290-313: Consider error resilience for catalog initialization.

If OIDC initialization fails for any single catalog in Promise.all, the entire Iceberg API initialization fails. Consider whether partial initialization (some catalogs working) would be preferable, or ensure upstream guarantees prevent individual failures.

Alternative with Promise.allSettled for partial success

- const catalogs = await Promise.all(
- deploymentRegion.iceberg.map(async warehouseConfig => {
+ const catalogResults = await Promise.allSettled(
+ deploymentRegion.iceberg.map(async warehouseConfig => {
const oidc_iceberg = await createOidc({
...mergeOidcParams({
oidcParams,
oidcParams_partial: warehouseConfig.oidcParams
}),
transformBeforeRedirectForKeycloakTheme,
getCurrentLang,
autoLogin: true,
enableDebugLogs: enableOidcDebugLogs
});

return {
name: warehouseConfig.catalog,
warehouse: warehouseConfig.warehouse,
endpoint: warehouseConfig.endpoint,
getAccessToken: async (): Promise => {
if (!oidc_iceberg.isUserLoggedIn) return undefined;
return (await oidc_iceberg.getTokens()).accessToken;
}
};
})
);
+
+ const catalogs = catalogResults
+ .filter((r): r is PromiseFulfilledResult ? T : never> => r.status === "fulfilled")
+ .map(r => r.value);
+
+ if (catalogs.length === 0) {
+ break init_iceberg_api;
+ }

Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@web/src/core/bootstrap.ts` around lines 290 - 313, The current Promise.all
over deploymentRegion.iceberg will reject the whole catalogs initialization if
any createOidc call throws; change to per-item error handling so partial
successes are allowed: replace the Promise.all([...map...]) pattern with
Promise.allSettled or wrap each async map entry in a try/catch around
createOidc/oidc_iceberg, log the error, and return undefined/failure marker,
then filter out failed entries before assigning catalogs; reference the map over
deploymentRegion.iceberg, createOidc, oidc_iceberg, and the resulting catalogs
array when implementing this change.


22-29: Pre-index tables by catalog to avoid repeated full scans.

Line 25 filters the full allTables array once per catalog. This is O(catalogs * tables) and can get expensive on large catalogs.

Suggested refactor

- const catalogs: CatalogMetadata[] = icebergCatalogConfigs.map(config => {
+ const tableEntriesByCatalog = new Map();
+ for (const entry of allTables) {
+ const entries = tableEntriesByCatalog.get(entry.catalog) ?? [];
+ entries.push(entry);
+ tableEntriesByCatalog.set(entry.catalog, entries);
+ }
+
+ const catalogs: CatalogMetadata[] = icebergCatalogConfigs.map(config => {
const namespaceMap = new Map();

- for (const entry of allTables.filter(t => t.catalog === config.name)) {
+ for (const entry of tableEntriesByCatalog.get(config.name) ?? []) {
const tables = namespaceMap.get(entry.namespace) ?? [];
tables.push(entry.name);
namespaceMap.set(entry.namespace, tables);
}

Prompt for AI Agents

) or directly into Map> before the icebergCatalogConfigs.map, then iterate each config and pull its precomputed list to build namespaceMap without repeated full scans.">Verify each finding against the current code and only fix it if needed.

In `@web/src/core/usecases/icebergCatalog/thunks.ts` around lines 22 - 29, The
current loop in thunks.ts builds each catalog's namespaceMap by filtering
allTables for every config (see icebergCatalogConfigs, allTables, catalogs:
CatalogMetadata[], namespaceMap, entry), causing O(catalogs * tables) work; fix
it by pre-indexing allTables once into a Map keyed by catalog (e.g., MapTableEntry[]>) or directly into Map> before the
icebergCatalogConfigs.map, then iterate each config and pull its precomputed
list to build namespaceMap without repeated full scans.