Have you ever needed to generate a realtime stream of json data in order to test an application? When thinking about a good source of streaming data, we often look to the Twitter stream as a solution, but that only gets us so far in prototyping scenarios and we often fall short because Twitter data only fits a certain amount of use cases. There are plenty of json data generator online (like json-generator, or mockaroo), but we couldn't find an offline data generator for us to use in our testing and prototyping, so we decided to build one. We found it so useful, that we decided to open source it as well so other can make use of it in their own projects.
For more information, check out the announcement blog post.
We had a couple of needs when it came to generating data for testing purposes. They were as follows:
- Generate json documents that are defined in json themselves. This would allow us to take existing schemas, drop them in to the generator, modify them a bit and start generating data that looks like what we expect in our application
- Generate json with random data as values. This includes different types of random data, not just random characters, but things like random names, counters, dates, primitive types, etc.
- Generate a constant stream of json events that are sent somewhere. We might need to send the data to a log file or to a Kafka topic or something else.
- Generate events in a defined order, at defined or random time periods in order to act like a real system.
We now have a data generator that supports all of these things that can be run on our own networks and produce streams of json data for applications to consume.
The Generator has the following basic architecture:
JsonDataGenerator- The main application that loads configurations and runs simulations.Simulation Configuration- A json file that represents your over all simulation you would like to run.Workflow Definitions- Json files that define a workflow that is run by your Simulation.
When you start the JsonDataGenerator, you specify your Simulation Configuration which also references one or many Workflow Definitions. The Simulation is loaded and the Workflows are created within the application and each workflow is started within its own thread. Json events are generated and sent to the defined Producers.
There are multiple pieces of information that you as a developer/integrator need to define within your configuration files. We'll break down what goes where in this section.
The Simulation Configuration defines the following:
| Property | Type | Description |
|---|---|---|
| workflows | Array | Defines a list of workflows to run |
| producers | Array | Defines a list of producers that events should be sent to |
A Workflow is defined with the following properties:
| Property | Type | Description |
|---|---|---|
| workflowName | String | A name for the workflow |
| workflowFilename | String | The filename of the workflow config file to run |
| instances | Integer | the number of identical copies of instances to run (optional, default value is 1) |
Here is an example of a Workflow configuration:
"workflows": [{
"workflowName": "test",
"workflowFilename": "exampleWorkflow.json",
"instances": 1
}]
A Producer is defined with the following properties:
| Property | Type | Description |
|---|---|---|
| type | String | The type of the producer |
| optional properties | String | Other properties are added to the producer config based on the type of Producer |
Currenty supported Producers and their configuration properties are:
Logger
A Logger Producer sends json events to a log file in the logs directory into a file called json-data.log. The logs roll based on time and size. Configure it like so:
{
"type": "logger"
}
No configuration options exist for the Logger Producer at this time.
File
A File Producer sends json events to a file. One event will be written to each file in the specified directory. Configure it like so:
{
"type": "file",
"output.directory": "/tmp/dropbox/test2",
"file.prefix": "MYPREFIX_",
"file.extension":".json"
}
The File Logger will attempt to create the directory specified by output.directory.
HTTP POST
A HTTP POST Producer sends json events to a URL as the Request Body. Configure it like so:
{
"type": "http-post",
"url": "http://localhost:8050/ingest"
}
If you need to send data to an HTTPS endpoint that required client certificates, you can provide the configuration of those certificates on the commandline using the javax.net.ssl.* properties. An example might be:
java -Djavax.net.ssl.trustStore=/path/to/trustsore.jks -Djavax.net.ssl.keyStore=/path/to/user/cert/mycert.p12 -Djavax.net.ssl.keyStoreType=PKCS12 -Djavax.net.ssl.keyStorePassword=password -jar json-data-generator-1.2.2-SNAPSHOT.jar mySimConfig.json
Kafka
A Kafka Producer sends json events to the specified Kafka broker and topic as a String. Configure it like so:
{
"type": "kafka",
"broker.server": "192.168.59.103",
"broker.port": 9092,
"topic": "logevent",
"flatten": false,
"sync": false
}
If sync=false, all events are sent asynchronously. If for some reason you would like to send events synchronously, set sync=true. If flatten=true, the Producer will flatten the json before sending it to the Kafka queue, meaning instead of having nested json objects, you will have all properties at the top level using dot notation like so:
{
"test": {
"one": "one",
"two": "two"
},
"array": [
{
"element1": "one"
},{
"element2": "two"
}]
}
Would become
{
"test.one": "one",
"test.two": "two",
"array[0].element1": "one",
"array[1].element2": "two"
}
Kafka on kerberized cluster
Valid only for kafka version 0.10.2.0 or higher.
If you need to send data to kafka with kerberos you can provide the next configuration:
{
"type": "kafka",
"topic": "logevent",
"flatten": false,
"sync": false,
"kerberos": {
"kerberos.conf.file": "<path_your_kerberos_file_krb5.conf>",
"kafka.brokers.servers": "broker1:port,broker2:port,broker3:port",
"kafka.jaas.file": "<path_your_jaas_file_kafka_jaas.conf",
"kafka.security.protocol": "SASL_PLAINTEXT",
"kafka.service.name": "kafka"
}
}
kerberos.conf.file: Sets the system property 'java.security.krb5.conf'kafka.jaas.file: Sets the system property 'java.security.auth.login.config'kafka.security.protocol: Sets kafka producer property 'security.protocol'kafka.service.name: Sets kafka producer property 'sasl.kerberos.service.name'
In a kerberized kafka cluster, the keytab file must be defined inside JAAS file "kafka.jaas.file": "<path_your_jaas_file_kafka_jaas.conf"
Kafka on SASL cluster
Valid only for kafka version 0.10.2.0 or higher.
If you need to send data to kafka with SASL authentication you can provide the next configuration:
{
"type": "kafka",
"topic": "logevent",
"flatten": false,
"sync": false,
"sasl": {
"kafka.brokers.servers": "broker1:port,broker2:port,broker3:port",
"kafka.jaas.file": "<path_your_jaas_file_kafka_jaas.conf",
"kafka.security.protocol": "SASL_PLAINTEXT",
"kafka.sasl.mechanism": "SCRAM-SHA-512"
}
}
kafka.jaas.file: Sets the system property 'java.security.auth.login.config'kafka.security.protocol: Sets kafka producer property 'security.protocol'kafka.sasl.mechanism: Sets kafka producer property 'sasl.mechanism'
In a SASL kafka cluster, the username and password must be defined inside JAAS file "kafka.jaas.file": "<path_your_jaas_file_kafka_jaas.conf"
Kafka "advanced configuration" for kerberized or SASL cluster
Valid only for kafka version 0.10.2.0 or higher.
kafka.client.id: Sets kafka producer property 'client.id'kafka.compression: Sets kafka producer property 'compression.type' (none, gzip, snappy, or lz4)kafka.batch.size: Sets kafka producer property 'batch.size'kafka.acks: Sets kafka producer property 'acks'kafka.linger.ms: Sets kafka producer property 'linger.ms'kafka.request.per.connection: Sets kafka producer property 'max.in.flight.requests.per.connection'
NATS
A nats logger sends json events to gnatsd broker specifed in the config. The following example shows a sample config that sends json events to a locally running NATS broker listening on the default NATS port.
{
"type": "nats",
"broker.server": "127.0.0.1",
"broker.port": 4222,
"topic": "logevent",
"flatten": false
}
Note that flatten has the same effect as the option in the Kafka producer.
Tranquility
A Tranquility Producer sends json events using Tranquility to a Druid cluster. Druid is an Open Source Analytics data store and Tranquility is a realtime communication transport that Druid supports for ingesting data. Configure it like so:
{
"type": "tranquility",
"zookeeper.host": "localhost",
"zookeeper.port": 2181,
"overlord.name":"overlord",
"firehose.pattern":"druid:firehose:%s",
"discovery.path":"/druid/discovery",
"datasource.name":"test",
"timestamp.name":"startTime",
"sync": true,
"geo.dimensions": "where.latitude,where.longitude"
}
When sending data to Druid via Tranquility, we are sending a Task to the Druid Overlord node that will perform realtime ingestion. Our task implementation currently does not allow users to specifc the aggregators that are used. We create a single "events" metric that is a Count Aggregation. Everything else if configured though the properties. The sync and flatten properties behave exactly the same ast the Kafka Producer. This works for us currently, but if you need more capability, please file an issue or submit a pull request!
When you start the Generator, it will contact the Druid Overlord and create a task for your datasource and will then begin streaming data into Druid.
MQTT
An MQTT Producer sends json events to the MQTT broker specified in the config. The following example shows a sample config that sends json events to a locally running MQTT broker listening on the default MQTT port. The example also includes the two optional fields: username and password
{
"type": "mqtt",
"broker.server": "tcp://localhost",
"broker.port": 1883,
"topic": "/logevent",
"clientId": "LogEvent",
"qos": 2,
"username": "whoami",
"password": "whatsmypassword"
}
The MQTT producer support step specific configuration for QOS and Topic. The entire configuration and each item in it are optional. Add an "mqtt" item to the "producerConfig" map:
"mqtt" : {
"topic": "/elsewhere",
"qos": 1
}
Azure IoT Hub
An Azure IoT Hub Producer sends json events to the Azure IoT Hub specified in the config. Choose a protocol from HTTPS, AMQPS or MQTT
{
"type": "iothub",
"connectionString": "<- Get from Azure portal or Device Explorer ->",
"protocol": "HTTPS",
}
Full Simulation Config Example
Here is a full example of a Simulation Configuration file:
exampleSimConfig.json:
{
"workflows": [{
"workflowName": "test",
"workflowFilename": "exampleWorkflow.json",
"instances" : 4
}],
"producers": [{
"type": "kafka",
"broker.server": "192.168.59.103",
"broker.port": 9092,
"topic": "logevent",
"sync": false
},{
"type":"logger"
}]
}
This simulation will run four instances (same workflow rules but on different threads) of Workflow and send all the events to both the defined Kafka topic and to the Logger producer that put the events in to a log file.
Use instances to create multiple traffic of the same workflow.
The Workflow is defined in seperate files to allow you to have and run multiple Workflows within your Simulation. A Workflow contains the following properties:
| Property | Type | Description |
|---|---|---|
| eventFrequency | integer | The time in milliseconds events between steps should be output at |
| varyEventFrequency | boolean | If true, a random amount (between 0 and half the eventFrequency) of time will be added/subtracted to the eventFrequency |
| repeatWorkflow | boolean | If true, the workflow will repeat after it finishes |
| timeBetweenRepeat | integer | The time in milliseconds to wait before the Workflow is restarted |
| varyRepeatFrequency | boolean | If true, a random amount (between 0 and half the eventFrequency) of time will be added/subtracted to the timeBewteenRepeat |
| stepRunMode | string | Possible values: sequential, random, random-pick-one. Default is sequential |
| steps | Array | A list of Workflow Steps to be run in this Workflow |
Workflow Steps
Workflow Steps are meat of your generated json events. They specify what your json will look like and how they will be generated. Depending on the stepRunMode you have chosen, the Generator will execute your steps in different orders. The possibilities are as follows:
- sequential - Steps will be run in the order they are specified in the array.
- random - Steps will be shuffled and run in a random order. Steps will be reshuffled each time the workflow is repeated.
- random-pick-one - A random step will be chosen from your config and will be run. No other steps will be run until the workflow repeats. When the workflow repeats, a different random step will be picked.
Step
Now that you know how Steps are executed, let's take a look at how they are defined.
| Property | Type | Description |
|---|---|---|
| config | array of objects | The json objects to be generated during this step |
| duration | integer | If 0, this step will run once. If -1, this step will run forever. Any of number is the time in milliseconds to run this step for. |
| producerConfig | map of objects | Optional: producer configuration for this step - optional and specific for each producer. (See producer documentation) |
Step Config
The configuration specified in the step config section is the actual json you want output. For example, you could put the following json as the config for a step in your workflow:
{
"test": "this is a test",
"test2": "this is another test",
"still-a-test": true
}
Specifying this as the step config would literly generate that json object every time the step ran. Now, that's interesting, but just echoing json isn't really what we are after. Our Generator support a number of Functions that can be placed into your json as the value. These Functions will be run everytime the json is generated. Here is an example:
{
"timestamp": "nowTimestamp()",
"system": "random('BADGE', 'AUDIT', 'WEB')",
"actor": "bob",
"action": "random('ENTER','LOGIN','EXIT', 'LOGOUT')",
"objects": ["Building 1"],
"location": {
"lat": "double(-90.0, 90.0)",
"lon": "double(-180.0, 180.0)"
},
"message": "Entered Building 1"
}
This config will generate json documents that look something like:
{"timestamp":1430244584899,"system":"BADGE","actor":"bob","action":"LOGOUT","objects":["Building 1"],"location":{"lat":-19.4224,"lon":-165.0512},"message":"Entered Building 1"}
Each time, the timestamp, system, action, and lat/lon values would be randomly generated.
Full Workflow Definition Config Example
Here is a full example of a Workflow Definition file:
exampleWorkflow.json:
{
"eventFrequency": 4000,
"varyEventFrequency": true,
"repeatWorkflow": true,
"timeBetweenRepeat": 15000,
"varyRepeatFrequency": true,
"steps": [{
"config": [{
"timestamp": "nowTimestamp()",
"system": "random('BADGE', 'AUDIT', 'WEB')",
"actor": "bob",
"action": "random('ENTER','LOGIN','EXIT', 'LOGOUT')",
"objects": ["Building 1"],
"location": {
"lat": "double(-90.0, 90.0)",
"lon": "double(-180.0, 180.0)"
},
"message": "Entered Building 1"
}],
"producerConfig" {
},
"duration": 0
},{
"config": [{
"timestamp": "nowTimestamp()",
"system": "random('BADGE', 'AUDIT', 'WEB')",
"actor": "jeff",
"action": "random('ENTER','LOGIN','EXIT', 'LOGOUT')",
"objects": ["Building 2"],
"location": {
"lat": "double(-90.0, 90.0)",
"lon": "double(-180.0, 180.0)"
},
"message": "Entered Building 2"
}],
"duration": 0
}]
}
This workflow would output the defined json about every 4 seconds and then it will wait about 15 seconds before starting again.
Now that you know how to configure the Generator, it's time to run it. You will need Maven to build the application until we put a release up to download.
First, clone/fork the repo:
git clone [email protected]:acesinc/json-data-generator.git
cd json-data-generator
Now build it!
mvn clean package
Once this completes, a tar file will have been generated for you to use. Unpack the tar file somewhere you want to run the application from:
cp target/json-data-generator-1.0.0-bin.tar your/directory
cd your/directory
tar xvf json-data-generator-1.0.0-bin.tar
cd json-data-generator
In the conf directory, you will find an example Simulation Configuration and an example Workflow Definition. You can change these or make your own configurations to run with, but for now, we will just run the examples. This example simulates an auditing system that generates events when a user performs an action on a system. To do so, do the following:
java -jar json-data-generator-1.0.0.jar exampleSimConfig.json
You will start seeing output in your console and data will begin to be generated. It will look something like this:
...
2015-04-28 14:21:08,013 DEBUG n.a.d.j.g.t.TypeHandlerFactory [Thread-2] Discovered TypeHandler [ integer,net.acesinc.data.json.generator.types.IntegerType ]
2015-04-28 14:21:08,013 DEBUG n.a.d.j.g.t.TypeHandlerFactory [Thread-2] Discovered TypeHandler [ timestamp,net.acesinc.data.json.generator.types.TimestampType ]
2015-04-28 14:30:02,817 INFO data-logger [Thread-2] {"timestamp":1430253002793,"system":"BADGE","actor":"bob","action":"ENTER","objects":["Building 1"],"location":"45.5,44.3","message":"Entered Building 1"}
2015-04-28 14:30:05,369 INFO data-logger [Thread-2] {"timestamp":1430253005368,"system":"AD","actor":"bob","action":"LOGIN","objects":["workstation1"],"location":null,"message":"Logged in to workstation 1"}
2015-04-28 14:30:07,491 INFO data-logger [Thread-2] {"timestamp":1430253007481,"system":"AUDIT","actor":"bob","action":"COPY","objects":["/data/file1.txt","/share/mystuff/file2.txt"],"location":null,"message":"Printed /data/file1.txt"}
2015-04-28 14:30:09,768 INFO data-logger [Thread-2] {"timestamp":1430253009767,"system":"AUDIT","actor":"bob","action":"COPY","objects":["/data/file1.txt","/share/mystuff/file2.txt"],"location":null,"message":"Printed /data/file1.txt"}
This example only outputs to the Logger Producer, so all the data is going to your console and it is also being written to json-data-generator/logs/json-data.log. The data written to json-data.log is a single event per line and does not contain the logging timestamps and other info like above.
If you were to create your own simulation config called mySimConfig.json, you would place that file and any workflow conigs into the conf directory and run the application again like so:
java -jar json-data-generator-1.0.0.jar mySimConfig.json
As you saw in the Workflow Definition, our Generator supports a number of different functions that allow you to generate random data for the values in your json documents. Below is a list of all the currently supported Functions and the arguments they support/require.
If you specify a literal value as the value of a property, it will just be echoed back into the generated json. For example:
{
"literal-bool": true
}
Will always generate:
{
"literal-bool": true
}
| Function | Arguments | Description |
|---|---|---|
alpha(#) |
The number of characters to generate | Generates a random string of Alphabetic charaters with the length specified |
alphaNumeric(#) |
The number of characters to generate | Generates a random string of AlphaNumeric charaters with the length specified |
firstName() |
n/a | Generates a random first name from a predefined list of names |
lastName() |
n/a | Generates a random last name from a predefined list of names |
uuid() |
n/a | Generates a random UUID |
stringMerge() |
Delimiter and then the string values to merge | Takes the input arguments and merges them together using the delimiter. This can be used with the this.prop or cur.prop keywords to merge generated values like: stringMerge(_, this.firstName, this.lastName) which would output something like Bilbo_Baggins |
| Function | Arguments | Description |
|---|---|---|
boolean() |
n/a | Random true/false |
double([min, max]) |
Optional min val or range | If no args, generates a random double between Double.MIN & MAX. If one arg, generates a double between that min and Double.MAX. If two args, generates a double between those two numbers. |
integer([min, max]) |
Optional min val or range | If no args, generates a random integer between Integer.MIN & MAX. If one arg, generates a integer between that min and Integer.MAX. If two args, generates a integer between those two numbers. |
long([min, max]) |
Optional min val or range | If no args, generates a random long between Long.MIN & MAX. If one arg, generates a long between that min and Long.MAX. If two args, generates a long between those two numbers. |
| Function | Arguments | Description |
|---|---|---|
date([yyyy/MM/ddTHH:mm:ss,yyyy/MM/ddTHH:mm:ss]) |
Optional min date or date range | Generates a random date between the min and max specified. If no args, generates a random date before today. If one arg, generates a date after the specified min. If two args, generates a date between those dates. |
now([#_unit]) |
Optional number and unit to add to the now date | If no args, generates the date of now. If one arg, it will use the # portion as the amount of time and the unit portion (y=year,d=day,h=hour,m=minute) to convert the # to the time to add. If you want to subtract time, make your number negative (i.e. -5_d). Generates date in iso8601 formatted string |
timestamp([yyyy/MM/ddTHH:mm:ss,yyyy/MM/ddTHH:mm:ss]) |
Optional min date or date range | Generates a timestamp (i.e. long value) of the specified range. Same rules as the date() Function apply |
nowTimestamp() |
n/a | Generates a timestamp (i.e. long value) of the now date. |
| Function | Arguments | Description |
|---|---|---|
random(val1,val2,...) |
Literal values to choose from. Can be Strings, Integers, Longs, Doubles, or Booleans | Randomly chooses from the specified values |
counter(name) |
The name of the counter to generate | Generates a one up number for a specific name. Specify different names for different counters. |
this.propName |
propName = the name of another property | Allows you to reference other values that have already been generated (i.e. they must come before). For example, this.test.nested-test will reference the value of test.nested-test in the previously generated json object. You can also specify a this. clause when calling other functions like date(this.otherDate) will generate a date after another generated date. |
cur.propName |
propName = the name of another property at the same level as this property | Allows you to reference other values at the same level as the current property being generated. This is useful when you want to reference properties within a generated array and you don't know the index of the array. |
randomIncrementLong(name, baseValue, minStep, maxStep) |
The name of the value to generate, its base value and boundaries for the step | Generates a random step number for a specific name. Specify different names for different counters. |
We have two super special Functions for use with arrays. They are repeat() and random(). The repeat() function is used to specify that you want the Generator to take the elements in the array, and repeat its generator a certain number of times. You can specify the number of times or if you provide no arguments, it will repeat it 0-10 times. Use it like so:
{
"values": [
"repeat(7)",
{
"date": "date('2015/04/01T00:00:00', '2015/04/25T00:00:00')",
"count": "integer(1, 10)"
}]
}
This will generate a json object that has 7 values each with different random values for date and count.
The random() function will tell the generator to pick a random element from the array and out put that element only. Use it like so:
{
"values": [
"random()",
{
"date": "date('2015/04/01T00:00:00', '2015/04/25T00:00:00')",
"count": "integer(1, 10)"
},{
"thing1": "random('red', 'blue')"
}]
}
This will generate an array with one element that is either the element with date & count or the element with thing1 in it.
Here is a Kitchen Sink example to show you all the different ways you can generate data:
{
"strings": {
"firstName": "firstName()",
"lastName": "lastName()",
"random": "random('one', \"two\", 'three')",
"alpha": "alpha(5)",
"alphaNumeric": "alphaNumeric(10)",
"uuid": "uuid()"
},
"primatives": {
"active": "boolean()",
"rand-long": "long()",
"rand-long-min": "long(895043890865)",
"rand-long-range": "long(787658, 8948555)",
"rand-int": "integer()",
"rand-int-min": "integer(80000)",
"rand-int-range": "integer(10, 20)",
"rand-double": "double()",
"rand-double-min": "double(80000.44)",
"rand-double-range": "double(10.5, 20.3)"
},
"dates": {
"rand-date": "date()",
"min-date": "date(\"2015/03/01T00:00:00\")",
"range-date": "date(\"2015/03/01T00:00:00\",\"2015/03/30T00:00:00\")",
"now": "now()",
"nowTimestamp": "nowTimestamp()",
"5days-ago": "now(-5_d)",
"timestamp": "timestamp(\"2015/03/01T00:00:00\",\"2015/03/30T00:00:00\")"
},
"literals": {
"literal-bool": true,
"literal-int": 34,
"literal-long": 897789789574389,
"literal-double": 45.33
},
"references": {
"after-5days-ago": "date(this.dates.5days-ago)",
"isItActive": "this.primatives.active"
},
"counters": {
"counter1": "counter('one')",
"counter2": "counter('two')",
"counter2-inc": "counter('two')"
},
"randomIncrementLongs": {
"randomIncrementLong1": "randomIncrementLong('one', 1, 1, 10)"
"randomIncrementLong2": "randomIncrementLong('two', 0, 0, 5000)"
},
"array": [
{
"thing": "alpha(3)"
},
{
"thing2": "alpha(3)"
},
{
"thing3": true
},
{
"thing4": "job"
}
],
"repeat-array": [
"repeat(3)",
{
"thing1": "random('red','blue')",
"thing2": "random('red', 'blue')"
}
]
}
Which generates the following json:
{
"strings": {
"firstName": "Bob",
"lastName": "Black",
"random": "two",
"alpha": "ihDra",
"alphaNumeric": "F4mmpuSRDI",
"uuid": "1f04040f-fc6c-4de0-a17f-c588d1b62e75"
},
"primatives": {
"active": false,
"rand-long": 8886422858603719071,
"rand-long-min": 4635399365989710039,
"rand-long-range": 6934933,
"rand-int": 433941228,
"rand-int-min": 519082863,
"rand-int-range": 17,
"rand-double": 7.339790593356424E307,
"rand-double-min": 1.7261947151579352E308,
"rand-double-range": 11.0806
},
"dates": {
"rand-date": "1996-07-18T13:50Z",
"min-date": "2015-03-09T13:50Z",
"range-date": "2015-03-07T13:50Z",
"now": "2015-04-28T13:50Z",
"nowTimestamp": 1430250600027,
"5days-ago": "2015-04-23T13:50Z",
"timestamp": 1425588600027
},
"literals": {
"literal-bool": true,
"literal-int": 34,
"literal-long": 897789789574389,
"literal-double": 45.33
},
"references": {
"after-5days-ago": "2015-04-24T13:50Z",
"isItActive": false
},
"counters": {
"counter1": 0,
"counter2": 0,
"counter2-inc": 1
},
"randomIncrementLongs": {
"randomIncrementLong1": "1"
"randomIncrementLong2": "0"
},
"array": [{
"thing": "Psu"
}, {
"thing2": "yoU"
}, {
"thing3": true
}, {
"thing4": "job"
}],
"repeat-array": [{
"thing1": "red",
"thing2": "blue"
}, {
"thing1": "blue",
"thing2": "blue"
}, {
"thing1": "blue",
"thing2": "red"
}]
}