how to make a DISTINCT of STRING_AGG - sql

Question

I am trying to do a DISTINCT of the offerDetails column of the STRING_AGG function
OUTER APPLY
(
SELECT
STRING_AGG(CAST(co.OfferId AS NVARCHAR(MAX)), ';') AS OfferDetails,
STRING_AGG(CAST(ct.ConferenceId AS NVARCHAR(MAX)), ';') AS CustomTransferDetails
FROM ConferenceTransfer ct
INNER JOIN ConferenceOffer co ON ct.ConferenceId = co.ConferenceId
WHERE
ct.ConferenceTransferTypeId = 1 AND
ct.DateStarted IS NOT NULL AND
ct.CallerUserId = #agentId AND
ct.DateInitiated BETWEEN #dateFrom AND #dateTo
) CallTransfered

Answer 1

I'm not aware of any native way of making string_agg distinct, but you can ignore all but the first instance of each value by using a subquery and row_number to identify the first instance, i.e.
OUTER APPLY
(
SELECT
STRING_AGG(CAST(CASE WHEN c.RowNumber = 1 THEN c.OfferId END AS NVARCHAR(MAX)), ';') AS OfferDetails,
STRING_AGG(CAST(c.ConferenceId AS NVARCHAR(MAX)), ';') AS CustomTransferDetails
FROM ( SELECT co.OfferId,
ct.ConferenceId,
RowNumber = ROW_NUMBER() OVER(PARTITION BY co.OfferId ORDER BY ct.ConferenceId)
FROM ConferenceTransfer ct
INNER JOIN ConferenceOffer co
ON ct.ConferenceId = co.ConferenceId
WHERE
ct.ConferenceTransferTypeId = 1 AND
ct.DateStarted IS NOT NULL AND
ct.CallerUserId = #agentId AND
ct.DateInitiated BETWEEN #dateFrom AND #dateTo
) AS c
) CallTransfered
The key part being CASE WHEN c.RowNumber = 1 THEN c.OfferId END, so for anything other than the first instance, you are passing NULL, so the end result will only contain each OfferId once.

Answer 2

You need to group by OfferId, then group again
OUTER APPLY
(
SELECT
STRING_AGG(CAST(c.OfferId AS NVARCHAR(MAX)), ';') AS OfferDetails,
STRING_AGG(c.CustomTransferDetails, ';') AS CustomTransferDetails
FROM (
SELECT
co.OfferId,
STRING_AGG(CAST(ct.ConferenceId AS NVARCHAR(MAX)), ';') AS CustomTransferDetails
FROM ConferenceTransfer ct
INNER JOIN ConferenceOffer co ON ct.ConferenceId = co.ConferenceId
WHERE
ct.ConferenceTransferTypeId = 1 AND
ct.DateStarted IS NOT NULL AND
ct.CallerUserId = #agentId AND
ct.DateInitiated BETWEEN #dateFrom AND #dateTo
GROUP BY
co.OfferId
) c
) CallTransfered
This obviously only works for one DISTINCT. If you need both to be distinct, you can re-split the values, distinct them and re-aggregate
OUTER APPLY
(
SELECT
(
SELECT
STRING_AGG(s.value, ';')
FROM (
SELECT DISTINCT s.value
FROM STRING_SPLIT(OfferDetails, ';') s
) s
) AS OfferDetails,
(
SELECT
STRING_AGG(s.value, ';')
FROM (
SELECT DISTINCT s.value
FROM STRING_SPLIT(ConferenceId, ';') s
) s
) AS CustomTransferDetails
FROM (
SELECT
STRING_AGG(CAST(co.OfferId AS NVARCHAR(MAX)), ';') AS OfferDetails,
STRING_AGG(CAST(ct.ConferenceId AS NVARCHAR(MAX)), ';') AS CustomTransferDetails
FROM ConferenceTransfer ct
INNER JOIN ConferenceOffer co ON ct.ConferenceId = co.ConferenceId
WHERE
ct.ConferenceTransferTypeId = 1 AND
ct.DateStarted IS NOT NULL AND
ct.CallerUserId = #agentId AND
ct.DateInitiated BETWEEN #dateFrom AND #dateTo
GROUP BY
co.OfferId
) c
) CallTransfered