amogh-jahagirdar commented on code in PR #14867:
URL: https://github.com/apache/iceberg/pull/14867#discussion_r2673131242
##########
core/src/main/java/org/apache/iceberg/rest/RESTCatalogProperties.java:
##########
@@ -37,12 +37,107 @@ private RESTCatalogProperties() {}
public static final String NAMESPACE_SEPARATOR = "namespace-separator";
- // Enable planning on the REST server side
- public static final String REST_SCAN_PLANNING_ENABLED =
"rest-scan-planning-enabled";
- public static final boolean REST_SCAN_PLANNING_ENABLED_DEFAULT = false;
+ // Configure scan planning mode
+ // Can be set by server in LoadTableResponse.config() or by client in
catalog properties
+ // Negotiation rules: ONLY beats PREFERRED, both PREFERRED = client wins
+ // Default when neither client nor server provides: client-preferred
+ public static final String SCAN_PLANNING_MODE = "scan-planning-mode";
+ public static final String SCAN_PLANNING_MODE_DEFAULT =
+ ScanPlanningMode.CLIENT_PREFERRED.modeName();
public enum SnapshotMode {
ALL,
REFS
}
+
+ /**
+ * Enum to represent scan planning mode configuration.
+ *
+ * <p>Can be configured by:
+ *
+ * <ul>
+ * <li>Server: Returned in LoadTableResponse.config() to advertise server
preference/requirement
+ * <li>Client: Set in catalog properties to set client
preference/requirement
+ * </ul>
+ *
+ * <p>When both client and server configure this property, the values are
negotiated:
+ *
+ * <p>Values:
+ *
+ * <ul>
+ * <li>CLIENT_ONLY - MUST use client-side planning. Fails if paired with
CATALOG_ONLY from other
Review Comment:
I'm still a bit skeptical that we even need the notion of preferences, e.g.
client-preferred, catalog-referred. it''s plausible that servers could have
more insights to give a more intelligent preference but it feels over
complicated compared to just having a "clients-choice" or something instead of
2 preferences. It simplifies the decision matrix logic below, and clients can
then use their own heuristics.
I think that's what the decision as to if preferences or not makes sense,
comes down to:
is it better to have clients just make intelligent choices when server side
planning is available but not required, or is it better for servers to indicate
preferences. My thought process is if a server really feels like it's
advantageous to do remote planning, may as well just send it back as required.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
