I know in computer graphic, we can set a world window in world coordinate system, and then mapping it to viewport which is the display window. Things that are not in the world window should not be displayed and we can do this through Clipping (like in OpenGL). I have a question that do we have alternative methods?
"Go not to the CG Elves for counsel for they will say both no and yes".
Yes and No. As noted in another post, these issues don't exist with ray tracing but I'll assume you are interested in standard rasterisation.
It's possible to dispose of near clipping by using homogeneous rendering but that can introduce additional cost. (e.g more expensive triangle set up, more difficulty in identifying which screen regions are covered in tile-based renderers)
You could eliminate XY clipping by using float representations for the X and Y coordinates in the rasteriser but that can lead to catastrophic cancellation when computing triangle determinants which, AFAICS, is a necessary step. In my experience, hardware rasterisers* usually work in fixed-point so that everything is 'exact'.
As a compromise to XY clipping, guard-band clipping was introduced so that clipping only needs to be done on triangles that extend a 'long'** way off screen. Culling, however, is still be done with the standard frustum planes to eliminate those that are trivially off-screen.
*Dreamcast being a notable exception.
**for certain definitions of 'long', e.g. a small integer multiple of the screen width/height.