Online vectorized map construction provides foundational spatial representations of drivable areas for autonomous driving systems. Although transformer-based methods demonstrate improved accuracy, they remain vulnerable to sensor limitations. Prior knowledge integration has emerged as a promising direction to mitigate these limitations. However, short-term temporal priors offer freshness but lack global context, while historical global priors provide comprehensive spatial knowledge but risk structural obsolescence. To resolve this, we introduce PriorsFusionMap, a unified framework that synergistically integrates both prior types. In the framework, we propose a Prior-based Temporal Feature Integration (PTFI) module to enhance current BEV features using short-term priors and an Adaptive Gated Feature Fusion (AGFF) module to dynamically fuse enhanced features with historical map features via spatial gating to prevent overreliance on outdated structures. In addition, we use the Distribution Correction-Based Query Initialization (DCQI) strategy to accelerate convergence by focusing queries on spatio-temporal prior ROIs. The framework employs lightweight rasterized representations for efficient historical map updating. Evaluations on nuScenes and Argoverse 2 datasets demonstrate state-of-the-art performance (75.1% and 78.5% mAP), confirming the effectiveness of the priors fusion framework.