This IPython extension makes it easier to interact with MarkLogic directly from your Jupyter notebooks by providing several magic functions.

NOTE: the extension is still a prototype, and is not currently published (via pip), but will be soon. Feedback is definitely welcome!

An example notebook is available to quickly check out the extension in action, which can be viewed here.

To run the example yourself, you'll need the following installed:

• Docker Desktop
• Gradle

Steps to run the example:

• In the project directory, run docker-compose up -d. This will setup a MarkLogic server and Jupyter server.
• Go to http://localhost:8001 to initialize the MarkLogic server. Follow the setup steps. When asked, set the admin password to admin.
• Once MarkLogic is setup, run ./gradlew mlDeploy.
• Run ./gradlew loadSampleData.
• Go to http://localhost:9999 to access Jupyter, use admin as the token/password.
• Click on the work folder and open the example.ipynb notebook.
• Open the Cell menu and click Run All.

You can use the extension in your own notebooks by copying the marklogic_ext directory in the same directory where your notebooks are located. Remember to restart the kernel if your notebook(s) are already open. This is a temporary process until the extension is published.

The extension uses magic functions to extend or change the behavior of notebook cells, allowing you to interact with a MarkLogic server in different ways. Magic functions are prefixed by % or %%.

Add these to a notebook cell to load the extension and connect to a MarkLogic server:

%load_ext marklogic_ext
%marklogic_connect http://username:password@hostname:port

The extension uses digest authentication by default. If you need to use basic authentication, add the -basic option:

%marklogic_connect http://username:password@hostname:port -basic

If the server uses SSL, specify https:

%marklogic_connect https://username:password@hostname:port

Run XQuery code directly in a cell:

%%marklogic_xqy

fn:collection("sample-data")

Run server-side Javascript directly in a cell:

%%marklogic_sjs

fn.collection("sample-data");

By default, the extension will try to coerce the script's output into corresponding Python types, such as numbers. If the script returns XML, JSON objects, or JSON arrays, the extension will try to coerce them into Python dicts.

You can use the -dataframe option to try coercing the output into a pandas.DataFrame object (or raise an error if it can't). It is best to construct the DataFrame using an array of dicts, so it is recommended the script return an array or sequence of JSON or XML nodes.

%%marklogic_sjs -dataframe

return [{ id: 1, value: 3 }, { id: 2, value: 2 }, { id: 3, value: 1}];

There is also a -raw option that returns the output as an array of strings, regardless of their type.

You can run SQL queries against your SQL views or TDEs:

%%marklogic_sql

SELECT * FROM schema.view LIMIT 50

SQL queries return pandas.DataFrame objects by default. The -raw option is also available.

You can insert documents directly from a cell:

insert_model = {
    "uri": "/models/example.onnx",
    "format": "binary",
    "content": onnx.SerializeToString()
}
%marklogic_doc_insert insert_model

In a Python cell, construct a dict variable with the document's uri and content. The content can be a string or bytes. By default, MarkLogic will try to decipher the document's mime type. Optionally, you can explicitly specify the document type using format (can be xml, json, or binary). Provide the variable as an argument to %marklogic_doc_insert.