(arrow-rs) branch 53.0.0-dev updated: Remove `impl> From for Buffer` that easily accidentally copies data (#6043)

[alamb]

This is an automated email from the ASF dual-hosted git repository.

alamb pushed a commit to branch 53.0.0-dev
in repository https://gitbox.apache.org/repos/asf/arrow-rs.git


The following commit(s) were added to refs/heads/53.0.0-dev by this push:
     new 8f76248222 Remove `impl<T: AsRef<[u8]>> From<T> for Buffer`  that 
easily accidentally copies data (#6043)
8f76248222 is described below

commit 8f762482221a4b2104a770d1c132dea3b58c117d
Author: Xiangpeng Hao <haoxiangpeng...@gmail.com>
AuthorDate: Tue Jul 16 18:06:57 2024 -0400

    Remove `impl<T: AsRef<[u8]>> From<T> for Buffer`  that easily accidentally 
copies data (#6043)
    
    * deprecate auto copy, ask explicit reference
    
    * update comments
    
    * make cargo doc happy
---
 arrow-buffer/src/buffer/immutable.rs | 33 +++++++++++++++++++++++----------
 1 file changed, 23 insertions(+), 10 deletions(-)

diff --git a/arrow-buffer/src/buffer/immutable.rs 
b/arrow-buffer/src/buffer/immutable.rs
index 52e201ca15..c53ef18ba5 100644
--- a/arrow-buffer/src/buffer/immutable.rs
+++ b/arrow-buffer/src/buffer/immutable.rs
@@ -356,16 +356,29 @@ impl Buffer {
     }
 }
 
-/// Creating a `Buffer` instance by copying the memory from a `AsRef<[u8]>` 
into a newly
-/// allocated memory region.
-impl<T: AsRef<[u8]>> From<T> for Buffer {
-    fn from(p: T) -> Self {
-        // allocate aligned memory buffer
-        let slice = p.as_ref();
-        let len = slice.len();
-        let mut buffer = MutableBuffer::new(len);
-        buffer.extend_from_slice(slice);
-        buffer.into()
+/// Note that here we deliberately do not implement
+/// `impl<T: AsRef<[u8]>> From<T> for Buffer`
+/// As it would accept `Buffer::from(vec![...])` that would cause an 
unexpected copy.
+/// Instead, we ask user to be explicit when copying is occurring, e.g., 
`Buffer::from(vec![...].to_byte_slice())`.
+/// For zero-copy conversion, user should use `Buffer::from_vec(vec![...])`.
+///
+/// Since we removed impl for `AsRef<u8>`, we added the following three 
specific implementations to reduce API breakage.
+/// See <https://github.com/apache/arrow-rs/issues/6033> for more discussion 
on this.
+impl From<&[u8]> for Buffer {
+    fn from(p: &[u8]) -> Self {
+        Self::from_slice_ref(p)
+    }
+}
+
+impl<const N: usize> From<[u8; N]> for Buffer {
+    fn from(p: [u8; N]) -> Self {
+        Self::from_slice_ref(p)
+    }
+}
+
+impl<const N: usize> From<&[u8; N]> for Buffer {
+    fn from(p: &[u8; N]) -> Self {
+        Self::from_slice_ref(p)
     }
 }