I did some tradeoff analysis on filament drivers for a project underway, and I concluded a dropping resistor is better than a current-regulator for filaments, at least in my case.
To cancel-out the voltage-gradient along a filament, AC rather than DC is commonly used. That makes a current-regulator more complex even if you just use a square-wave.
The basic problem with filaments is that their cold-resistance is several times lower than their operating resistance, and without any current-limiting they will undergo a surge-current at turn-on that is many times greater than their operating current. It's basically why incandescent bulbs almost always fail when switched-on.
If you drive a 1V filament from a constant 1V supply, the filament current is entirely determined by the filament's resistance. When it's energized, it rapidly heats up, which increases the resistance which lowers the current, hence lowers the temperature. The current decreases until equilibrium is reached.
The filament in the tube I'm working with has a cold-resistance of 2.7 ohms, and measured 7 ohms at it's operating current of 200mA. That's almost a 3:1 range.
If you drove this from a 5V supply, you would need about 18ohms of series resistance for 200mA at 5V. When you first turn it on, the surge current will be 5/(2.7+18) = 241mA. On the other hand, if you just used the 1V supply, the surge-current is 370mA. Clearly the higher-voltage supply with the dropping resistor will put far less stress on the tube. The tradeoff is wasted energy (280mW for the tube, vs 720mW for the dropping resistor).
I did some web research on this, and found a good paper about extending the life of radio-transmitter tubes. They basically do the same thing: Add series-resistance to limit the surge current, and the tube life was dramatically increased. I think the same reasoning can apply to VFDs, magic eyes, & NIMOs.
The one thing I could not draw a conclusion on was how to determine when to leave filaments running, vs turning them off. Even with current-limiting a filament will undergo mechanical degradation from expansion and contraction when power-cycled. But a filament will also undergo some degradation just from being hot. I wont have enough time and tubes to determine this thru experiments. I've seen some info about thoriated filaments (ie, containing Thorium) degrading if they are kept idle at currents significantly below their normal operating current. I'm inclined to keep my clock filaments on while the clock is plugged-in, energize them via a PIR sensor, and unplug the clock when I put it away for a few months.