I've been seeing this issue come up a lot lately in various different places, both in relation to Jinja and Django Templates. There seems to be a fundamental misunderstanding (among some users) about how templates systems work and how that relates to Markdown text which is rendered to HTML and inserted into a template. I’ll try to explain this clearly. Note that while the answer below applies to most templating systems (including Jinja and Django), the examples use Jinja for illustrative purposes (after all, the original question specifically asks about Jinja). Simply adapt the code to match the API of your templating system of choice, and it should work just as well.
First of all, Markdown has no knowledge of template syntax. In fact, Markdown has been around longer than Jinja, Django or various other popular templating systems. Additionally, the Markdown Syntax Rules make no mention of template syntax. Therefore, your template syntax will not be processed simply by passing some Markdown text which contains template syntax through a Markdown parser. The template syntax needs to be processed separately by the template engine. For example:
from jinja2 import Environment
# Set up a new template environment
env = Environment()
# Create template with the markdown source text
template = env.from_string(text_from_markdown_editor)
# Render that template. Be sure to pass in the context (post in this instance).
template_processed_markdown = template.render(post=post)
# Now pass the Markdown text through the Markdown engine:
post_body = markdown(template_processed_markdown)
Note that the above first processes the template syntax, then parses the Markdown. In other words, the output of the template processing is still Markdown text with the tags replaced by the appropriate values. Only in the last line is the Markdown text converted to HTML by the Markdown parser. If you want the order of processing to be reversed, you will need to switch the code around to run the Markdown parser first and then pass the output of that through the template processor.
I assume that some of the confusion comes from people passing the Markdown text through a templating system. Shouldn’t that cause the template syntax to get processed? In short, No.
At its core, a templating system takes a template and a context. It then finds the various tags in the template and replaces those tags with the matching data provided in the context. However, the template has no knowledge about the data in the context and does no processing of that data. For example, this template:
Hello, {{ name }}!
And this context:
output = template(name='John')
Would result in the following output:
Hello, John!
However, if the context was this instead:
output = template(name='{(some_template_syntax)}')
then the output would be:
Hello, {{some_template_syntax}}!
Note that while the data in the context contained template syntax, the template did not process that data. It simply considered it a value and inserted it as-is into the template in the appropriate location. This is normal and correct behavior.
Sometimes however, you may have a legitimate need for a template to do some additional processing on some data passed to the template. For that reason, the template system offers filters. When given a variable in the context, the filter will process the data contained in that variable and then insert that processed data in the template. For example, to ensure that the name in our previous example is capitalized, the template would look like the following:
Hello, {{ name|capatalize }}!
Passing in the context output = template(name='john') (note that the name is lowercase), we then get the following output”
Hello, John!
Note, that the data in the name variable was processed by having the first letter capitalized, which is the function of Jinja’s built-in filter capitalize. However, that filter does not process template syntax, and therefore passing template syntax to that filter will not cause the template syntax to be processed.
The same concept applies to any markdown filter. Such a filter only parses the provided data as Markdown text and returns HTML text which is then placed into the template. No processing of template syntax would happen in such a scenario. In fact, doing so could result in a possible security issue, especially if the Markdown text is being provided by untrusted users. Therefore, any Markdown text which contains template syntax must have the template syntax processed separately.
However, there is a note of caution. If you are writing documentation which includes examples of Template syntax in them as code blocks (like the Markdown source for this answer), the templating system is not going to know the difference and will process those tags just like any template syntax not in a code block. If the Markdown processing was done first, so that the resulting HTML was passed to the templating system, that HTML would still contain unaltered template syntax within the code blocks which would still be processed by the templating system. This is most likely not what is desired in either case. As a workaround, one could conceivably create some sort of Markdown Extension which would add syntax processing to the Markdown processor itself. However, the mechanism for doing so would differ depending on which Markdown processor one is using and is beyond the scope of this question/answer.
