jhorstmann commented on code in PR #9746:
URL: https://github.com/apache/arrow-rs/pull/9746#discussion_r3098982962
##########
parquet/src/encodings/rle.rs:
##########
@@ -514,16 +514,28 @@ impl RleDecoder {
break;
}
{
+ #[cold]
+ #[inline(never)]
+ fn oob(max_idx: u32, dict_len: usize) -> ! {
+ panic!(
+ "dictionary index out of bounds: the len is
{dict_len} but the index is {max_idx}"
+ )
+ }
+ const CHUNK: usize = 16;
let out = &mut buffer[values_read..values_read +
num_values];
let idx = &index_buf[..num_values];
- let mut out_chunks = out.chunks_exact_mut(8);
- let idx_chunks = idx.chunks_exact(8);
+ let dict_len = dict.len();
+ let mut out_chunks = out.chunks_exact_mut(CHUNK);
+ let idx_chunks = idx.chunks_exact(CHUNK);
for (out_chunk, idx_chunk) in
out_chunks.by_ref().zip(idx_chunks) {
- let dict_len = dict.len();
- assert!(
- idx_chunk.iter().all(|&i| (i as usize) <
dict_len),
- "dictionary index out of bounds"
- );
+ // u32 max-reduction instead of `.all(|&i| ..)`:
`.all`
+ // short-circuits and blocks autovectorisation.
Negative
+ // i32 cast to u32 becomes a large value so the
bounds
+ // check still rejects it.
+ let max_idx = idx_chunk.iter().fold(0u32, |acc,
&i| acc.max(i as u32));
+ if (max_idx as usize) >= dict_len {
+ oob(max_idx, dict_len);
Review Comment:
Now I remember there was a draft pr about improving the error handling
(#9365), so this could also be done separately.
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
