repeated set appears to be malfunctioning

[Pomax]

I'm using the following nunjucks fragment:

      {% set title = "photostream" %}
      {% set prev = photo.prev %}
      {% set next = photo.next %}
      {% include "streamnav.html" %}

      {% for setObj in photo.sets %}
        {% set photoset = photosets[setObj.id] %}
        {% set title = photoset.title._content %}
        {% set prev = setObj.prev %}
        {% set next = setObj.next %}
        {% include "streamnav.html" %}
      {% endfor %}
    </div>

However, when templating is performed inside the for loop, the title and other variables, despite having been rebound using the set command, are still the value they received on first `set.

[mattbasta]

Can you pare down your snippet into a standalone example? I think I know why this is happening and have an idea how to fix it, but want to make sure it works with your use case.

[Pomax]

no problem. Reduced:

test.html

{% set a = b %}
{% include "a.html" %}
{% for c in d %}
  {% set a = c %}
  {% include "a.html" %}
{% endfor %}

a.html

a = {{ a }}

app.js

app.get('/test', function(req, res) {
  res.render("test.html", {
    b: "b",
    d: ["1", "2", "3"]
  })
})

[mattbasta]

So this is why you're seeing the issue:

1. When you do the original set a = b, the compiler generates a context.set('item', b)[1].
2. When you do set a = c, the compiler generates frame.set('item', c).
3. When you do {{ a }} in a.html, the compiler generates runtime.contextOrFrameLookup(context, frame, 'item').
4. contextOrFrameLookup does a context.lookup('item') and sees that it has a value (b's value) and returns that.

I'm not sure why contextOrFrameLookup looks at context before frame, as that seems like a bug (frame will always override the context, I think). I have a PR that fixes the issue, but I'll let @jlongster look at it and see if there's a strong reason to not flip the context/frame lookup order.

[1] Simplified for the sake of explanation

[jlongster]

@Pomax does this fix your issue?

[Pomax]

Completely missed the pull request - I'll test this in a few hours!

[Pomax]

@jlongster yup: that seems to do the trick!