[Bug]: Formatting taking more than 500ms for single files

[Schlafhase]

Description

Csharpier has been quite slow for a while now (at least for me). It often takes more than 500ms to format a single file (it doesn't take much more time to format a whole solution). I use csharpier with conform.nvim but it also takes that long when I run it from the terminal.
Has anyone else noticed this or is it just me?

I'd appreciate if someone would look into this :)

Steps to Reproduce

Format a c# file using csharpier format.

Expected Behavior

Should take 0-100ms.

Actual Behavior

For me, it takes 300-1000ms

[belav]

Formatting a single file has always been slow because of the overhead of starting up the dotnet tool. (200-300ms for me) That said, there have been some issues that caused slow down around how csharpier deals with finding options and determining if a file is ignored or not.

1. Is the time it takes to format the same file significantly different from an old version to the latest version?
2. Is the time it takes to format a single file in a brand new empty git repo significantly different than the time it takes to format a single file in an existing project of yours?

[Schlafhase]

Hi, thanks for the fast answer!
Would it be possible to cut down the startup overhead by running something in the background? Like prettierd does it, maybe? Not sure how much effort that would take to implement but it might be worth it (it should be opt-in, of course).

1. I think it does take longer than older versions. I use csharpier with conform.nvim and a timeout of 500ms. In the past, it never timed out but now I often get a warning about csharpier timing out. However, this might have nothing to do with the version but rather the git repository as seen in 2.
2. I ran the formatter 10 times on the same file, one time in a git repository (9 commits but some big files like images and videos) and the other time in a folder. These are the results (The last line is the average of each benchmarks):

In a git repository
╭────┬───────────────────╮
│  0 │ 441ms 878µs 203ns │
│  1 │  439ms 754µs 47ns │
│  2 │ 440ms 756µs 872ns │
│  3 │ 439ms 190µs 427ns │
│  4 │ 431ms 653µs 542ns │
│  5 │  437ms 65µs 372ns │
│  6 │ 436ms 229µs 405ns │
│  7 │ 433ms 535µs 939ns │
│  8 │   435ms 2µs 709ns │
│  9 │ 439ms 800µs 653ns │
│ 10 │ 437ms 486µs 716ns │ <-- avg
╰────┴───────────────────╯

Not in a git repository
╭────┬───────────────────╮
│  0 │ 505ms 561µs 366ns │
│  1 │ 188ms 234µs 646ns │
│  2 │ 193ms 623µs 238ns │
│  3 │ 189ms 955µs 702ns │
│  4 │ 187ms 832µs 725ns │
│  5 │ 190ms 556µs 548ns │
│  6 │ 190ms 100µs 481ns │
│  7 │ 193ms 702µs 913ns │
│  8 │ 187ms 880µs 971ns │
│  9 │ 188ms 525µs 848ns │
│ 10 │ 221ms 597µs 443ns │ <-- avg
╰────┴───────────────────╯

So you can clearly see that the git repo makes it a bit slow. I think the first value of the normal directory is the highest because the later ones just use a cache, right? So that's 505ms for a modified file which is still pretty slow, I'd say.

I just ran the benchmark again, this time modifying the file before each format and the format times are indeed a lot slower:

Git repo
╭────┬───────────────────╮
│  0 │ 757ms 731µs 240ns │
│  1 │  758ms 268µs 79ns │
│  2 │ 750ms 933µs 594ns │
│  3 │ 745ms 157µs 348ns │
│  4 │  751ms 555µs 61ns │
│  5 │ 767ms 577µs 658ns │
│  6 │ 748ms 990µs 642ns │
│  7 │ 763ms 667µs 109ns │
│  8 │ 762ms 343µs 289ns │
│  9 │ 776ms 262µs 697ns │
│ 10 │ 758ms 248µs 671ns │ <-- avg
╰────┴───────────────────╯

Not git repo
╭────┬───────────────────╮
│  0 │ 509ms 611µs 967ns │
│  1 │ 499ms 206µs 825ns │
│  2 │ 502ms 501µs 401ns │
│  3 │  502ms 495µs 81ns │
│  4 │ 509ms 776µs 905ns │
│  5 │ 499ms 413µs 842ns │
│  6 │ 507ms 828µs 456ns │
│  7 │ 500ms 947µs 680ns │
│  8 │ 500ms 268µs 370ns │
│  9 │ 502ms 254µs 448ns │
│ 10 │ 503ms 430µs 497ns │ <-- avg
╰────┴───────────────────╯