Mikhail Korobov

Hi David, this issue is resolved in my fork: http://bitbucket.org/kmike/django-qsstats-magic/src

David, hi again. This issue is also resolved here: http://bitbucket.org/kmike/django-qsstats-magic/src

I think adding pyqt5 to `install_requires` could make things worse; it is not always possible to install pyqt5 using pip, and with pyqt5 in `install_requires` an innocent-looking `pip install -U...

I believe regression happened when lxml document cache was removed, as a part of https://github.com/scrapy/scrapy/pull/1409. The idea was to use response.selector explicitly; we fixed it for link extractors (https://github.com/scrapy/scrapy/pull/1409/files#diff-952f1c77a1cfacb08eed58f08daa8870R67), but...

@juraseg I think your approach is fine, but there is an extra tricky part: when you access `response.selector.__class__`, response body could be parsed in order to fill the "selector" attribute...

TakeFirstN sounds fine to me, though I haven't used item loaders for a long time.

> The tests for py3.5 are failing because of f-strings. Are we going to deprecate py3.5 soon? Currently Scrapy supports Python 3.5.1+ because it is Python availlable in Ubuntu 16.04...

+1 to this proposal.

I agree with @hugovk; I'm also not using reruns on success, and not going to use them.

I like the idea. Something like `{% render_field form.text readonly?=value %}`? Pull requests are welcome.