Prevent Duplicate Answers in an XLSForm Repeat

To stop the same answer appearing twice in an XLSForm repeat, put count(../../member[member_id = current()]) = 1 in the constraint column of the question inside the repeat. It counts the repeat instances whose member_id equals this answer, the current instance included, so any value above 1 is a duplicate. The same constraint works for text, integer and barcode questions in ODK, KoboToolbox and SurveyCTO.

Download example form (.xlsx)All five examples in one XLSForm, one group each. Tested with pyxform and ODK Validate; free to reuse under CC BY 4.0.Download

What is the constraint that blocks duplicates in a repeat?

A household roster records an ID card number for every member. Two members cannot share a number, so the ID question rejects any number already entered for another member:

type	name	label	constraint	constraint_message
begin repeat	member	Household member
text	member_name	Name of household member
text	member_id	National ID number of ${member_name}	count(../../member[member_id = current()]) = 1	This ID is already recorded for another member.
end repeat	member

◀ ▶surveychoicessettings+

Read the expression from the inside out:

1. current() is the answer being checked, here this member's ID.
2. ../../member walks up from the question: .. is this repeat instance, ../.. is the level that holds the repeat, and /member selects every instance of the repeat.
3. [member_id = current()] keeps the instances whose ID equals this answer.
4. count(...) = 1 passes when only one instance has it, which is the current one.

Use the repeat's name from the begin repeat row, not its label. Constraints do not run on blank answers, so two members with no ID yet are not flagged; make the question required if it must always be filled.

Does the same check work for barcodes and numbers?

Yes. The constraint does not care how the answer was entered. A lab sample log scans one tube per child, and a tube scanned twice means another tube was skipped:

type	name	label	constraint	constraint_message
begin repeat	sample	Blood sample
integer	child_line	Line number of the child on the roster
barcode	tube_code	Scan the barcode on the sample tube	count(../../sample[tube_code = current()]) = 1	This tube was already scanned in this form. Scan a new tube.
end repeat	sample

◀ ▶surveychoicessettings+

For an integer question the comparison is numeric, so 7 and 07 count as the same answer. For text and barcode it is an exact text match: A12 and a12 are different answers. When enumerators type IDs by hand, prefer numeric IDs or a scan over free text names.

Why does my join() and selected() check fail when I finalize?

The most shared forum answer for this task builds a list of every ID outside the repeat and checks the new one against it:

type,name,label,calculation,constraint
calculate,all_ids,,join(' ', ${member_id}),
begin repeat,member,Household member,,
text,member_id,Member ID,,not(selected(${all_ids}, .))
end repeat,member,,,

It compiles, and it often seems to work while you type. Then Collect refuses to finalize and sends you back to the first member with the constraint message. The reason: all_ids includes the current member's own ID. While you type, the constraint can run before all_ids updates. When you finalize, every constraint runs again against the full list, every ID finds itself in it, and every member fails.

The count(...) = 1 constraint has no such problem, because it expects to find the current answer once. If you have a form built on join(), replace the constraint and delete the all_ids calculate.

How do I hide options already picked in an earlier instance?

For a select_one inside a repeat, it is kinder to remove the used options than to reject them. A crop module asks about each crop the household grew, and each crop should come up once:

type	name	label	calculation	choice_filter
begin repeat	crop_rep	Crop grown this season
calculate	crop_pos		position(..)
calculate	crops_other		join(' ', ../../crop_rep[crop_pos != current()/../crop_pos]/crop)
select_one crop	crop	Which crop is this?		not(selected(${crops_other}, name))
integer	crop_area	Area planted with this crop (decimals)
end repeat	crop_rep

◀ ▶surveychoicessettings+

list_name	name	label
crop	1	Rice
crop	2	Maize
crop	3	Wheat
crop	4	Jute
crop	5	Potato
crop	-96	Other

◀ ▶surveychoicessettings+

How it works:

• crop_pos stores the number of the current instance with position(..).
• crops_other joins the crops of every other instance into a space-separated list, such as 1 4. Excluding the current instance matters: if it included its own answer, the chosen crop would vanish from its own list when you go back to it.
• The choice_filter keeps only the choices whose name is not in that list.

The same filter works with a select_multiple in the repeat, since crops_other then holds every ticked name. To filter choices by a select_multiple outside a repeat, see choice filters with select_multiple.

How do I block a duplicate combination of two answers?

Names repeat in many households, so a roster that checks the name alone rejects real members. Check the name and the age together instead, on the second of the two questions:

type	name	label	constraint	constraint_message
begin repeat	person	Household member
text	person_name	Name of household member
integer	person_age	Age of ${person_name} in completed years	count(../../person[person_name = current()/../person_name and person_age = current()]) = 1	A member with this name and age is already on the roster. Check for a double entry.
end repeat	person

◀ ▶surveychoicessettings+

Inside the brackets, a bare name like person_name means the field of each instance being counted. current()/../person_name means the name in the instance being checked: current() is the age, .. is its instance, and /person_name is the name next to it.

How do I check duplicates in nested repeats?

The number of ../ steps decides how far the check reaches. A farm survey lists farmers, and each farmer's plots inside a nested repeat. Plot codes restart for every farmer, so the plot code only has to be unique within one farmer:

type	name	label	constraint	constraint_message
begin repeat	farmer	Farmer
text	farmer_name	Name of the farmer
begin repeat	plot	Plot
integer	plot_no	Plot number for ${farmer_name}	count(../../plot[plot_no = current()]) = 1	This farmer already has a plot with this number.
text	gps_tag	Survey tag on the plot boundary marker	count(../../../farmer/plot[gps_tag = current()]) = 1	This marker tag is already recorded for another plot in this form.
end repeat	plot
end repeat	farmer

◀ ▶surveychoicessettings+

• ../../plot stays inside the current farmer: .. is this plot, ../.. is this farmer.
• ../../../farmer/plot goes up one more level to the form and back down through every farmer, so the boundary marker tag must be unique across the whole form.

To read single answers from other instances rather than count them, use indexed-repeat().

Common duplicate-check mistakes and how to fix them

The error text below is what ODK Validate returned for each broken form. Several mistakes compile without any error and only show up on the phone:

You wrote	Write instead	What you will see
not(selected(${all_ids}, .)) with all_ids outside the repeat	count(../../member[member_id = current()]) = 1	No error at upload; the form will not finalize and flags every member
count(../../member[member_id = current()]) = 0	... = 1	No error at upload; every answer is rejected, because the count includes the current instance
count(../../member[${member_id} = current()]) = 1	count(../../member[member_id = current()]) = 1	No error at upload; every answer is rejected, because ${member_id} becomes ../member_id inside the brackets and matches nothing
count(../member[member_id = current()]) = 1	count(../../member[member_id = current()]) = 1	No error at upload; every answer is rejected, because one ../ stays inside the current instance
join(' ', ${member_id}) in a calculate inside the repeat	Put the calculate outside the repeat, or use count()	No error at upload; the list holds only the current answer
count(...) == 1	count(...) = 1	contains invalid constraint expression ... Bad node

When a check rejects answers it should accept, add a temporary calculate inside the repeat with count(../../member[member_id = current()/../member_id]) and show it in a note, to see what the form is counting. More messages are explained in common XLSForm errors.

Does this work the same in ODK, KoboToolbox and SurveyCTO?

Yes. count(), current(), position(..), join() and selected() are part of the ODK XPath used by all three, and relative paths like ../../member read the same way. What differs is where you type the rule:

Platform	Where the duplicate check goes
ODK (XLSForm)	constraint column of the question inside the repeat; filters in choice_filter
KoboToolbox Formbuilder	Question settings, Validation Criteria, manually entered expression; use XLSForm upload for the choice filter
KoboToolbox (XLSForm upload)	Same as ODK
SurveyCTO	constraint column of the field inside the repeat group (skip logic goes in a column named relevance)

Test the finished form in both SurveyLoopr Collect and the web form before fieldwork. Fill two instances with the same answer, check that the second is refused, then correct it and finalize, which is the step the join() method fails.

Why duplicate checks belong in the form

A duplicated household member or a tube scanned twice is cheap to fix while the enumerator is still in the house and expensive to fix in the office. After upload, someone has to decide which of two identical rows is real, and the missing tube may never be found. A constraint costs one cell, and it catches the error at the moment it happens.

Build my first form free

Next: XLSForm repeat groups →