Trying to use an iterator as a section fails after the first time

[bdesham]

Consider the following simple use of Pystache under Python 3:

import pystache

context = {
    'numbers': [
        {'number': 1},
        {'number': 2},
        {'number': 3}
    ]
}

template = '''
    {{ #numbers }}
    * {{ number }}
    {{ /numbers }}

    {{ #numbers }}
    * {{ number }}
    {{ /numbers }}
'''

def double(item):
    item['number'] *= 2
    return item

context['numbers'] = map(double, context['numbers'])

renderer = pystache.Renderer()
print(renderer.render(pystache.parse(template), context))

The expected output is, modulo whitespace,

* 2
* 4
* 6

* 2
* 4
* 6

But the actual output is just

* 2
* 4
* 6

The problem is that context['numbers'] is an iterator, not an ordinary list. Once we’ve iterated through it once, its items are “used up” and so it acts like an empty list on subsequent iterations. I’d argue that this is a bug—rendering a Mustache template shouldn’t ever alter the context.

In the meantime, you can get around this bug by manually converting the iterator to a list. In the example above, using the line

context['numbers'] = list(map(double, context['numbers']))

gives the expected output.

(I tried to fix this by changing part of RenderEngine.fetch_section_data:

try:
    iter(data)
except TypeError:
    # Then the value does not support iteration.
    data = [data]
else:
    if is_string(data) or isinstance(data, dict):
        # Do not treat strings and dicts (which are iterable) as lists.
        data = [data]
    else:
        # Otherwise, treat the value as a list.  We call list() on it
        # in case it's a generator--if we didn't do this then iterating
        # through it once would "use up" all of the values and trying
        # to iterate a second time would produce nothing.
        data = list(data)

but that didn’t do it. I think the call to list() needs to be made somewhere earlier in the call stack.)