Read ManifestList, Manifest

[ZENOTME]

For ManifestList, Manifest, I think #29 already as a good example to refer. We can define a unified view in memory and distinguish version1 and version2 in representation for serde.

And for the representation in memory, we can refer icelake.

[ZENOTME]

I'm glad to prepare PR for it.

[JanKaul]

Thanks for taking this on!

There is one issue with the upper & lower bounds. The upper and lower bounds are stored as bytes, which can only be interpreted with the schema. So we will not be able to deserialize directly into the unified view. There needs to be a function which takes the schema into account.

Alternatively we could store the bytes in the unified view, but I think that's not the way to go.

Serialization should be fine.

[liurenjie1024]

Maybe we should complete #37 first.

[ZENOTME]

There is one issue with the upper & lower bounds. The upper and lower bounds are stored as bytes, which can only be interpreted with the schema. So we will not be able to deserialize directly into the unified view. There needs to be a function which takes the schema into account.

Alternatively we could store the bytes in the unified view, but I think that's not the way to go.

Yes, I agree with you. Maybe we can store the bytes in representation of serde first. And then take the schema to interpreted them to unified view in memory.

[JanKaul]

You could deserialize into serde_bytes::ByteBuf and then use Literal::try_from_bytes.

[Fokko]

There is one issue with the upper & lower bounds. The upper and lower bounds are stored as bytes, which can only be interpreted with the schema. So we will not be able to deserialize directly into the unified view. There needs to be a function which takes the schema into account.

In other languages, we keep this as bytes before deserialization. The metrics can become quite large and therefore become pretty expensive.

[liurenjie1024]

Close by #79