytes. So you know I have this many bytes so far, and it's this big in total. So, therefore, I can calculate your 1%, 2% done, and that's that's really handy to be able to do. And so I realized that The the thing with web streams is if you read or work with a web stream It consumes it. Right? It consumes it. You can't use that again. Right? And there's there's a property on there called body used or done Mhmm. That you can access to see if it is done. But I we talked about it on the stream show with the dot t. There's also a dot clone method, which you can basically clone a stream. So what I did here is I cloned it.

I used one of them to actually view the progress, and I used the other one just to return the same as is. I didn't wanna, like, You can collect the chunks and put them in an array and then put the array together and put them into a blob and then return the the value. But that's annoying because, like, Once you start fussing with the parts of some data, like, you're responsible for all of it. I'd rather the browser just do it all for me. So what you can do is you can just clone it. You use 1 to, like, watch for the progress. And then once all the progress is done, you just return the original untouched Version of that. I have a tweet I'll link up in the show notes here of the actual code behind it. And it was really nice because you still can just use your regular await fetch to return the data, but you can tag a dot then log progress onto the end of any fetch, and you immediately have A progress indicator being streamed in, and you could you could put that into like a React store, Svelte store or something like that, which would be reactive. You could display it to the user. In In my case, I was just console logging the actual data. Nice. I I did, this clone business. I had a middleware hook where I didn't wanna consume

The, response, but I wanted to access and and throw the data into, like, locals, or something inside of SvelteKit for actions. So I made a little middleware. That's what the 1st time I hit that issue where it's like, oh, you consumed the response already.

You can't do that

If you wanna use it twice or whatever. It's it's kinda weird how you can't use it twice. It makes sense. And we talked about that on the streams as why you can't, but the the error message is hilarious The body is not usable.

Yeah. The body is not usable.

Next one we have here is you can cancel Streams.

This is something you used to not be able to do. It's been around for a couple years, so it's in all the browsers, in all the server things. But the way that works is you create a abort controller, and that gives you sort of like a handle. And then when you have a fetch, you pass it a signal, which is controller dot signal. And then if you want to ever cancel A fetch request, you simply just run controller. Cancel and that will stop the or sorry controller. Abort And that will stop the fetch request from running, and that that can be really handy. It's a little bit weird that you now have to have 2 different things, and And that's why a lot of people will reach for a library or at the very least, just a nice little wrapper function that would return The something with a cancel request

and a promise on it as well. What's what's the use case? Or, like, you would typically be doing this for your Streaming a lot of data, and then you needed to abort and do something else. You wanted to have, like, a user initiated cancel or something. I I think the the most

Common example that I can think of is the autocomplete text. So you type into a search box and then you type a little bit more.

If you fire off 2 requests, there's a possibility that 1 would come back after the other and not in the order that you're expecting.

So What you would do is you cancel any in flight fetches and then send the next one over.

It would also work on server if you're downloading a large MP 3 and then for some reason you you realize, oh, I no longer need this thing. Or maybe you're trying to download from 2 sources and you're not sure which one will be faster. If one finishes first, then cancel the other one. Interesting. Next one we have here is testing thing if JSON is returned. So one funky thing about JSON is if the server throws an error, You might get HTML back, and that is weird because you're expecting JSON. Like, you might you might be getting, like, You need to log in 1st, and they send you to their HTML login page when you're expecting JSON to come back.

And or you might be simply getting, like, a 500 internal error or whatever.

So one little trick that I've used in the past, Because you don't want the thing to choke when you call JSON dot parse on HTML. That's the classic.

Like, if you JSON parse HTML, you get, like, the open angle bracket error. It says, hey. I was not expecting open angle bracket. Right? So what you can do Is you can use instead of dot JSON, you can use dot text, and that will give you The actual text that's resolved, then you can say Parse it or wrap yeah. You can wrap your JSON parse in a try catch.

I In case there, we'll try parse it. And if it parses, great. Then we have we have an object.

And if it doesn't parse, Then, you know, Okay, something went wrong and maybe you can run that text through.

You can you can regex it. You could run it through like a DOM a Parser library, or you could display it to the user. You could just tell the user, something went wrong. Yeah. Ideally, you would get a JSON response that says something went wrong. But I was gonna say yeah. When you're working with third party APIs, you don't necessarily always have the option to to make it do what you want it to do. Right? Yeah. Next 1. Async, await, catch. I've said this a couple times, is you can mix and match await and .catch.

So if you have a fetch or or even you can you can mix and match await and .then.

So I'll often do this where I'll run a fetch and then I'll tag a little dot then response response that JSON, Right. Like you look, I just want the JSON. I want it in one line. You can tag a dot then dot JSON on the end, And then you can use an await to get the syntax that you do want. So it is totally fine to mix and match await and the chaining dot then promise syntax. Especially, in my case, I'll often use await to get the data, happy path, And I'll use dot catch to actually catch any error. I find that to be a bit nicer than having to wrap the whole thing in a try catch.

Yeah. I do I I like that a little bit better myself as well. And and try catch is one of those things that just straight up feels

annoying. Yeah. Unless it is Explicitly in a function to handle errors or explicitly in something like I was talking about earlier to try to parse it. I don't like putting try catch in my regular logic code because it messes the flow up and messes the scope up. All of that stuff is really annoying. Yeah. It's annoying to write.

So the next one is is something that we talked about on the Rust episode, which I think is coming out on the Wednesday after this, and that is If you want, in many cases, you want both the error and the happy path data result On the same level, you can wrap your promise functions, specifically in fetch in this case, in Another function that will try basically, this is a function that takes in a promise And we'll try to await that promise, and then it'll wrap it in a try catch. And then that function will either return An error.

It'll return a tuple. The first thing is going to be an error, and the second thing is going to be the data. And it's either gonna return to you An error and no data or no error and the data. Right? It's gonna return 1 or the other. And this is what Rust does with their response. Right? Or what's it called? Reply? Okay error? Okay or error. Yes. Okay or error. And it's nice because you don't have to, like, have a callback of dot then. You don't have a try catch. It simply returns to you the error or the data at a single high level. And you can return this in an array tuple or you can return it in an object it. With an error and a success property, whatever is, sort of what you like. So I have A video on YouTube that shows how to write one of these yourself. I have a video in my upcoming TypeScript that shows how to type this because It's very generic. Right? You could you could pass literally any promise function into it, and and the types have to be inferred all the way through.

And then there's also a really popular library out there called Await2JS, And it will do that. You just basically wrap it in. And, honestly, I think this is how it should have worked from the get go. Yeah. Because, essentially, you say, alright. I wanna try to run this function. The next line, I wanna check if there's an error. And if there is, I'll do something. Otherwise, After that checking of the error, then I have my data, and I can go ahead and do whatever it is I want. Yeah. I I prefer this slow. Next Tip we have here is DevTools. You can right click, click copy as fetch. So often if you want to replicate a request that has happened in dev tools. You can right click copy as fetch.

Firefox dev tools will Automatically copy all headers on with you, and then I believe the Chrome Dev Tools has one that's copy as Node. Js fetch, Which I think it will also take cookies along for for the ride, which is really nice. So if you are ever trying to do any sort of scraping or reverse engineering, Copy as fetch is a or or base or even just try to debug how does something work. What are the pieces that are needed to make this entire Request work in a silo that is going to give you all of those pieces, and you don't have to, like, guess. Oh, does it need this cookie? It doesn't need that. It will give you all the pieces and then you can kind of go from there.

Last tip I have here is you can programmatically create request And response and header objects, this one's not all that useful. But if you're trying to build a fetch request and you want all the nice typings and everything that goes along with that, you can use new request, new response, and new headers, And you'll get both the autocomplete for that as well as the Headers API has a really nice set of methods for checking, adding, removing, setting, All those good things that you're used to, and then you just pass your request to a fetch function. Yeah. That's sick.

I I like that. I've never done that myself, but I I like it,

would keep things clean. Yeah. It's it's not all that All that useful every now and then, but, you can do it. You sometimes need to programmatically craft a fetch request where you are setting different parameters on the fetch before you actually fire it off. And that's a great way to do it. Yeah. Totally.

Well, sick. I I thought this was really informative. I thought this was really cool, especially the stream stuff. That that stuff is the stuff that I I know the least about. But, You know, I I do love these flows in in working with the responses and errors. And, yeah, we we take Fetch for granted because we do the same thing over and over again with it. And,

It's, nice to see some some different stuff. Alright. That's it for today. Thanks for tuning in. We will catch you later. Peace. Peace.

Head on over to syntax.fm for a full archive of all of our shows, and don't forget to subscribe in your podcast player Or drop a review if you like this show.