Python Liquid

A Python implementation of Liquid, the safe, customer-facing template language for flexible web apps.

from liquid import Template

template = Template("Hello, {{ you }}!")
print(template.render(you="World"))  # "Hello, World!"
print(template.render(you="Liquid"))  # "Hello, Liquid!"

Installing

Install Python Liquid using Pipenv:

$ pipenv install python-liquid

Or pip:

$ python -m pip install -U python-liquid

Or from conda-forge:

$ conda install -c conda-forge python-liquid

Related Projects

liquid-babel: Internationalization and localization for Liquid templates.
LiquidScript: A JavaScript and TypeScript engine for Liquid with a similar high-level API to Python Liquid.
django-liquid: A Django template backend for Liquid. Render Liquid templates in your Django apps.
Flask-Liquid: A Flask extension for Liquid. Render Liquid templates in your Flask applications.
golden-liquid: A test suite for Liquid. See how various Liquid template engines compare to the reference implementation.

Compatibility

We strive to be 100% compatible with the reference implementation of Liquid, written in Ruby. That is, given an equivalent render context, a template rendered with Python Liquid should produce the same output as when rendered with Ruby Liquid.

See the known issues page for details of known incompatibilities between Python Liquid and Ruby Liquid, and please help by raising an issue if you notice an incompatibility.

Benchmark

You can run the benchmark using make benchmark (or python -O performance.py if you don't have make) from the root of the source tree. On my ropey desktop computer with a Ryzen 5 1500X and Python 3.11.0, we get the following results.

Best of 5 rounds with 100 iterations per round and 60 ops per iteration (6000 ops per round).

lex template (not expressions): 1.2s (5020.85 ops/s, 83.68 i/s)
                 lex and parse: 5.0s (1197.32 ops/s, 19.96 i/s)
                        render: 1.4s (4152.92 ops/s, 69.22 i/s)
         lex, parse and render: 6.5s (922.08 ops/s, 15.37 i/s)

And PyPy3.7 gives us a decent increase in performance.

Best of 5 rounds with 100 iterations per round and 60 ops per iteration (6000 ops per round).

lex template (not expressions): 0.58s (10308.67 ops/s, 171.81 i/s)
                 lex and parse: 3.6s (1661.20 ops/s, 27.69 i/s)
                        render: 0.95s (6341.14 ops/s, 105.69 i/s)
         lex, parse and render: 4.6s (1298.18 ops/s, 21.64 i/s)

On the same machine, running rake benchmark:run from the root of the reference implementation source tree gives us these results.

/usr/bin/ruby ./performance/benchmark.rb lax

Running benchmark for 10 seconds (with 5 seconds warmup).

Warming up --------------------------------------
             parse:     3.000  i/100ms
            render:     8.000  i/100ms
    parse & render:     2.000  i/100ms
Calculating -------------------------------------
             parse:     39.072  (± 0.0%) i/s -    393.000  in  10.058789s
            render:     86.995  (± 1.1%) i/s -    872.000  in  10.024951s
    parse & render:     26.139  (± 0.0%) i/s -    262.000  in  10.023365s

I've tried to match the benchmark workload to that of the reference implementation, so that we might compare results directly. The workload is meant to be representative of Shopify's use case, although I wouldn't be surprised if their usage has changed subtly since the benchmark fixture was designed.

Contributing

Please see Contributing to Python Liquid.