alamb commented on issue #97: URL: https://github.com/apache/arrow-rs/issues/97#issuecomment-826798446
Comment from Neville Dipale(nevi_me) @ 2020-08-25T03:59:35.414+0000: <pre>Once arrays are built, they're meant to be immutable. Wouldn't this better belong in ArrayBuilder?</pre> Comment from Francesco Gadaleta(frag) @ 2020-08-26T13:05:41.441+0000: <pre>Yes, it makes sense to me. Generally speaking if arrays are immutable, there are some operations that should directly modify elements at specific index. How are you going about that? </pre> Comment from Paddy Horan(paddyhoran) @ 2020-09-07T02:50:04.369+0000: <pre>Essentially, we are not addressing that. I don't believe any of the other implementations support this. It's kind of against the general principles of Arrow. However, there is nothing stopping you adding this for yourself in your application. You have access to underlying buffers. We are trailing behind the C++ impl in terms of what we support in Rust so we will likely focus on "catching up" for now.</pre> Comment from Jorge Leitão(jorgecarleitao) @ 2020-09-08T04:16:41.818+0000: <pre>> Generally speaking if arrays are immutable, there are some operations that should directly modify elements at specific index. if arrays as immutable, there should be _no_ operations that modify its elements, right? I also understood that arrays are immutable. Some general reasons is that this allows to pass slices of data by reference, both within a thread and across threads, without the need to worry about data races or threads waiting around. It is a whole computational model. In rust, it allow us to use {{Arc}} to safely share arrays across threads, as otherwise we would need a Mutex or other mechanism. To "modify" elements, we create a new array with the modified elements (see e.g. {{src/compute}}). </pre> Comment from Francesco Gadaleta(frag) @ 2020-09-08T07:13:00.133+0000: <pre>But that can be extremely inefficient. If one needs to change a dozen values in a column of millions of elements, that can become prohibitive. In-place value changes are quite a common operation in data science. </pre> -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org