JSON

JsonDataSet reads and writes JSON dataset documents. Each top-level key corresponds to a table, where the value is an array of row objects. Each row is a JSON object of column:value pairs.

JSON dataset document sample:

{
  "TEST_TABLE": [
    {"COL0": "row 0 col 0", "COL1": "row 0 col 1", "COL2": "row 0 col 2"},
    {"COL1": "row 1 col 1"}
  ],
  "SECOND_TABLE": [
    {"COL0": "row 0 col 0", "COL1": "row 0 col 1"}
  ],
  "EMPTY_TABLE": []
}

To specify null values, just omit the corresponding column key. In the above example, the missing COL0 and COL2 keys of TEST_TABLE’s second row represent null values. Writing a dataset is the mirror image: any column whose value is null is omitted from that row’s JSON object rather than written as a JSON null.

In contrast to Flat XML, table metadata is deduced from the sum of all rows for each table, not just the first — so a column that first appears on a later row is still picked up.

Because columns are inferred entirely from row content, a table with declared columns but zero rows cannot carry that column information through JSON: an empty array has no rows to infer columns from, so writing such a table and reading it back produces a table with no columns instead. Formats that declare columns independently of row content, such as CSV’s per-table header row, do not have this limitation.

Table names are always case-sensitive — unlike some other formats, there is no case-insensitive table lookup fallback.

A table name repeated as a second top-level key throws AmbiguousTableNameException while the document is being read.

Load a dataset:

IDataSet dataSet = new JsonDataSet(new File("dataset.json"));

Write a dataset:

JsonDataSet.write(dataSet, new FileOutputStream("dataset.json"));

<dependency>
  <groupId>com.fasterxml.jackson.core</groupId>
  <artifactId>jackson-databind</artifactId>
  <version>VERSION</version>
</dependency>