Build: don't track changed files

[stsewd]

We aren't caching per file,
and to do this we are tracking the md5 for each file.
Which involves reading te file, calculate the md5, and query the db for
each html file.

We also don't need the HTMLFile model at all..
But, we can remove that in another PR (#7875),
this should help to make indexing fast.

Less and faster code.

[ericholscher]

This makes sense to me. I feel like we were using logic for some other reason, but I can't remember what it was. We should verify in the DB nobody has cdn_enabled set.

The embed UI is the only other place I remember using ImportedFile's, but we only want HTMLFile's there too.

[stsewd]

We should verify in the DB nobody has cdn_enabled set.

We aren't using that attribute in the code base (-ext and .com). But here are results from the db

.org

In [3]: Project.objects.filter(cdn_enabled=True).count()
Out[3]: 2

In [5]: Project.objects.filter(cdn_enabled=True).values_list('slug')
Out[5]: <ProjectQuerySetBase [('time-test',), ('write-the-docs',)]>

.com

In [1]: Project.objects.filter(cdn_enabled=True).count()
Out[1]: 0

The embed UI is the only other place I remember using ImportedFile's, but we only want HTMLFile's there too.

Yeah, we don't use imported files that aren't html.

readthedocs.org/readthedocs/projects/views/public.py

Lines 514 to 516 in 4e398fa

	files = version.imported_files.filter(
	name__endswith='.html',
	).order_by('path')

[ericholscher]

This is a useful improvement I think. We should next remove the md5 from the DB 👍