Temporal Knowledge Graphs (TKGs) capture dynamic event behaviors with temporal information. However, existing TKG link prediction methods are predominantly embedding-based, lacking interpretability and transparency. This paper presents TempRuLe, a novel method for temporal rule learning on TKGs. By leveraging a new search strategy for temporal path patterns, TempRuLe learns a general class of temporal rules while maintaining full interpretability by applying symbolic reasoning. Experiments on three benchmark datasets demonstrate its superior performance. In the context where embedding and neural network-based representation learning methods dominate, we once again demonstrate the feasibility and superiority of pure symbolic logic reasoning. Our code, results, and datasets can be obtained through this repository: https://github.com/Xianglong-Bao/TempRuLe .