Michael Sowka

deadpool-postgres with connections to multiple hosts

3

currently we use tokio_postgres order of connecting to hosts. This means we go in the order of the hosts. I would prefer to round robin over the hosts and allow...

ivf-hnsw

6

I stumbled over https://github.com/dbaranchuk/ivf-hnsw Did anyone tried to benchmark their implementation? Is there conzeptual difference between ivf-hnsw and hnsw from hnswlib or nswlib?

brotli compression

8

Many modern Browsers support brotli compression. We should add optional brotli compression. It does not need to be best compressionlevel. https://caniuse.com/#search=brotli

Allow html Comments when using html-min

3

I'm getting a yarte (feature=["html-min"]) error for ``` ``` obviously this html comment is more than a simple html comment and should therefore work. error I'm getting: "No use html...

benchmark with fxhash

1

ramhorns uses fnvhash as default, however the hasher is interchangeable. Could you add a benchmark with https://github.com/cbreeden/fxhash ? you have mentioned fxhash as faster alternative in https://maciej.codes/2019-03-03-ramhorns.html

Comparison with Tera and Askama

1

It seems ramhorns is pretty fast, however could you explain in some words, why it's faster than even askama?

I'm using deadpool as async connectionpool for tokio-postgres. Can we add a test for compatibility and maybe an example? alternativly using feature-flags to enable such pools?

stroke size controlling

8

In your paper you mention changing the decoder of adain to change the stroke size. what is the difference between the decoders e.g. "decoder_stroke_perceptual_loss_1.pth" ? In my case, I would...

BTreeMap vs HashMap

1

I have tested an implementation where I replaced the hashmap with btreemap. (https://github.com/godofdream/ramhorns/tree/btreemap) It seems that in most cases the btreemap is faster on my machine (AMD Ryzen 5 3600...