Jack Grigg

Results 892 comments of Jack Grigg
trafficstars

I'd like to see if there's a way to make the grease more palatable before considering whether we merge this. I know that grease in non-interactive protocols like the age...

This is going to interact with #403. In a 30-second chat with @FiloSottile we figured it will make sense to define grease as having an empty label set, so it...

The logic they implement can be implemented outside the `age` crate, because `IdentityFileEntry` is a transparent enum. That being said, they are nice helpers to have available. The only issue...

Hmm, this might end up interacting with the refactor I need to do as mentioned in https://github.com/str4d/rage/pull/334#issuecomment-1236127574. So while I _can_ place the `callbacks` arg behind the `plugin` feature flag,...

I see you've pulled out the high-viscosity grease with a lithium complex soap thickener 😆 The performance issue is almost certainly this parser: https://github.com/str4d/rage/blob/3ff541a0a59864b782aeaca4d33cee1c8d61c2e9/age-core/src/format.rs#L191-L218 My expectation is that `full_chunks` is...

I measured a flamegraph, and indeed the perfomance issue is in `wrapped_encoded_data`, but it's not the allocations! ![image](https://user-images.githubusercontent.com/4993799/219985931-16232c0c-d594-464f-bf58-a18b93ca697a.png) The bottleneck is this function: https://github.com/str4d/rage/blob/3ff541a0a59864b782aeaca4d33cee1c8d61c2e9/age-core/src/format.rs#L145-L152 For a given slice, `is_base64_char` is...

I added a `Decryptor::new_buffered` API, and implemented a basic change to use `BufRead::read_until('\n', ..)` instead of reading precisely what the parser asks for. The effect on the header-parsing benchmarks (Ryzen...

#377 implements the above, which also required `impl io::BufRead for ArmoredReader` in order for `rage` to use it. I still need to implement `Decryptor::new_async_buffered` and `impl AsyncBufRead for ArmoredReader` to...

#393 implements `Decryptor::new_async_buffered` and `impl AsyncBufRead for ArmoredReader` to provide the equivalent performance improvements for the `async` APIs. I'm going to deploy these performance improvements in `rage 0.9.2`, and defer...