The pioneering work of Haller  on physically investigating bathymetry-controlled rip currents in the laboratory is a standard benchmark test for verifying numerical nearshore circulation models. In this paper, a numerical model based on higher-order Boussinesq equations was developed to reproduce the number of experiments involved in such an investigation, with emphasis on the effect of computational domain size on the numerical results. A set of Boussinesq equations with optimum linear properties and second-order full nonlinearity were solved using a higher-order finite difference scheme. Wave breaking, moving shoreline, bottom friction, and mixing were all treated empirically. The developed model was first run to simulate the rip current under full spatial and time-domain conditions. The computed mean quantities, including wave height, mean water level, and mean current, were compared with the experimental data and favorable agreements were found. The effects of computational domain size on the computation results were then investigated by conducting numerical experiments. The Willmott index was introduced to evaluate the agreements between the computed results and data. Inter-comparisons between the computation results and measurements demonstrated that the computational domain size significantly influenced the numerical results. Thus, running a Boussinesq wave model under full spatial and time-domain conditions is recommended to reproduce Haller's experiment.