In-Memory Data Node Retention via Priority Scoring

The patent applies local quality by differentiating memory retention strategies across different data nodes based on their priority scores. Critical data nodes that are frequently accessed or处于浅层位置 retain their data in memory, while less critical nodes clear their data. This creates a non-uniform memory management approach where each data node has a different retention policy, resolving the contradiction between memory capacity and re-materialization time.

The system performs preliminary actions by pre-calculating priority scores for all data nodes before memory management decisions are made. These priority scores are used to determine which nodes should retain data in memory proactively, before memory pressure occurs. This preliminary classification allows the system to quickly make retention decisions during runtime without recalculating priorities, thus reducing re-materialization time while efficiently using memory capacity.

The patent segments the processing graph into independent data nodes with calculated priority scores. When a change is detected, the system identifies and re-processes only the affected segments (nodes with priority above threshold) rather than the entire graph. This segmentation allows developers to maintain debugging accuracy by re-processing relevant portions while significantly improving processing speed by avoiding unnecessary re-computation of unrelated nodes.

The system applies partial action by re-processing only the necessary subset of data nodes required to observe the effects of changes, rather than performing excessive full-graph re-processing. The priority score threshold mechanism enables the system to perform just enough re-processing to achieve debugging goals, balancing accuracy requirements with productivity concerns.

The patent changes the parameter of memory retention by introducing dynamic priority scores that determine retention status. Instead of a static all-or-nothing approach, the system continuously adjusts which data nodes retain memory based on their priority scores, which are calculated based on factors like depth in processing graph and accessibility importance. This parameter-based approach optimizes the balance between viewing speed and memory consumption.

The memory retention strategy is made dynamic through the priority score mechanism. As data nodes are accessed, modified, or re-processed, their priority scores are updated, which dynamically changes their memory retention status. This dynamic approach allows the system to adapt memory consumption to actual usage patterns, maintaining fast viewing speed for critical data while freeing memory for less critical nodes.