FGL is a great example of how to make a "nice" high-level graph interface suited for functional programming. I'm a big fan. But it's orders of magnitude too slow and memory-inefficient for performance-sensitive graph computations—if you have even moderately sized graphs and graph algorithms are a bottleneck, you'll need to use something else, and probably something domain-specific. Given the way the interface works, I don't think you could have a high-performance version that would scale to large graphs.

In my experience this leaves FGL in an awkward spot: on the one hand, it isn't sufficient for heavy-duty graph processing; on the other, if you don't need fancy high-performance graph algorithms, chance are that encoding your problem as a graph is going to be more awkward than defining some domain-specific types for what you're doing. Graphs are such a general structure that they're usually the wrong level of abstraction for higher-level domain-specific logic.

Of course, sometimes you're writing graph code specifically and you need a nice way to express your graph algorithms without worrying about performance. In that case, FGL is great. I wrote a tutorial about using it to [generate mazes][1] and it helped me express the algorithms better than I would have been able to do without it. But that still leaves it as too narrow for something to be "the" graph representation in a language's standard library.

[1]: https://jelv.is/blog/Generating-Mazes-with-Inductive-Graphs/