Embedded python?

example.erb

<%#
  #
  # Embedded python?
  #
  # See:
  #   http://mail.python.org/pipermail/python-ideas/2011-December/013003.html
  #   http://readthedocs.org/docs/ncoghlan_devs-python-notes/en/latest/pep_ideas/suite_expr.html
  # and related discussion:
  #   http://mail.python.org/pipermail/python-ideas/2011-December/thread.html#13003
  #
  # (Updated to add the new :} closing brace syntax and add a couple of missing
  # close braces before the elses).
  #
  # Some background:
  #
  # I recently was doing some Django coding for the first time. I discovered
  # that I needed to learn a templating language in order to write templates in
  # Django.
  #
  # So I'm a Ruby guy. In ruby we have ERB. You don't have to 'learn' ERB since
  # ERB is ruby. You never have to wonder how to do something in a template, or
  # if a particular thing you want to do is possible.  So the answer to "can it
  # be done?" is always yes, and the question "how is it done?" never comes up,
  # since, again, it's ruby.
  #
  # As a ruby programmer looking at python, the idea of needing to learn a
  # templating language is a strange concept. And the fact that there
  # are many different and competing python templating languages feels like a
  # sign that something isn't right.
  #
  # As has been stated already, python's significant whitespace, for all its
  # benefits, is largely what makes embedded python impractical. (See the first
  # anti-example here: http://wiki.python.org/moin/Templating ). The proposal
  # might change this, and I think that would be a good thing.
  #
  # (The following is to get a feel for what embedded python might look like).
  #
%>

<% if danger_level > 3 {: %>
  <div class="alert"><% if danger_level == 5 {: %>EXTREME <% :} %>DANGER ALERT!</div>

<% :} elif danger_level > 0 {: %>
  <div>Some chance of danger</div>

<% :} else {: %>
  <div>No danger</div>

<% :} %>

<% for a in ['cat', 'dog', 'rabbit'] {: %>
  <h2><%= a %></h2>
  <p><%= describe_animal(a) %></p>

<% :} %>

Linkage:

• http://mail.python.org/pipermail/python-ideas/2011-December/013003.html
• http://readthedocs.org/docs/ncoghlan_devs-python-notes/en/latest/pep_ideas/suite_expr.html
• http://mail.python.org/pipermail/python-ideas/2011-December/thread.html#13003
• http://wiki.python.org/moin/Templating

It's been demonstrated time and time again that having language constructs in in templates encourages bad separation. So I have to ask: Why you want to have embedded python?

As far as templating in python goes, the one I like is Jinga2, which is a bit like the stuff in django, but more.

You are correct. Being able to write arbitrary code in a template doesn't mean you should. And in practice it's not hard to learn a templating syntax.

But imagine a developer doing a 10 minute 'Hello World' tutorial with a python based framework like Django. When it gets to the part about templates she effectively must learn a templating language with a new unfamiliar syntax just to display hello world in a browser. I don't think it should be that way. Why not have embedded python?

And then, after the initial playful phase, do you take away the Python-based templating and move to a more (positively) constrained one?

At the point when a developer knows why they want to use a non-standard[1] templating engine they are qualified to choose to do so.

[1] As devil's advocate, I will refer to any python templating language that isn't embedded python as "non-standard". ;)

Fair enough - I'm all for choice (and rope). Plus it's an interesting intellectual exercise.