Sampling methods for digital simulations of the optical diffraction based on scalar diffraction theory and fast Fourier transform are analyzed. There exist three sampling methods according to the sampling objects: the transfer function sampling, the point-spread function sampling and the weighting function sampling. Theoretical analysis reveals that these sampling methods hold true in different ranges of diffractive distances, which can be distinguished by a feature diffraction distance. Under definite sampling space and sampling number, the sampling methods hold true only when the diffraction distance is smaller than, larger than and equal to the feature distance, respectively. For demonstration of the correctness of the theoretical analysis, a digital simulation based on Matlab solfware and an experimental result are given and discussed. Finally, an adaptive sampling algorithm is proposed for reconstruction of a digital hologram, which could be suitable for any diffraction distances.