It is not clear exactly what you are trying to do. Instead of latching the inputs, perhaps it is better to latch the encoded output. If you use something like the CD4532 it has an output signal GS that goes true whenever there is an active input. You could use it to clock a set of D type flip-flops to store the encoded output. With switches there is also contact bounce- if this is of consequence, you can debounce by using the GS to clock a monostable with a period longer than the bounce time, and use the trailing edge to clock your output register. If you have to latch the inputs then simply feed the input signal to the clock of a D type FF (with the D input tied high)- you then have to figure out how and when to clear it, but that is true for the suggestion I made above. As an aside, and perhaps you already know this, all the unused inputs to the 4532 (and all CMOS devices) should be tied low, and the used inputs should have a pull up resistor if they go directly to the switch. Protection of the inputs is a whole different ball of wax, if this is going into production. This design actually takes me back 42 years, so there may be more recent devices...
