I don't know what it says precisely in sedra smith but you can't do anything about input offset current. However, if you mean input bias current then that is eradicated by making the dc resistance seen by both inputs the same value. So, if there is a 100k feedback and a 100k input resistor to the inverting terminal then, ideally, you should have 50k seen by the non-inverting terminal to 0V. Thus bias currents produce an error voltage at each input terminal that is identical and thus cancel out.
Input offset current is the difference between the bias currents on each input terminal and this is not so easily got rid of.
Of course if the bias current is only 1nA and you are using 1k resistors (i.e. quite small values) then the worst case dc voltage error produced (if you don't have a resistor in the inverting input) is 1 uV - probably far less than the input offset voltage of the op-amp and therefore not worth bothering about.
See this (MT-038: Op Amp Input Bias Current - Analog Devices) for further reading on the subject.