I've never used a switching regulator before. For my project I plan for an input voltage between 24V and 7.8V and regulating it to 5V for Arduino use. The maximum load would be about 200ma.
Is the LM2575 a good choice?
Would it need a heatsink?
The LM2575 needs 4 external components. Is there an alternative that requires fewer parts?
Would it need a heatsink?
Probably not.
However making a switch mode regulator is tricky and is very sensitive to layout and you need a PCB. I haven't known a engineer working for me that got the layout right fist time. (that includes me). I would advise you to get a ready built module if possible.
I would advise you to get a ready built module if possible.
Good to know. Thanks.
What would happen if the layout is not right?
Can I get something like this, copy the trace layout, and incorporate it into my own pcb?
Or is it trial and error to get a setup that works with each project?
You would be wasting a lot of time, money, and board space trying to beat what you can buy in the way of switching power regulators from Asian sellers on E-Bay. Here is just one example of hundreds of avalible types:
^That seems like a deal. I'll try it out. 
The RC modeling crowd use switching regulators a lot, to supply radios and servos from high voltage Li-ion packs. They're referred to as BEC or UBEC (ultimate battery eliminator circuit) and run about $4. Here's an example: http://www.hobbyking.com/hobbyking/store/uh_viewItem.asp?idProduct=4319
92% efficient, input voltage range 5.5v ... 23v, 3A continuous.
What would happen if the layout is not right?
The output would oscillate when certain currents are drawn. So it might look stable at some current draws but not at others.
If I follow the data sheets from the manufacturer, is this still a problem?
And is the only way to detect this to run it through a variable load with a scope on the output?
If I follow the data sheets from the manufacturer, is this still a problem?
Yes.
Do you think professional engineers don't follow the advice?
And is the only way to detect this to run it through a variable load with a scope on the output?
No you find the circuit you are trying to power is unreliable and malfunctions. This is not some sort of esthetic nicety.
I don't know. Not all switching regulators have a reputation for being THAT difficult to use. The tinier, the fussier (because they operate at higher frequencies.) The lm2575 doesn't seem too bad (only about 50kHz), and the MC34063 is very common in automotive converters, cheaply made with single sided PCBs and wildly varying components...
Pretty much the shorter the datasheet, the easier it is to use. If the datasheet includes a PCB layout for a four-layer board, that's probably a bad sign!
Grumpy_Mike:
If I follow the data sheets from the manufacturer, is this still a problem?
Do you think professional engineers don't follow the advice?
From personal experience as long-time lead in other fields, getting engineers to actually read best practices rather than rolling their own is a bit like herding cats.
And is the only way to detect this to run it through a variable load with a scope on the output?
No you find the circuit you are trying to power is unreliable and malfunctions. This is not some sort of esthetic nicety.
I got that. I was asking about tools to detect the problem before you get to a unit that can fail on the bench.
I also assume that a 42 kHz spread-spectrum device is easier to get stable than, say, a 1.6 MHz fixed frequency part? Electrolytes have OK performance below 100 kHz AFAICT, so adding capacitance is cheap.
Building a switching power supply from scratch is probably some what like trying to make tires for your car from scratch, probably best to just go to Tire Kingdom for the tires and ebay for a UBEC.
When you're building a "final product" with a PCB, you want to avoid external "boxes" and put everything on the PCB. There are circuits that you can buy that are "switching regulators." These are much more efficient than the "linear regulators" that you can buy -- LM7805, LM1117, etc.
However, these typically require a couple of capacitors, maybe a diode and a resistor. And, most importantly, the values and placement of these capacitors (and, in case of boost regulators, an inductor) is crucial. If you get it wrong, the whole thing will have a resonant mode and oscillate. I've been aware of this problem theoretically, but according to Grumpy_Mike, this problem happens even if you follow the (usually detailed) instructions available in the data sheets for the circuits.
Look at the data sheet for a typical part like the LM2574: Technical Documentation
Around page 8, it starts talking about oscillation and what to do about it. Apparently, this is more of a black art than what reading that description might lead you to believe. (And this description is thorough, going into detail about capacitor ESR varying by type and temperature, parasitic inductance/capacitance of PCB copper, etc)