mirror of
https://github.com/windmill-labs/windmill.git
synced 2026-08-20 16:02:19 +00:00
Ruben Fiszel
eba70ce735
Folder labels are exposed verbatim as `inherited_labels` (via the
`folder_labels` SQL function) and rendered in keyed `{#each}` blocks that
throw Svelte's `each_key_duplicate` on a repeated key, crashing the list
views. The UI dedups labels on entry, but API / CLI / git-sync writes do
not, so a folder.yaml with `labels: [foo, foo]` persists duplicates.
- Dedup on write in create_folder and update_folder (order-preserving).
- Make folder_labels() dedup on read so it is resilient regardless of how a
row was populated, plus a one-time cleanup of already-persisted duplicates
so direct folder.labels reads (folder list, editor) are safe too.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
34 lines
1.3 KiB
PL/PgSQL
34 lines
1.3 KiB
PL/PgSQL
-- Folder labels are exposed verbatim as `inherited_labels` (via folder_labels) and
|
|
-- rendered in keyed `{#each}` blocks in the UI, which throw `each_key_duplicate` on a
|
|
-- repeated key. The write paths now dedup, but make the read resilient regardless of
|
|
-- how a row was populated, and clean up any duplicates already persisted.
|
|
|
|
-- Dedup while preserving first-seen order.
|
|
CREATE OR REPLACE FUNCTION folder_labels(w_id text, item_path text) RETURNS text[]
|
|
LANGUAGE sql STABLE SECURITY DEFINER SET search_path = public AS $$
|
|
SELECT (
|
|
SELECT array_agg(l ORDER BY first_ord)
|
|
FROM (
|
|
SELECT u.l, min(u.ord) AS first_ord
|
|
FROM unnest(f.labels) WITH ORDINALITY AS u(l, ord)
|
|
GROUP BY u.l
|
|
) deduped
|
|
)
|
|
FROM folder f
|
|
WHERE f.workspace_id = w_id AND item_path LIKE 'f/%' AND f.name = split_part(item_path, '/', 2)
|
|
$$;
|
|
|
|
-- One-time cleanup of rows that already contain duplicate labels, so direct reads of
|
|
-- folder.labels (folder list, editor) are also safe.
|
|
UPDATE folder
|
|
SET labels = (
|
|
SELECT array_agg(l ORDER BY first_ord)
|
|
FROM (
|
|
SELECT u.l, min(u.ord) AS first_ord
|
|
FROM unnest(labels) WITH ORDINALITY AS u(l, ord)
|
|
GROUP BY u.l
|
|
) deduped
|
|
)
|
|
WHERE labels IS NOT NULL
|
|
AND cardinality(labels) <> (SELECT count(DISTINCT x) FROM unnest(labels) AS x);
|