What is Avro Schema?

An Avro schema describes a record: its name and its typed fields. It is the contract, not the data.

Who uses it

Kafka pipelines and anything in the Hadoop family.

Writing Avro Schema

Types are inferred per column from the values present. Any column with a gap in it becomes a union with null, because a reader rejects a record whose non-nullable field is missing.

What it can hold

Avro is a data interchange format. Avro carries its own schema, so the column types travel with the data instead of being inferred on the way in. A machine can read the table back out of it without guessing, which is what makes it safe to convert into and out of.

It keeps

Types. A number stays a number and a boolean stays a boolean, so whatever reads the file next does not have to guess from how a value looks.
The difference between empty and missing, which most table formats cannot express at all.
Nested values. A cell can hold a list or an object rather than only a single value.
A line break inside a cell, which survives rather than ending the row.
Its own record boundaries. A reader can take one row at a time without loading the file, and a copy that was cut off part way through still yields everything before the cut.

It cannot express

Any formatting. Bold, colour, column widths and merged cells have nowhere to go.
A header row. Column names are repeated on every record instead, which is why a file in this format is larger than the table it holds.

What it looks like

Example Avro Schema

{
  "type": "record",
  "name": "staff",
  "fields": [
    {
      "name": "id",
      "type": "long"
    },
    {
      "name": "name",
      "type": "string"
    },
    {
      "name": "role",
      "type": "string"
    },
    {
      "name": "started",
      "type": "string"
    },
    {
      "name": "hours",
      "type": [
        "null",
        "double"
      ]
    }
  ]
}

Questions

How do I open Avro Schema files?

Avro Schema is written by Tablizer rather than read by it. Open the file in whichever tool the format belongs to.

Is Avro Schema a good format for sharing data?

It depends who is receiving it. Kafka pipelines and anything in the Hadoop family. If the recipient is outside that group, a plain CSV travels further than anything else.