fix: optimize notification_log_wallet_address insert performance#1861
fix: optimize notification_log_wallet_address insert performance#1861
Conversation
Batch wallet address inserts into groups of 100 to reduce Hasura CTE query overhead, and fix v2 schema logging wallet addresses once instead of duplicating per localisation. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
|
@codex review |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 406900875e
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".
| const logMessage = | ||
| schemaVersion === "v1" | ||
| ? (parsedParams as SendNotificationBodyV1).message | ||
| : (parsedParams as SendNotificationBodyV2).localisations?.[0]?.message; |
There was a problem hiding this comment.
When schemaVersion is v2, the logged message is taken from localisations?.[0]?.message, which depends on client-supplied array order. This means the same notification payload can produce different notification_log.message values if localisations are reordered, and logs can end up in a non-English language even though validation guarantees an en localisation exists. After consolidating to one log row per send, pick a stable locale (e.g. en with fallback) so log data remains consistent and queryable.
Useful? React with 👍 / 👎.
Pick the "en" localisation for the log message instead of relying on client-supplied array order, with fallback to the first entry. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2 participants
PR Type
Description
Optimizes the
notification_log_wallet_addressbulk insert query that was consuming a large percentage of DB query runtime. Two changes:Batch inserts into groups of 100 — Hasura wraps bulk inserts in a CTE with
RETURNING *and constraint checks. Large batches (hundreds of wallet addresses) produce oversized queries with significant planning overhead. Batching keeps each query small and efficient.Fix v2 schema duplicating wallet address logs per localisation — Previously, for v2 schema notifications,
logNotificationwas called once per localisation entry, each time inserting the same full set of wallet addresses. This multiplied insert volume by the number of localisations. Now wallet addresses are logged once regardless of schema version.Also adds error handling for insert failures that were previously silently swallowed (fire-and-forget without try/catch).
Checklist