Skip to content

Separate header parsing and block reading #20

@martindurant

Description

@martindurant

In order to allow parallel reading by blocks, would prefer to have the class be able to read the header on instantiation, and provide the schema and sync marker. This happens here in fastavro, could copy that.

I appreciate that in cyavro's case, the schema is held in C variables, which we cannot directly serialize. That means, that we need to be able to grab as many bytes as were required for the first part, and pass those around so that a reader with initialized vars can be recreated without having to open and interpret the same file.

The second part will be to be able to read a block of bytes between sync markers. That might need prepending the block with the bytes required to parse the header.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions