Kevin Pis

@yingsu00 the DATA_DAY in parquet is byte_array， the default type in velox is VARBINARY

> @kevincmchen Will you please attach the Parquet file here? [bb4fc17563709e2c-39e8c8be00000000_2087878052_data.0.parq.zip](https://github.com/user-attachments/files/16177042/bb4fc17563709e2c-39e8c8be00000000_2087878052_data.0.parq.zip) @yingsu00 this is the Parquet file.

@majetideepak @Yuhta can you take a look ？

> Does the error from S3 indicate an invalid bucket? it will not indicate a invalid bucket error instead of some error like bucket is empty/null. > But I wonder...

@Yuhta @majetideepak the last commit caused poor performance in the Conbench performance report, besides the issue you commented. so I rewrote the code about how to get the requested type....

@Yuhta could you please help me review this pr

@majetideepak could you please help me review this pr ？

> @kevincmchen are you able to address this comment? [#10399 (comment)](https://github.com/facebookincubator/velox/pull/10399#discussion_r1681824613) I replied to this comment and @ you, please take a look。 @majetideepak

> This change looks good. I will add the merge label after fixing the `nit: auto&` I have fixed this issue `nit: auto&`, please take a look. @majetideepak cc @Yuhta

> @kevincmchen thanks! This should be merged soon. Thank you for helping me review code and provide a lot of good suggestions! @majetideepak @Yuhta