{% include JB/setup %}
Apache Zeppelin Interpreter is a language backend. For example to use scala code in Zeppelin, you need a scala interpreter. Every Interpreters belongs to an InterpreterGroup. Interpreters in the same InterpreterGroup can reference each other. For example, SparkSqlInterpreter can reference SparkInterpreter to get SparkContext from it while they're in the same group.
InterpreterSetting is configuration of a given InterpreterGroup and a unit of start/stop interpreter. All Interpreters in the same InterpreterSetting are launched in a single, separate JVM process. The Interpreter communicates with Zeppelin engine via Thrift.
In ‘Separate Interpreter(scoped / isolated) for each note’ mode which you can see at the Interpreter Setting menu when you create a new interpreter, new interpreter instance will be created per note. But it still runs on the same JVM while they're in the same InterpreterSettings.
Creating a new interpreter is quite simple. Just extend org.apache.zeppelin.interpreter abstract class and implement some methods. You can include org.apache.zeppelin:zeppelin-interpreter:[VERSION]
artifact in your build system. And you should put your jars under your interpreter directory with a specific directory name. Zeppelin server reads interpreter directories recursively and initializes interpreters including your own interpreter.
There are three locations where you can store your interpreter group, name and other information. Zeppelin server tries to find the location below. Next, Zeppelin tries to find interpreter-setting.json
in your interpreter jar.
{ZEPPELIN_INTERPRETER_DIR}/{YOUR_OWN_INTERPRETER_DIR}/interpreter-setting.json
Here is an example of interpreter-setting.json
on your own interpreter.
[ { "group": "your-group", "name": "your-name", "className": "your.own.interpreter.class", "properties": { "properties1": { "envName": null, "propertyName": "property.1.name", "defaultValue": "propertyDefaultValue", "description": "Property description" }, "properties2": { "envName": PROPERTIES_2, "propertyName": null, "defaultValue": "property2DefaultValue", "description": "Property 2 description" }, ... }, "editor": { "language": "your-syntax-highlight-language", "editOnDblClick": false } }, { ... } ]
Finally, Zeppelin uses static initialization with the following:
static { Interpreter.register("MyInterpreterName", MyClassName.class.getName()); }
Static initialization is deprecated and will be supported until 0.6.0.
The name will appear later in the interpreter name option box during the interpreter configuration process. The name of the interpreter is what you later write to identify a paragraph which should be interpreted using this interpreter.
%MyInterpreterName some interpreter specific code...
You can add editor
object to interpreter-setting.json
file to specify paragraph editor settings.
If the interpreter uses a specific programming language (like Scala, Python, SQL), it is generally recommended to add a syntax highlighting supported for that to the note paragraph editor.
To check out the list of languages supported, see the mode-*.js
files under zeppelin-web/bower_components/ace-builds/src-noconflict
or from github.com/ajaxorg/ace-builds.
If you want to add a new set of syntax highlighting,
mode-*.js
file to zeppelin-web/bower.json (when built, zeppelin-web/src/index.html will be changed automatically).language
field to editor
object. Note that if you don‘t specify language field, your interpreter will use plain text mode for syntax highlighting. Let’s say you want to set your language to java
, then add:"editor": { "language": "java" }
If your interpreter uses mark-up language such as markdown or HTML, set editOnDblClick
to true
so that text editor opens on pargraph double click and closes on paragraph run. Otherwise set it to false
.
"editor": { "editOnDblClick": false }
Once you have built your interpreter, you can place it under the interpreter directory with all its dependencies.
[ZEPPELIN_HOME]/interpreter/[INTERPRETER_NAME]/
To configure your interpreter you need to follow these steps:
conf/zeppelin-site.xml
.Property value is comma separated [INTERPRETER_CLASS_NAME]. For example,
<property> <name>zeppelin.interpreters</name> <value>org.apache.zeppelin.spark.SparkInterpreter,org.apache.zeppelin.spark.PySparkInterpreter,org.apache.zeppelin.spark.SparkSqlInterpreter,org.apache.zeppelin.spark.DepInterpreter,org.apache.zeppelin.markdown.Markdown,org.apache.zeppelin.shell.ShellInterpreter,org.apache.zeppelin.hive.HiveInterpreter,com.me.MyNewInterpreter</value> </property>
Add your interpreter to the default configuration which is used when there is no zeppelin-site.xml
.
Start Zeppelin by running ./bin/zeppelin-daemon.sh start
.
In the interpreter page, click the +Create
button and configure your interpreter properties. Now you are done and ready to use your interpreter.
Note : Interpreters released with zeppelin have a default configuration which is used when there is no
conf/zeppelin-site.xml
.
Inside of a note, %[INTERPRETER_NAME]
directive will call your interpreter. Note that the first interpreter configuration in zeppelin.interpreters will be the default one.
For example,
%myintp val a = "My interpreter" println(a)
Inside of a note, %[INTERPRETER_GROUP].[INTERPRETER_NAME]
directive will call your interpreter.
You can omit either [INTERPRETER_GROUP] or [INTERPRETER_NAME]. If you omit [INTERPRETER_NAME], then first available interpreter will be selected in the [INTERPRETER_GROUP]. Likewise, if you skip [INTERPRETER_GROUP], then [INTERPRETER_NAME] will be chosen from default interpreter group.
For example, if you have two interpreter myintp1 and myintp2 in group mygrp, you can call myintp1 like
%mygrp.myintp1 codes for myintp1
and you can call myintp2 like
%mygrp.myintp2 codes for myintp2
If you omit your interpreter name, it'll select first available interpreter in the group ( myintp1 ).
%mygrp codes for myintp1
You can only omit your interpreter group when your interpreter group is selected as a default group.
%myintp2 codes for myintp2
Checkout some interpreters released with Zeppelin by default.
We welcome contribution to a new interpreter. Please follow these few steps:
pom.xml
.docs/interpreter/
. Follow the Markdown style as this example. Make sure you list config settings and provide working examples on using your interpreter in code boxes in Markdown. Link to images as appropriate (images should go to docs/assets/themes/zeppelin/img/docs-img/
). And add a link to your documentation in the navigation menu (docs/_includes/themes/zeppelin/_navigation.html
).