HTTPX Tutorial - A next-generation HTTP client for Python
Vložit
- čas přidán 27. 07. 2024
- In this Python Tutorial we learn about HTTPX, a modern HTTP client for Python with async support.
Resources:
www.python-httpx.org/
www.twilio.com/blog/asynchron...
Get my Free NumPy Handbook:
www.python-engineer.com/numpy...
✅ Write cleaner code with Sourcery, instant refactoring suggestions in VS Code & PyCharm: sourcery.ai/?... *
⭐ Join Our Discord : / discord
📓 ML Notebooks available on Patreon:
/ patrickloeber
If you enjoyed this video, please subscribe to the channel:
▶️ : / @patloeber
~~~~~~~~~~~~~~~ CONNECT ~~~~~~~~~~~~~~~
🖥️ Website: www.python-engineer.com
🐦 Twitter - / patloeber
✉️ Newsletter - www.python-engineer.com/newsl...
📸 Instagram - / patloeber
🦾 Discord: / discord
▶️ Subscribe: / @patloeber
~~~~~~~~~~~~~~ SUPPORT ME ~~~~~~~~~~~~~~
🅿 Patreon - / patrickloeber
#Python
Timeline:
00:00 Introduction
00:47 Installation & Features
02:25 Basic Usage
03:55 Advanced Usage with Client
05:27 Async API
06:20 Web App Testing
07:33 Speed Tests
----------------------------------------------------------------------------------------------------------
* This is an affiliate link. By clicking on it you will not have any additional costs, instead you will support me and my project. Thank you so much for the support! 🙏
What's your favorite HTTP library in Python?
httpx
requests
aiohttp
requests
Aiohttp because you can change proxies per request in a client rather than having to set the proxy at the time of creating the client. Httpx would be the better option if they had that one feature
This was brilliant, the more involved topic, the comparisons, discussing correct practices, everything! Keep it up!
Small nitpick, can you please use time.perf_counter() instead of time.time() since a lot of beginners watch your videos and it's the better function
Why is it better?
@@cobyiv because perf_counter uses cpu clocks to measure the difference of two function calls, while time.time returns the system time, which is subject to changes by the operating system.
@@cobyiv the system time, also called as ‘real world time’, is always subject to changes because your operating system periodically tries to calibrate the clock with clock synchronization protocol. You should not use system time to measure the difference of two time points because the time measured latter can be smaller than the time measured earlier due to these changes.
@@cobyiv check out pep418 peps.python.org/pep-0418/#rationale time.perf_counter() has the highest resolution for time deltas and is the one that should be used.
Could you instead use time.process_time()?
Great content, looking forward for more.
It's funny I recently ran into domaintools switching from requests to httpx and didn't know why. Thanks for the run down!
I love Python! ... the simplicity of Pythonic ethos reverberates through even 3rd party Python modules
Thx Python Engineer for making High Quality Videos!
Thankx for breaking it down.. great tutorial for beginners
I used this in our project… best library for async and sync requests
Thanks mate !! Great one
Brilliant video!!
GREAT CONTENT, THANK YOU!
Very helpful!
Excellent. Thanks.
This is a great video that came at the right time! I was working with an API for the past week and the part of the pipeline that took the longest was the APi request through the requests library.
Now I'm going to implement this HTTTPX client and also try to work in the Async feature! Thank you! You're awesome!
Interesting. Thanks
Any thoughts on whether additional code would be needed to handle a server's responses such as HTTP rate-limits?
Great video thanks! Now is this also able to handle well in a multiprocess async fashion? I am looking into aiomultuprocess and not sure if it would be better to run this over aiohttp.
wow thanks for the tip
I was trying to use this, but you can't pass context to event handlers. This means, for instance, that a response handler only knows the response, but it has no knowledge of what the request was. Handlers are only assigned at the client level (no per-request handlers). So to get different handling for each request, you need a new client. This breaks connection pooling. Ultimately, I went with aiohttp, as it allows passing context per request. Hopefully they add this feature to httpx.
have you tried http2 for making request ? what was the performance difference between http 1.1 vs 2 ? I think requests does not support h2 but httpx does.
It would have been great if you verified that all the 150 Pokemon were successfully gathered after each time trial. We are just trusting that the outcome of the 0.57 second async gather is equivalent to the 7 second synchronous gathering of Pokemon details. I believe you, but I am skeptical and an extra 2 or 3 seconds in the video to print the contents of the lists after would have been appreciated! Thanks
Nice.
httpx is very good library. amazing asyncio support. at my job we have replaced all httprequests in our code base with this. we used aiohttp and requests in the past.
Are there benefits compared to aiohttp? Performance should be comparable, but I'm curious if there is more... probably on testing flask / fastAPI as already mentioned?
so i think, as a HTTP client the bast solution for now is httpx, because AIOHTTP provides httpserver and httpclient as well.
aiohttp is faster than async httpx, but need more code to write and less intuitive (like requests library). and httpx supports http/2 which is faster
Hi, thanks for all useful info
Could you please make video list for your shorts videos
Well what is the difference between it and the requests library?
I always wondered if something like this would work to build backend webframework with
What is the VS code theme used ? I really like it !
What theme do you use in VS code?
I am new to your channel
nice demo, the requests does not support asyncio, so there is aiohttp can help. now, we have another library which much convenient than aiohttp. the httpx have the same style as requests..thanks !
Do you know why memory leak is happening when using in fastapi on linux?
So they combined requests with aiohttp? Don't think I'm going to bother changing my code to just have a single import.
try using timeit to time method execution, python time library is not meant to time method execution
How well does this intergrate with Django?
How Can i Use Place Method with Frames in Tkinter ?
Can you please give a lecture about 'with' ?
the last example sometimes is giving "RuntimeError: Event loop is closed" or "httpx.ConnectTimeout" but sometimes it is executing properly, I am using python 3.8. , windows 10,, can you please explain what is happening, and what is the solution?
got the solution , just need to increase the timeout param
Make a video about socket
But can't you use requests in asyncio via to_thread?
Looks cool but not enough features to switch from requests
can you build a desktop app.
determining the area of any engineering shapes
like rectangle ,..... .
Can you please make a video on Aws Cloud Front
How can I Use This http module with
Python
import httpx as requests. Boom no refactoring required
Hope some day, httpx supports (Async+Socks5 proxy)
99% modern internet is adult content or advertisement. I remember the times when a page opened in a browser was tens of kilobytes in size, and not like modern monstrous pages in megabytes and tens of megabytes.
I'm pretty sure there is ways to archive similiar result with request module and threading.
That's the method I currently use. But this looks to be faster.
And using async you dont need more threads to just wait requests. You can take advantage of a single thread on those kinda of tasks.
Speed comparison starts at 7:33 - czcams.com/video/qAh5dDODJ5k/video.html
Nice explanation Patrick, but 'Pokémon' is plural. Not 'pokemons'
Why does python use single-line quotes instead of double quotes when defining a string literal? Golang always uses doubles. What could be the reason for this choice?
You can use double or single-line quotes. For example print("text"more text"text") won't work, but print("text'more text'text") will work
@@tolia5587 I didn't mean it, it's all clear. I'm talking about the fact that it would be more logical to always use double quotes.
@@int0matar Yeah, I agree with you
Which faster httpx or aiohttp
Whats the point?
If request is working perfectly then i dont see the need for more bloat and dependencies.
🎯 Key Takeaways for quick navigation:
00:00 📖 Introduction to HTTPX
00:27 🚀 Getting Started with HTTPX
01:37 📜 HTTPX Features
02:58 🧪 Basic Functionality of HTTPX
04:05 🔧 Advanced Usage with HTTPX Client
05:27 ⚡ Asynchronous API in HTTPX
06:24 🧪 Testing Web Applications with HTTPX
07:33 🚀 Speed Comparisons
10:58 🎉 Conclusion
I'd still choose aiohttp over httpx for pure async code
wow
i think requests + asyncio.to_thread way simple and clean kkkk
Hi Patrick !
Bateman
its missing http/3
In the last example there is no need for `asyncio.create_task`. You can do it just adding coroutines to list `tasks.append(get_pokemon(client, url))`.
What's more - to make this code more "pythonic" you can do it like:
`pokemons = await asyncio.gather(*[get_pokemon(client, f"{base_pokemon_url}/{i}") for i in range(1, 151)])` :D
True. I tried it without asyncio.ensure_future and it worked prefectly.
Is this better than uvicorn?
Uvicorn Gunicorn is for server side hosting. In contrast, Request and httpx module is client side to request result from server.
It's even better than Excel
"There is no way to get async with requests" ... well there is, using greenlets and gevent.
threading with request would have beeen as fast...
New gen? You mean someone just made helper methods deal with this for you
MEMORY LEAKS: httpx.Client and httpx.AsynClient is leaking memory when you are not passing custom ssl_context. So if you use it in production, be cautious behind tutorials in vacuum space != your real code.
Nice, but honestly for a language that itself does not perform well and therefore always was the perfect wrapper and easy to handle, this async function (the right way) looks way to complicated. They should have made it WAY more simple.
No http/3…
This is useless without semaphores, you must divide the requests if you have many requests like over 1000
.
You're comparing parallelism to synchronous code. This is not really a fair comparison.
hell no why every new framework say "next generation framework" just stop saying just GO and do it or you RUST
trash expansions upon trash
you do thrash just for money sake, nothing useful, therefore trash
trash bloats around everything