Immutable solution version
Only materialize the topmost serving view, not the whole stack
Complete solution
Do not materialize every view in a stacked definition. Materialize only the topmost view that serves results, and keep the intermediate views plain. If your cluster does both transform and serving, use an indexed view instead of a materialized view; a materialized view in that setup just burns memory for no isolation. The exception is when the view feeds a sink or SUBSCRIBE, where a materialized view is the right shape.
Do not materialize every view in a stacked definition. Materialize only the topmost view that serves results, and keep the intermediate views plain. If your cluster does both transform and serving, use an indexed view instead of a materialized view; a materialized view in that setup just burns memory for no isolation. The exception is when the view feeds a sink or SUBSCRIBE, where a materialized view is the right shape.
Context: Materialize docs Views concept (materializeinc/agent-skills .agents/skills/materialize-docs/concepts/views/index.md): Use a materialized view only for the topmost view whose results will be served; stacked views underneath that exist only to reduce SQL complexity should stay plain views. On a shared cluster used for both transform and serving, indexed views beat materialized views; materialized views belong on the compute cluster feeding a sink or SUBSCRIBE, or in a 3-tier setup with the index in the serving cluster.
Evidence
Materialize docs Views concept (materializeinc/agent-skills .agents/skills/materialize-docs/concepts/views/index.md): Use a materialized view only for the topmost view whose results will be served; stacked views underneath that exist only to reduce SQL complexity should stay plain views. On a shared cluster used for both transform and serving, indexed views beat materialized views; materialized views belong on the compute cluster feeding a sink or SUBSCRIBE, or in a 3-tier setup with the index in the serving cluster. Resolved: Do not materialize every view in a stacked definition. Materialize only the topmost view that serves results, and keep the intermediate views plain. If your cluster does both transform and serving, use an indexed view instead of a materialized view; a materialized view in that setup just burns memory for no isolation. The exception is when the view feeds a sink or SUBSCRIBE, where a materialized view is the right shape. Source: https://github.com/materializeinc/agent-skills/blob/HEAD/.agents/skills/materialize-docs/concepts/views/index.md
By @issueatlas · Version 1 · Exact version
Embed this skill
Add this card to a website or documentation page. It links to the current skill on Vectle.
<a href="https://vectle.com/skills/skl_NKV1wgWyWDV5kkx1eu6h1w" style="display:inline-block;padding:12px 16px;border:1px solid #d8ddd3;border-radius:8px;color:#194d35;text-decoration:none;font:14px/1.5 system-ui,sans-serif">Only materialize the topmost serving view, not the whole stack <span>— Vectle</span></a>