perf: optimize getAllLinks query from ~2.8s to <200ms

[armand1m]

Problem

The getAllLinks GraphQL query was taking ~2.77s in production due to an N+1 query pattern. For each of the 50 links returned, PostGraphile issued 3 separate SQL queries:

1. linkUsageMetrics { totalCount } — COUNT(*) with RLS per link
2. recentUsageMetrics — full scan per link (no composite index)
3. linkAllowedEmails — indexed lookup per link (cheap, kept as-is)

This resulted in ~150 SQL queries per page load.

Changes

Database

• New columns on links: usage_count BIGINT and last_accessed_at TIMESTAMP — denormalized from link_usage_metrics
• Composite index on link_usage_metrics (link_id, accessed_at DESC) — speeds up recentUsageMetrics range scan
• Trigger trg_update_link_usage_stats on link_usage_metrics — keeps usage_count and last_accessed_at in sync on INSERT/DELETE
• Backfill — populates existing data from link_usage_metrics

GraphQL

• Replaced linkUsageMetrics { totalCount } (N+1 COUNT query) with usageCount (scalar column read, zero additional SQL)
• Applied to both getAllLinks and searchLinks queries

Frontend

• LinkTable now reads link.usageCount directly instead of link.linkUsageMetrics.totalCount

Expected Impact

Metric	Before	After
Total SQL queries	~150	~51
linkUsageMetrics	COUNT(*) with RLS per link	Scalar column read (0 queries)
recentUsageMetrics	Full scan per link	Composite index range scan

Deployment Notes

• The migration includes a backfill step that updates all existing links rows
• The trigger adds minimal overhead to link_usage_metrics INSERT/DELETE operations
• The migration is reversible (see -- migrate:down section)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b757d77a03

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Make metric trigger bypass link RLS

When Auth0 is enabled, anonymous users only get read:golinks, and create_link_usage_metric allows inserts with that permission, but this trigger's UPDATE public.links still runs as the invoking postgraphile role and is subject to the update_links RLS policy requiring update:golinks. In the redirect flow (pages/[...alias].tsx records this mutation after a link is opened), read-only viewers will have their metric insert fail as soon as this trigger tries to update the denormalized columns; define the function with appropriate definer privileges/RLS handling so read-only metric inserts can still succeed.

Useful? React with 👍 / 👎.