Xform rendering hides choices when relevance condition and choice filter are applied to a secondary question

[kitsao]

Describe the bug
Choices for a secondary select_one/select_multiple field do not show when both relevance condition and choice filter are applied. Consider the following design xls immunization section:

Question/Note/Label	Type	Required	Choices	Condition (Relevance/Show only if + Constraints/Validation)
Select the vaccines that were given on {{date}} Tick every condition/symptom that applies	select_multiple	yes	BCG, Birth Polio (OPV 0), OPV, PCV, DPT-HepB-Hib, IPV, Rota, Vitamin A, Measles	Don't show vaccines already reported. Birth Polio(OPV 0) if age <= 2weeks, OPV, PCV, Rota and DPT-HepB-Hib if age >= 6 weeks, Vit A if age >= 6 months, Measles if age >= 9 months
Select which dose of OPV was given on {{date}}	select_one	yes	OPV 1/2/3	If OPV is selected. Show OPV 2 if age >= 10 weeks, OPV 3 if age >= 14 weeks. Don't show doses already reported.
Select which dose of PCV was given on {{date}}	select_one	yes	PCV 1/2/3	If PCV is selected. Show PCV 2 if age >= 10 weeks, PCV 3 if age >= 14 weeks. Don't show doses already reported.
Select which dose of DPT-HepB-Hib was given on {{date}}	select_one	yes	DPT-HepB-Hib 1/2/3	If DPT-HepB-Hib is selected. Show DPT-HepB-Hib 2 if age >= 10 weeks, DPT-HepB-Hib 3 if age >= 14 weeks. Don't show doses already reported.
Select which dose of Rota was given on {{date}}	select_one	yes	Rota 1/2	If Rota is selected. Show Rota 2 if age >= 10 weeks. Don't show doses already reported.
Select which dose of Measles was given on {{date}}	select_one	yes	Measles 1/2	If Rota is selected. Show Measles 2 if age >= 18 months. Don't show doses already reported.

To Reproduce
Steps to reproduce the behavior:

1. Configure the above, a sample can be found here
2. Convert and upload the form and pick a client profile to complete it
3. The behavior in the image below is observed, doses do not show
   

Expected behavior
The expected behavior is to have the doses (choices) showing appropriately, as shown below. This happens when I either remove the relevance condition or the choice filter.

Logs
No logs on browser console.

Screenshots
See captures above.

Environment

• Instance: pih-malawi.dev.medicmobile.org
• Browser: Chrome
• Client platform: Linux
• App: webapp
• Version: 3.4.0

Additional context
This was experienced in 3.3.0 too. The workaround applied is merging vaccines and doses, i.e. instead of asking to select dose say for OPV, have vaccines as OPV 1, OPV 2 and OPV 3 ...

cc @derickl @smbuthia

Why Now?
Partner(s): Internal
Reason(s): A partner will have CHWs/Supervisors with a detrimental experience, thus providing bad care

[SCdF]

Worked in 2.14 / 2.x (according to @derickl and @smbuthia). We should work out which version of Enketo broke this, and if it's been subsequently fixed in a later version.

[tookam]

@kitsao Thanks for sharing the sample but it fails to load ..

has the form been updated?
(pih-malawi.dev.medicmobile.org produces 502)

[garethbowen]

Possibly related: #4249

[Hareet]

@wtekeu
pih-malawi.dev.medicmobile.org is back up. Please check again.

[tookam]

@kitsao After doing some research, it looks like this is due to the use of the following skip logic:
<bind nodeset="/under_5_screening/vaccines_given/visit_details" relevant="../visits_reported != 0"/>
When you remove this line it does work. Can you think of an alternative way to build the skip logic?

[abbyad]

@wtekeu does it work with different relevant conditions, such as relevant="true()"?

[tookam]

@abbyad it does work but if you select Zero immunization to report then click on Next then back with Prev then select One immunization to report it does not work.

[tookam]

But when you remove the line, it works in both cases. The only drawback is that if you select Zero you get a blank page with just the title of the form and the buttons Prev, Next and the link Cancel

[abbyad]

Are you confirming that a valid relevant statement doesn't work as expected? Is this related to our version of Enketo, or Enketo proper? It would be good for us to handle this rather than not using relevant statements.

[garethbowen]

Bumped to 3.8.0 so it doesn't hold up the release.

[garethbowen]

No update for some time. Bumping.

[mrjones-plip]

@jkuester - can you tell if the enketo uplift in 4.0 will fix this? If yes, I'll move it there. Thanks!

[garethbowen]

@jkuester Any update on this?

[jkuester]

@garethbowen shoot! This fell off my radar somehow.

@kitsao, sorry to resurrect this ancient history, but I am having some trouble recreating this and was wondering if you could give me access to the Google xlsx form that you linked in the original issue (if it still exists). I have requested access via the link already. I tried cobbling together a form based on a combination of the PIH under_5_screening and immunization forms, but even the earliest versions of those forms in the medic repos appear to have already been refactored to fix the issue....

[jkuester]

Okay, spoke too soon! I realized that my test form did not have anything set for the choice_filter. I am now able to recreate this behavior. TLDR is that this issue appears to have been fixed by the Enketo uplift!

Here is the most basic config that I have come up with that demonstrates the issue:

type	name	label::en	relevant	appearance	choice_filter	repeat_count
begin group	my_group	NO_LABEL		field-list
integer	question_count	Repeat Count
begin repeat	my_repeat	NO_LABEL	${question_count} != 0			${question_count}
select_one yes_no	my_question	You should be able to see the choices	true()		true()
end repeat
end group

The best I can tell, what seems to be happening in the old version of Enketo is that the order of evaluations gets confused so the choice_filter value is not properly evaluated for questions inside a repeat when the repeat's relevant and repeat_count logic are both calculated (yeah the chain of dependencies here is mind-numbing).

This is what I see when running the above form on 3.4:

And this is what I see on 6345-enketo-uplift:

[garethbowen]

Great! I'll move this issue to 4.0.0 to line up with the upgrade. Thanks.

[jkuester]

@medic/quality-assurance this issue is ready for AT on #7256

[jkuester]

@tatilepizs here is the xml for a sample for that you can use to test this behavior:

<?xml version="1.0"?>
<h:html xmlns="http://www.w3.org/2002/xforms" xmlns:h="http://www.w3.org/1999/xhtml" xmlns:ev="http://www.w3.org/2001/xml-events" xmlns:xsd="http://www.w3.org/2001/XMLSchema" xmlns:jr="http://openrosa.org/javarosa" xmlns:orx="http://openrosa.org/xforms">
  <h:head>
    <h:title>Enketo Widgets</h:title>
    <model>
      <itext>
        <translation lang="en">
          <text id="/enketo_widgets/my_group/my_repeat/my_question:label">
            <value>You should be able to see the choices</value>
          </text>
          <text id="/enketo_widgets/my_group/question_count:label">
            <value>Repeat Count</value>
          </text>
          <text id="static_instance-yes_no-0">
            <value>Yes</value>
          </text>
          <text id="static_instance-yes_no-1">
            <value>No</value>
          </text>
          <text id="static_instance-yes_no-2">
            <value>Don't Know</value>
          </text>
          <text id="static_instance-yes_no-3">
            <value>Not Applicable</value>
          </text>
        </translation>
      </itext>
      <instance>
        <enketo_widgets id="enketo_widgets" prefix="J1!enketo_widgets!" delimiter="#" version="2022-08-16 00:00:00">
          <my_group>
            <question_count/>
            <my_repeat_count/>
            <my_repeat jr:template="">
              <my_question/>
            </my_repeat>
          </my_group>
          <meta tag="hidden">
            <instanceID/>
          </meta>
        </enketo_widgets>
      </instance>
      <instance id="contact-summary"/>
      <instance id="yes_no">
        <root>
          <item>
            <itextId>static_instance-yes_no-0</itextId>
            <name>yes</name>
          </item>
          <item>
            <itextId>static_instance-yes_no-1</itextId>
            <name>no</name>
          </item>
          <item>
            <itextId>static_instance-yes_no-2</itextId>
            <name>dk</name>
          </item>
          <item>
            <itextId>static_instance-yes_no-3</itextId>
            <name>na</name>
          </item>
        </root>
      </instance>
      <bind nodeset="/enketo_widgets/my_group/question_count" type="int"/>
      <bind nodeset="/enketo_widgets/my_group/my_repeat_count" type="string" readonly="true()" calculate=" /enketo_widgets/my_group/question_count "/>
      <bind nodeset="/enketo_widgets/my_group/my_repeat" relevant=" /enketo_widgets/my_group/question_count  != 0"/>
      <bind nodeset="/enketo_widgets/my_group/my_repeat/my_question" type="select1" relevant="true()"/>
      <bind nodeset="/enketo_widgets/meta/instanceID" type="string" readonly="true()" calculate="concat('uuid:', uuid())"/>
    </model>
  </h:head>
  <h:body class="pages">
    <group appearance="field-list" ref="/enketo_widgets/my_group">
      <input ref="/enketo_widgets/my_group/question_count">
        <label ref="jr:itext('/enketo_widgets/my_group/question_count:label')"/>
      </input>
      <group ref="/enketo_widgets/my_group/my_repeat">
        <label ref="jr:itext('/enketo_widgets/my_group/my_repeat:label')"/>
        <repeat nodeset="/enketo_widgets/my_group/my_repeat" jr:count=" /enketo_widgets/my_group/my_repeat_count ">
          <select1 ref="/enketo_widgets/my_group/my_repeat/my_question">
            <label ref="jr:itext('/enketo_widgets/my_group/my_repeat/my_question:label')"/>
            <itemset nodeset="instance('yes_no')/root/item[true()]">
              <value ref="name"/>
              <label ref="jr:itext(itextId)"/>
            </itemset>
          </select1>
        </repeat>
      </group>
    </group>
  </h:body>
</h:html>

[tatilepizs]

Config: Default
Environment: Local with docker helper script
Platform: WebApp
Browser: Chrome

---

The file used to test was provided by @jkuester in this comment

Reproducible on Master

Options are not shown.

Desktop

Mobile

Fixed on 6345-enketo-uplift

Options are shown successfully.

Desktop

Mobile