Allowing xml-internal question type.

[nribeka]

Fix #176.

This would allow question of type xml-internal.

[nribeka]

@lognaturel see also my comment in #176. Is this the expected behavior?

[lognaturel]

@nribeka is this behavior on master or just on this branch? What happens if you actually include a select on the fruits list, is it inlined or added as an instance?

[nribeka]

@lognaturel, this is in master.

As long as you add choice_filter it will add everything from the choices sheet as separate instance even though it's not being used by any question.

If you add select on fruits, then it will add it to the body and also create an instance of it.

      <instance id="fruits">
        <root>
          <item>
            <label>Apple</label>
            <name>apple</name>
          </item>
          <item>
            <label>Banana</label>
            <name>banana</name>
          </item>
        </root>
      </instance>

    <select1 ref="/data/fruit">
      <label>The Fruit</label>
      <item>
        <label>Apple</label>
        <value>apple</value>
      </item>
      <item>
        <label>Banana</label>
        <value>banana</value>
      </item>
    </select1>

[nribeka]

I think it's coming from this line: https://github.com/XLSForm/pyxform/blame/master/pyxform/xls2json.py#L1119

                if row.get("choice_filter"):
                    if select_type == "select one external":
                        new_json_dict["query"] = list_name
                    else:
                        new_json_dict["itemset"] = list_name
                        json_dict["choices"] = choices
                        if choices.get(list_name):
                            new_json_dict["list_name"] = list_name
                            new_json_dict[constants.CHOICES] = choices[list_name]

This assignment: json_dict["choices"] = choices is adding the whole choices sheet to the json.

[lognaturel]

There should not be primary instance nodes created for xml-internal questions. I think that ideally when a list name that doesn't exist in the choices tab is specified, that should lead to an error.

It would be helpful to add a test that verifies that extra columns' values are added to the secondary instance. Luckily, it looks like that behavior already works.

For this form:

            | survey  |                    |            |           |            |                   |
            |         | type               | name       | label     | hint       |appearance         |
            |         | xml-internal       | states     | The state |            |                   |
            |         | xml-internal       | yes_no     |           |            |                   |
            | choices |                    |            |           |            |                   |
            |         | list_name          | name       | label     | capital    | color             |
            |         | states             | VA         | Virginia  | Richmond   | orange            |
            |         | states             | IN         | Indiana   | Indianapolis|                  |
            |         | fruits             | apple      | Apple     |            | red               |
            |         | fruits             | banana     | Banana    |            | yellow            |

we expect to see capital and color nodes in the secondary instances where those have values in the form definition.

[lognaturel]

fruits isn't used anywhere so I wouldn't expect it be generated. I think this is existing behavior so maybe we don't need to change it here?

[lognaturel]

The two assertion blocks seem redundant to me. Perhaps it's best to go with the first variant so the count can be verified.