我正在使用postgres 9.2，我试图构造一个查询，以生成特定格式的json。我已经非常接近一个简单的json_agg表达式，但现在我被卡住了。
我有一个简单的三表模式，其定义是：

CREATE TABLE project (
    id      INTEGER PRIMARY KEY,
    name    varchar(128) NOT NULL,
    UNIQUE (name)    
);

CREATE TABLE test (
    id          INTEGER PRIMARY KEY,
    name        varchar(128) NOT NULL,
    project_id  integer,
    FOREIGN KEY (project_id) REFERENCES project(id),
);

CREATE TABLE data (
    id              INTEGER PRIMARY KEY,
    date_entered    timestamp with time zone NOT NULL,
    data            json NOT NULL,
    test_id         integer,
    FOREIGN KEY (test_id) REFERENCES test(id)
);

插入一些数据后，如下所示：

INSERT INTO project (id, name) VALUES (0, 'my_project');
INSERT INTO test (id, name, project_id) VALUES (0, 'test0', 0);
INSERT INTO data (date_entered, data, test_id) VALUES (TIMESTAMP WITH TIME ZONE '2014-04-15T09:34:41.454999 z', '["some", "data"]', 0);
INSERT INTO test (id, name, project_id) VALUES (1, 'test1', 0);
INSERT INTO data (date_entered, data, test_id) VALUES (TIMESTAMP WITH TIME ZONE '2014-04-15T09:34:41.454999 z', '["some", "data"]', 1);

我想构造一个查询，它返回：

{
  "test0": {
    "first_data": "2014-04-15 09:35:10.394+00",
    "data_points": 1
  },
  "test1": {
    "first_data": "2014-04-15 09:35:10.394+00",
    "data_points": 1
  }
}

我最接近这个解决方案的是这个查询：

SELECT
    json_agg(data) as data
FROM (
    SELECT
        test.name as test_name,
        min(data.date_entered) as first_data,
        count(data.id) as data_points
    FROM test
    INNER JOIN data on data.test_id = test.id
    INNER JOIN project on test.project_id = project.id
    WHERE project.name = 'my_project'
    GROUP BY test.name
) as data;

它返回以下内容：

[
  {
    "test_name":"test0",
    "first_data":"2014-04-15 09:34:41.454999+00",
    "data_points":1
  },
  {
    "test_name":"test1",
    "first_data":"2014-04-15 09:34:41.454999+00",
    "data_points":1
  }
]

我已经尝试了row_to_json和array_to_json的各种奇怪用法，但我似乎无法将test_name值转换为外部字典中的键。
这可能吗？我是不是在滥用postgres的json生成函数？