If conditions can cause massive confusion (here in django coverage plugin and even in pytest-cov/coverage)

At first I would like to thank you for this wonderful plugin - we love it! We noticed the CRLF vs. LF problem a while ago but that just occured accidentally when we where testing through Gitlab Web IDE - so that's not a real blocker.

We recently looked closer into if conditions and think there is room for improvement, as sometimes 100% coverage is reported, but it just isn't true.

Let's create a very simple example: Imagine you could pass in a get param ?red=1, which would e.g. highlight the title:

<h1 
    {% if request.GET.red %}
        class="red"
    {% endif %}
>Headline</h1>

When you don't have a test covering passing in red=1 at least line 3 will be correctly reported as not covered. So far so good.

Sadly this unpopular but possible case

<h1 
    {% if request.GET.red %}
        class="red"    {% endif %}
>Headline</h1>

or this very popular case

<h1 {% if request.GET.red %}class="red"{% endif %}>Headline</h1>

will report 100% coverage - which is not correct at all.

This caught my attention, so I thought I should review it in Python in the view logic as well. Let's pretend the code above would rely on a variable:

<h1 {% if red %}class="red"{% endif %}>Headline</h1>

In our example we'd like to achieve the same behaviour so our view would look like this

def index(request):
    red = False
    if request.GET.get("red", False):
        red = True
    return render(request, "base/index.html", {"red": red})

Again this works as expected through pytest-cov, so in case you don't hand over ?red=1 in one of your the tests it will put line 4 on your list.

But sadly even in this extremely popular pattern it is not working as expected and stating 100% coverage, even when the condition is not covered.

def index(request):
    red = False if request.GET.get("red", False) else True
    return render(request, "base/index.html", {"red": red})

I checked it again with a colleague and still: maybe we've made a mistake somewhere, but these examples are very straight forward... so we doubt it. Could somebody spent a couple of minutes and confirm this?

We used:

• Python 3.12.3 (main, Feb 4 2025, 14:48:35) [GCC 13.3.0] on linux
• django-coverage-plugin==3.1.1
• coverage==7.10.1

In combination with:

• pytest==8.4.1
• pytest-cov==6.2.1
• pytest-django==4.11.1
• pytest-playwright==0.7.0

Reproduce the inline Django template condition and the Python conditional-expression examples using the versions listed in the issue. Compare the coverage results when each branch is exercised and when it is not; the investigation is complete when the incorrect 100% reports are explained and a focused change or test target is identified.