How to display markdown output in databricks notebook from a python cell

Question

With IPython/Jupyter it's possible to output markdown using the IPython display module and its MarkDownclass.

Question

How can I accomplish this with Azure Databricks?

What I tried

Databricks display

Tried using Databrick's display with the IPython Markdown class:

from IPython.display import Markdown
display(Markdown('*some markdown* test'))

but this results in the following error:

Exception: Cannot call display(<class 'IPython.core.display.Markdown'>)

IPython display

I then tried to use IPython's display:

from IPython.display import display, Markdown
display(Markdown('*some markdown* test'))

but this just displays the text:

<IPython.core.display.Markdown object>

IPython display_markdown

Tried using IPython's display_markdown:

from IPython.display import display_markdown
display_markdown('# Markdown is here!\n*some markdown*\n- and\n- some\n- more')

but this results in nothing showing up:

Looking up documentation

Also tried checking Azure Databricks documentation. At first I visited https://www.databricks.com/databricks-documentation which leads me to https://learn.microsoft.com/en-ca/azure/databricks/ but I wasn't able to find anything via searching or clicking the links and I usually find Microsoft documentation quite good.

Checking Databrick's display source

As Saideep Arikontham mentioned in the comments, Databricks version 11 and above is using IPython kernel so I dug a bit deeper.

According to Databrick's source for the display function, it will readily render any object that implements _repr_html().

However I'm having a hard time being able to get the raw html output that I'm assuming IPython.display.Markdown should be able to output. I can only find _repr_markdown_() and _data_and_metadata() where the former just calls the latter and the output, at least in Databricks, is just the original raw markdown string.

Answer 1

As of 2024, using a Databricks Runtime 15.4 LTS, it is possible to use IPython.display suggested in the original question.

from IPython.display import Markdown, display

# Generate welcome markdown text
welcome_text = '''
# Welcome to my notebook!

## Examples of titles and subtitles

### Title 1
This is an example of a title.

## Example of a table

| Column 1 | Column 2 | Column 3 |
|----------|----------|----------|
|   A      |    B     |    C     |
|   D      |    E     |    F     |
|   G      |    H     |    I     |
'''

# Display the welcome markdown text
display(Markdown(welcome_text))

Will successfully display the expected markdown

Answer 2

Markdown and display_markdown are not giving desired output when used in Azure Databricks. I have done the following in Databricks 11.1 runtime.

• Taking inputs from the question, I understood that when a class has _repr_html(), it is able to output the desired result. But when this method is absent in class, it is returning an object.
• So, for Markdown to work, I have written my own Markdown class where I used Python's markdown library.

from IPython.display import DisplayObject, TextDisplayObject

class Markdown(TextDisplayObject):

    def __init__(self,TextDisplayObject):
        import markdown as md
        
        #converting markdown to html
        self.html = md.markdown(TextDisplayObject)
        
    
    def _repr_html_(self):
        return self.html

• Now, this class is not completely same as the IPython.display.Markdown. I have formatted your sample markdown '# Markdown is here!\n*some markdown*\n- and\n- some\n- more' as following to get the desired result.

Markdown('''# Markdown is here!\n
*some markdown*\n
- and\n
- some\n
- more''')

NOTE:

• For display_markdown() to display output, we must specify another argument raw as True (display_markdown(<markdown_str>, raw=True)). However, in Databricks it is returning undefined (NoneType).

• Please do install markdown library first using %pip install markdown in Databricks cell.