Html minification middleware with (but not only) django support and good

django, html, html5, htmlmin, minification, python, python38
pip install django-hmin==0.5.4



Build and publish PyPI version codecov Code style: black

Django (2.0+) oriented HTML minification function and middleware (another one).
Key feature — speed. 10x (on large html's it can be 50x, 100x, 200x or even more) time faster, than htmlmin.

Compress html code and removes html comments, but ignores conditional comments (IE) by default.
Uses cache by default (can be disabled), so minification overhead is greatly reduced.
Also it can be used as solo function.
For best expirience use it with

Written in modern python 3.7+ with fully typing-covered codebase.

Full support of:


  • Python 3.7+
  • Django 2.0+ (not required)


Regular way

For install django-hmin, run on terminal:

$ pip install django-hmin


If there is no release, or you dont want/cant use pypi, then:

  • git clone this repo
  • pipenv install to install requirements or just pip install flit
  • flit install Yes, we dont use classic, we use flit + pyproject.toml. God bless PEP-518 + PEP-621.

Using with Django as midleware

All you need to do is add two middlewares to your MIDDLEWARE_CLASSES:

    # other middleware classes

If you're using Django's caching middleware, MarkMiddleware should go after FetchFromCacheMiddleware, and MinMiddleware should go after UpdateCacheMiddleware:

    # other middleware classes

You can optionally specify the HTML_MINIFY setting:

HTML_MINIFY: bool = True

The default value for the HTML_MINIFY setting is not DEBUG. You only need to set it to True if you want to minify your HTML code when DEBUG is enabled.

URL exclude

Specify setting:

HMIN_EXCLUDE: tuple = ('^base/', '^admin/')

Keep HTML comments

Specify settings:



By default hmin middleware uses cache via django caches framework (very useful for small and middle web sites, and for big you definitely will use ngx_pagespeed or other "big" solutions). You can disable it by specify setting:

HMIN_USE_CACHE: bool = False

Also you can change time and cache backend (if you want, by default time is 3600, cache backend — "default"):

HMIN_CACHE_BACKEND: str = 'my_cache'

More using scenarios


Just import decorator minify_plain: from hmin.decorators import minify_plain, than you can minify any function you want:

def my_cool_func():
    return some_plain_html

Or, if you want to keep html comments:

def my_cool_func():
    return some_plain_html


Just import function minify. Function definition: def minify(content, remove_comments=True). Example:

from hmin import html_minify

html_minify('<div>     hello</div>') # <div>hello</div>


$ python -m hmin filename.html > filename.min.html

Benchmarking (wannabe)

New data

I try to compress 1.1mb of very dense html on my i7 laptop processor (2020 edition!) and measure raw time with chrome inspector.
I got following data (this data comes from field "Waiting (TTFB)", not "Content Download"):

  • htmlmin took about 2.5 seconds!
  • this package took:
    • ~100ms (verion 0.5.2)
    • ~60ms (version 0.5.3) — i fixed some bug in regexp, now package doesnt strip spaces BEFORE open tag, cause it was incorrect behaviour. So, now speed is like previous version. Yes! :)

As you can see, hmin 0.5+ are slightly slower than 0.3, but this is comes from updated minifcation logic. In previous version minification was not so accurate, as i think. But, hmin still very much faster.
Maybe in the next releases i made some experiments with cython or numba to improve speed.

Old data (version < 0.5)

Stupid speed benchmark (1)

I try to compress 1mb of html (i think, your usual html is slightly thiner) on my i7 laptop processor and measure time with django-debug-toolbar.
Django overhead took about 40ms (all), this is time without minification, just plain html, django, etc.
Then i install hmin and htmlmin and just look at the debug toolbar numbers (this is very silly and simple "benchmark"):

  • with hmin cpu was about 60ms (min)
  • with htmlmin cpu was about 1200ms (min) Minus overhead, plain time: hmin — 20ms, htmlmin — 1160ms.
    Probably, you can get other numbers. But hmin really faster than htmlmin.

Stupid speed benchmark (2)

Also i try to compress 2mb of html on my desktop i3 (sandy bridge).
Debug toolbar time:

  • with hmin cpu was about 220ms without cache, and 87ms with cache
  • with htmlmin cpu was about 125000ms Django overhead was about 80ms.
    Minus overhead, plain time: hmin — 140ms (7ms with cache), htmlmin — ok.


You can check release page.

Current possible problems

  • Doesnt respect CDATA