Dramatic performance regression in import/no-cycle

[Blasz]

We experienced a large performance regression in our linting after upgrading to 2.22.0 from 2.20.1.

Running linting on a single file.

Before upgrade:

Rule                                | Time (ms) | Relative
:-----------------------------------|----------:|--------:
import/no-cycle                     |  3382.383 |    91.2%
...

After upgrade:

Rule                                     | Time (ms) | Relative
:----------------------------------------|----------:|--------:
import/no-cycle                          | 73731.106 |    99.3%
...

I narrowed the problem down to this PR #1689. The isEsModuleInterop function tries to load a tsconfig file from disk and doesn't cache the result. The function is called for each file that is parsed by the plugin and is called multiple times in the same file. This is particularly bad when running import/no-cycle over a file with a large dependency graph as lots of files end up being parsed.

The fix would be to only lookup tsconfig once and cache the result.

[Blasz]

I'm happy to raise a PR for this.

…

On Thu, 12 Nov 2020, 2:52 pm Michael, ***@***.***> wrote: We experienced a large performance regression in our linting after upgrading to 2.22.0 from 2.20.1. Running linting on a single file. Before upgrade: Rule | Time (ms) | Relative :-----------------------------------|----------:|--------: import/no-cycle | 3382.383 | 91.2% ... After upgrade: Rule | Time (ms) | Relative :----------------------------------------|----------:|--------: import/no-cycle | 73731.106 | 99.3% ... I narrowed the problem down to this PR #1689 <#1689>. The isEsModuleInterop function tries to load a tsconfig file from disk and doesn't cache the result. The function is called for each file that is parsed by the plugin and is called multiple times in the same file. This is particularly bad when running import/no-cycle over a file with a large dependency graph as lots of files end up being parsed. The fix would be to only lookup tsconfig once and cache the result. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1943>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AARAYSWWFUNCIFTEAIEFUE3SPNLYZANCNFSM4TSYQ4CQ> .

[soryy708]

Here's an idea for how to make import/no-cycle faster:
#2937