I want to charge capacitor with a switching regulator

[LM21]

I am thinking about old and familiar problem where we have two capacitors with different voltages. If those capacitors are then connected in parallell, sparks fly and some energy will be lost. But a switching regulator is able to convert voltages and currents with little losses.
I am not really thinking of building any of such systems, but what kind of regulator would be best. Some simple constant current fly back perhaps.

[Circlotron]

A diode and inductor in series with one of the connections between the two capacitors will transfer energy one way with relatively low loss.

[ajb]

It depends a lot on how different the voltages are, and what rate of charge transfer (current) you require.  In that sense it's no different from selecting a switching regulator for any other kind of application.  You will of course want current control, but that can be added to any converter topology.

Control of the converter may be a bit different than with typical converter applications depending on what you're actually trying to do.  If you have some big monster cap at a relatively high voltage and you're trying to use it to charge a smaller cap up to a defined voltage below that, then you just set the output voltage and desired current limit as normal and you're done.  If you're trying to equalize the voltage on two caps or something, then you would have to use the difference between the two voltages to drive the error amp rather than the difference from a fixed setpoint. 

[Doctorandus_P]

Some food for thought:

The energy stored in a capacitor rises with the square of the voltage over the capacitor.
This means, that if one battery starts with a higher voltage then the other, and you connect them in parallel, the end result is that the sum of the energies stored in both capacitors is less after the connection. It is a closed system, but the energy has to go somewhere. With a resistor the difference is dissipated as heat, by just connecting a wire you create the sparks. By just adding an inductor,  the energy is stored in the inductor, which then starts charging the other capacitor. (With a dampened oscillation because of losses), or it stops by adding that diode.

For any practical use, the size of the capacitors will have a big influence on the circuit.
If it's just for your curiousity, then combining it with learning spice simulations could be a good idea.