Getting to that "a ha!" moment when I figured out that this would help (a lot) was very hard. Forcing my code to behave that way was definitely non-intuitive and challenging. Having merely gotten it working and proving my theory (it works) that it's a huge efficiency gain, I now get another crack at the next version.
I don't think it's necessary for a compiler to do the work for you, or to have it built in as a fundamental feature of a language (that certainly helps). But I do agree that without a library or (dare I say) framework, it is a bit challenging to code from scratch.
Probably the hardest part (for me), is that it breaks so many fundamental notions of programming.
One obvious feature that I'm making heavy use of is that I can get data out of the state machine on every call (that's almost the whole point for me). I can use that data and determine if I want to call it any more or just kill it (or maybe put it aside to get called later).
Perhaps more interesting, I have the opportunity to pass data into the state machine on every call down to it (not just on the initial call). That seems powerful. I already have an idea on how to utilize that.
I do agree, that at some point this technique would be more efficient if implemented in assembly than in a high-level language. My approach is probably similar to implementing a binary search in a high-level language and comparing it so a linear search in C. At this point I'll take the flexibility of the high-level language. Almost certainly a lower level implementation of the same pattern would kick butt.
Getting to that "a ha!" moment when I figured out that this would help (a lot) was very hard. Forcing my code to behave that way was definitely non-intuitive and challenging. Having merely gotten it working and proving my theory (it works) that it's a huge efficiency gain, I now get another crack at the next version.
I don't think it's necessary for a compiler to do the work for you, or to have it built in as a fundamental feature of a language (that certainly helps). But I do agree that without a library or (dare I say) framework, it is a bit challenging to code from scratch.
Probably the hardest part (for me), is that it breaks so many fundamental notions of programming.
One obvious feature that I'm making heavy use of is that I can get data out of the state machine on every call (that's almost the whole point for me). I can use that data and determine if I want to call it any more or just kill it (or maybe put it aside to get called later).
Perhaps more interesting, I have the opportunity to pass data into the state machine on every call down to it (not just on the initial call). That seems powerful. I already have an idea on how to utilize that.
I do agree, that at some point this technique would be more efficient if implemented in assembly than in a high-level language. My approach is probably similar to implementing a binary search in a high-level language and comparing it so a linear search in C. At this point I'll take the flexibility of the high-level language. Almost certainly a lower level implementation of the same pattern would kick butt.