Will Jones

Results 534 comments of Will Jones

> This seems very promising to me, but why was it only implemented in Rust 4 years ago? ([ARROW-2385](https://issues.apache.org/jira/browse/ARROW-2385), https://github.com/apache/arrow/pull/1829) Ah it was actually specifically only for integration tests, but...

I started by working on the Take implementation for primitive values, so I could familiarize myself with how the Take kernels work. But based on the benchmark I added, it...

> The numbers seem a bit low to be honest You are correct on that. Both too low in test and baseline, by about the same factor. I was creating...

After some more testing it seems like concatenating buffers and indexing into that always wins over using `ChunkResolver`, both in primitive and string case. From my quick test of Take...

> I'd be curious how the benchmark compares when using a large number of chunks: 10, 100, 1000 which is where the concatenation penalty is noticeable. Obviously, the sizes of...

> @wjones127 These numbers are for the random or monotonic use case? That's for random. Here is it including monotonic, which makes it more complex: ![results](https://user-images.githubusercontent.com/5488879/191867122-faa5e3dd-e274-4767-95cd-9e96e635681d.png) So it seems like...

I was having this problem as well (same OS and R version) and increasing the timeout startup setting did not work in my case. I ran the extension in debug...

Sure. Here is testing with default settings: ``` [info....] [21:28:34.255] [ ] Custom server listening on localhost:55002 [info....] [21:28:34.297] [ ] Request handled in VS Code: [info....] [21:28:34.297] [ ]...

Nevermind about the shell vs executable idea. I tested on an Intel Mac (OSX 11.6, R 4.0.2) and it works just fine with default settings. I'll do some more testing...