Computer programming is regarded as an important skill for the future. However, many K-12 students face challenges and difficulties in learning traditional text-based programming. Block-based visual programming (BVP) can reduce the difficulty of learning programming and is seen as a potential programming education tool. Nevertheless, the effects of BVP on K-12 student learning outcomes remain unclear, and there is no evidence of meta-analysis. This is the "first" meta-analysis that aims to examine the effects of BVP on K-12 students' learning based on "42" effect sizes from 29 independent experimental studies (2000-2023). The result suggests that BVP has an upper-medium effect on learning (SMD = 0.769, 95% CI = [0.554, 0.984], p < 0.001), particularly in the cognitive outcome (SMD = 0.698, p < 0.001). Moreover, the BVP tool, study design, and region have moderating effects. The BVP's effects are larger when: (1) with 61~100 students, (2) for junior high school, (3) with the learning duration of >1 month, (4) targeting math, (5) among novice students, (6) using Scratch, ScratchJr, Alice, and App Inventor, (7) employing project-based learning and project-based + game-based learning, (8) base on quasi-experimental design, (9) in Africa and Europe, (10) with latter publication years. Finally, discussions, implications, limitations, and future directions are put forward.