Open mhrivnak opened 6 years ago
Yeah. It's a mess. Sorry.
By the way, I suspect the reason the Flask throughput was so terrible was because it was running through the Flask web server. In real life you use Gunicorn to run Flask and then the performance characteristics are super different.
Also, I ran the benchmark with Sanic (which is based on uvloop) and much more similar to NodeJS and it gives the same numbers as NodeJS.
All of this is poorly documented. The point was, I think, just to show how fast Go can be out-of-the-box.
I'm tempted to delete all mention of benchmarking because it wasn't very rigorously done.
The nodejs code is missing from the main post, even though you mention it and show test results. I did find the source in this git repo, but it would be valuable to post it on the main comparison page.
I also noted that the main comparison page includes a falcon-based implementation that you never mention or show test results for. But in my testing, it's faster than flask or tornado!