Embed custom filter definition into jinja2 template?

Question

I'm writing some Jinja2 templates that I'd like to be able to reuse as painlessly as possible in tangentially related projects. I have a set of custom convenience filters that I'd like the templates to "carry around" with them. Is there a Jinja2 syntax for embedding filter definitions into a template itself? Or a mechanism for embedding any kind of pure Python function into a Jinja2 template that can act on variables passed into the template? I used to use mako, and there it was trivial to do this, but templating LaTeX in mako is painful because of the lack of custom syntax, so I had to make the switch.

Solution 1

There is NO way you can embed python directly into a Jinja2 Template, the way that I know of is to define in your application and add them to your Jinja2 environment instance. Like the following example taken from https://jinja.palletsprojects.com/en/2.11.x/api/#writing-filters.

import jinja2

loader = jinja2.FileSystemLoader('/tmp')
env = jinja2.Environment(autoescape=True, loader=loader)

def upperstring(input):
    """Custom filter"""
    return input.upper()

env.filters['upperstring'] = upperstring
temp = env.get_template('template.html')
temp.render(name="testing")

Here the Template I am using

{{ name | upperstring }}

Result is this

TESTING

Solution 2

There is an easy way to add custom filters in jinja2 template. FILTERS is the dictionary containing all filters that we can use to render the template. However, we can add more filters to it.

Here is a quick example to add new filters to it.

from jinja2 import Template
from jinja2.filters import FILTERS, environmentfilter


@environmentfilter
def do_reverse_by_word(environment, value, attribute=None):
    """
    custom max calculation logic
    """
    if attribute:
        return [list(reversed(i.get(attribute).split())) for i in value]

    return list(reversed(value.split()))


FILTERS["reverse_by_word"] = do_reverse_by_word
print(Template("{{ name | reverse_by_word }}").render({"name": "Stack Overflow"}))
print(Template("{{ names | reverse_by_word(attribute='name') }}").render({"names": [{"name": "Stack Overflow"}, {"name": "Stack Exchange"}]}))

Outputs

['Overflow', 'Stack']
[['Overflow', 'Stack'], ['Exchange', 'Stack']]

please comment below in case you've more complex use cases. I'll be happy to answer all your queries.

Comments

[Tanmay Baranwal]

Any idea on how to remove all the old filters?

[flazzarini]

If you mean removing all existing default filters you could simple replace env.filters with an empty dictionary by doing env.filters = {}. Not sure however why would want to do that.

[Tanmay Baranwal]

Sorry, I meant all the old "custom" filters. But I found that Jinja environment sets up the basic filters explicitly, so can't assign to empty dictionary.

[superloopnetworks]

@flazzarini How do you call upperstring and feed in the variable "input"? If I wanted a custom filter for boolean, how do I return the variable true/false that I pass in?

[flazzarini]

Not sure if I understand your question correctly, but If you would like to define a filter that would return a boolean type you could write something here to transform any given strings into boolean values using a jinja2 filter.