Saturday, February 25, 2017

Bleach and html5lib

Good news for #1262 : Will Kahn-Greene (author of bleach) has fixed issue229. He asked me to test whether it works now. Here we go:

$ pip uninstall -y weasyprint html5lib bleach
$ pip install
$ pip install weasyprint html5lib

Yes, the import itself now works:

$ python -c "import bleach"

That is, the incompatability is fixed. We can close #1262.

I modified lino.mixins.bleached (which was deactivated until now) so that now the module should be active. I also extended the docstring. I activated it for lino.modlib.comments.models.Comment. That is, every comment is going to be bleached.

Now we should test whether this is actually what we want. New ticket #1522.

For example it is possible that I forgot to specify in allowed_tags some tags that we do not want to remove.

I think the best test is to use it on the battle field.

One challenge here is that the usage is currently “destructive”: when we deploy it to our production site, all comments will automatically get bleached with our data migration. And we won’t even get a notification of which rows are concerned. We need at least a kind of “bleach logger”.

A first idea is to override after_ui_save() instead of full_clean(). This makes everything less destructive. (I forgot to check this in, so it is visible only in 20170226)

TODO: Lino should log at least a bit of bleach’s “activity”. I mean for example that Bleached.before_ui_save should log an info message saying “removed tags x, y, z from short_text”

Almost all test suites seem to be passing now when Lino branch Comment.reply_to_common_work is activated. An exeception is still Lino Presto which says Exception: TicketDetail on tickets.Tickets has no data element 'changes.ChangesByMaster'. It is technically trivial, but conceptually complex: we need to decide whether Lino Presto wants lino.modlib.changes or not…