Support Parquet read of pages with Encoding::PlainDictionary and non-optional values

[mdrach]

I'm reading a parquet file generated by Snowflake and am hitting an error here:

arrow2/src/io/parquet/read/fixed_size_binary.rs

Line 230 in 9d4107c

(Encoding::Plain, _, false) => read_required(page.buffer(), additional, size, values),

Decoding "PlainDictionary"-encoded, dictionary-encoded required V1 pages is not yet implemented for FixedSizeBinary

Since PlainDictionary encoding with optional values is implemented, would it be simple to implement the non-optional version?
(Encoding::PlainDictionary, Some(dict), true) => ...

(I actually took a stab at this here but am missing a lot of context.)

[jorgecarleitao]

Hey. Thanks! Yes, that looks about right. I would just add an integration test against pyarrow here. What I usually do is:

1. add a new column here for the type
2. add the corresponding data here
3. add a test like this: https://github.com/jorgecarleitao/arrow2/blob/main/tests/it/io/parquet/read.rs#L139 targeting that column

so that we prove that we can read the type when it is written by pyarrow (which uses the c++ parquet implementation)

[jorgecarleitao]

Closed by #683