我们可以使用手动编写的Schema来序列化嵌套的结构。以下是示例代码：

from avro import schema
from avro import io
from avro import datafile

# 构建外层的Schema
outer_schema = schema.Parse("""
{
    "namespace": "example.avro",
    "type": "record",
    "name": "OuterRecord",
    "fields": [
        {"name": "inner", "type": "InnerRecord"}
    ]
}
""")

# 构建内层的Schema
inner_schema = schema.Parse("""
{
    "namespace": "example.avro",
    "type": "record",
    "name": "InnerRecord",
    "fields": [
        {"name": "value1", "type": "string"},
        {"name": "value2", "type": "int"}
    ]
}
""")

# 构建Record对象
inner_record = {"value1": "hello", "value2": 4}
outer_record = {"inner": inner_record}

# 序列化数据并写入文件
data_writer = io.DatumWriter(outer_schema)
data_file_writer = datafile.DataFileWriter(
    open("example.avro", "wb"),
    data_writer,
    outer_schema,
    codec="snappy"
)

data_file_writer.append(outer_record)
data_file_writer.close()

# 读取并反序列化数据
data_file_reader = datafile.DataFileReader(
    open("example.avro", "rb"),
    io.DatumReader()
)

for record in data_file_reader:
    print(record)
    
    # 访问内层数据
    inner_value1 = record["inner"]["value1"]
    inner_value2 = record["inner"]["value2"]

    print(f"inner_value1: {inner_value1}")
    print(f"inner_value2: {inner_value2}")
    
data_file_reader.close()

上述代码中，我们手动编写了外层和内层的Schema，并且使用DatumWriter进行序列化，使用DataFileWriter将数据写入文件，使用DataFileReader读取并反序列化数据。

值得注意的是，我们可以通过record["inner"]["value1"]访问内层数据。