tqwewe
Performance improvements
I'm wondering if there's some low-hanging fruit for performance improvements. I'm seeing a big performance slowdown after adding this extension on a real world project. Running prettier check takes 6.5x as long with this extension added on my Intel Macbook Pro. I tested using Node 12.16.1 and Node 14.6.0 with similar results.
I could dig around, but curious if others are seeing similar slowdowns. It seems like this plugin shouldn't be quite that computationally expensive.
Without extension
With extension
That is very slow I agree. I don't know of a good way to solve this.
For now I'm kind of waiting for https://github.com/prettier/prettier/issues/4424 to be solved - when this happens I'll probably rewrite this plugin (might be silly, but would be fun to tackle the rewrite in Rust if possible).
There is a sorter write by rust: https://github.com/avencera/rustywind, but it can't read the configuration file.
https://github.com/PutziSan/prettier-plugin-sort-class-names is faster but can't format vue files
@Acidic9 I would start by looking here. The algorithmic complexity of that underlying class name sort function is off the charts—so many trivial optimizations possible that would dramatically improve performance.
I've got a 8.84s check time without this plugin, which inflated to 168.10s once I installed it, little bit more than 19x slower. It's not like my machine is slow, Ryzen 5600x, 64GB of RAM @ 3200 MT/s, 512 GB Gen 3 NVMe SSD, and it's not like it's under a high load at the moment. At close to 3 minutes, this is kinda unusable, as it would make my git hooks and CI runs that much slower.
That said, give me a poke if you have a test version and want to see how much that puts a dent in it. Can't share the project as it's my work stuff, but I can give you the metrics for improving the speed of it. I'd like to see it be much faster so I could actually include it in our config.
@Spice-King I did some messing around a couple of days ago and tried using WASM to make it faster.. but after some tests it turned out that JavaScript was way faster. But the code I was testing was completely rewritten and different to this projects code. I'll make a Github repo with some instructions for testing.. it might be quite useful getting some statistics from you