You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "plewis110 (via GitHub)" <gi...@apache.org> on 2024/04/04 21:44:10 UTC
[I] Zero Copy Support [arrow-rs]
plewis110 opened a new issue, #5593:
URL: https://github.com/apache/arrow-rs/issues/5593
**Is your feature request related to a problem or challenge? Please describe what you are trying to do.**
<!--
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
(This section helps Arrow developers understand the context and *why* for this feature, in addition to the *what*)
-->
Hi, I've been having trouble integrating with the [`arrow_ipc`](https://docs.rs/arrow-ipc/51.0.0/arrow_ipc/) crate due to the requirement that a [`RecordBatch`](https://docs.rs/arrow-array/51.0.0/arrow_array/struct.RecordBatch.html) object can only be constructed from a stream of data via the [`Read`](https://doc.rust-lang.org/nightly/std/io/trait.Read.html) trait. This is problematic when I have an already existing buffer of data in-memory which is expected to represent an entire and valid [`RecordBatch`](https://docs.rs/arrow-array/51.0.0/arrow_array/struct.RecordBatch.html) already.
For example, I have a [`bytes::Bytes`](https://docs.rs/bytes/latest/bytes/struct.Bytes.html) instance which has been provided from another library. `bytes::Bytes` implements `Read` trait but I would like to avoid the copy which is forced when an object which implements the `Read` trait is passed to [`arrow_ipc::reader::StreamReader::try_new_unbuffered(...)`](https://docs.rs/arrow-ipc/51.0.0/arrow_ipc/reader/struct.StreamReader.html#method.try_new_unbuffered).
Because Arrow prides itself on being a "zero-copy" format, I expected an interface which did not force a copy to be provided.
**Describe the solution you'd like**
<!--
A clear and concise description of what you want to happen.
-->
For both the [`FileReader`]() and the [`StreamReader`]() instances, the following methods would be helpful:
```rust
pub fn try_new_from_bytes(
buffer: bytes::Bytes,
projection: Option<Vec<usize>>,
) -> Result<Self, ArrowError>;
```
**Describe alternatives you've considered**
<!--
A clear and concise description of any alternative solutions or features you've considered.
-->
The [`Read`](https://doc.rust-lang.org/nightly/std/io/trait.Read.html) trait is not acceptable for any zero-copy use case due to its trait methods passing around byte-slices which provide no means of lifetime extension and by extension, no way to avoid a copy of the data:
```rust
pub trait Read {
// Required method
fn read(&mut self, buf: &mut [u8]) -> Result<usize>;
// Provided methods
...
}
```
**Additional context**
<!--
Add any other context or screenshots about the feature request here.
-->
It appears that under the hood the underlying array type also uses `bytes::Bytes`, so this probably wouldn't be too difficult of a change.
--
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.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Zero Copy Support [arrow-rs]
Posted by "plewis110 (via GitHub)" <gi...@apache.org>.
plewis110 commented on issue #5593:
URL: https://github.com/apache/arrow-rs/issues/5593#issuecomment-2040289238
> You want to use the lower level FileDecoder and StreamDecoder (not yet released)
Ah neat, just took a look at the implementation and that's exactly what I need. I'll also be able to delete some of my `Stream` parsing code with that interface.
Thank you!
--
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.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Zero Copy Support [arrow-rs]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold commented on issue #5593:
URL: https://github.com/apache/arrow-rs/issues/5593#issuecomment-2039227220
You want to use the lower level FileDecoder and StreamDecoder (not yet released)
--
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.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] Zero Copy Support [arrow-rs]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold closed issue #5593: Zero Copy Support
URL: https://github.com/apache/arrow-rs/issues/5593
--
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.
To unsubscribe, e-mail: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org