I have had a soft spot for the Forth programming language for at least 10 years now. One of my favorite personal projects was writing an implementation of Forth in Z80 assembly for my TI-83 calculator, complete with an interpreter, loading files, and drawing on the screen. I've done Arduino projects with Forth, Advent of Code problems, and other projects, but I wouldn't say that I've really used it "in anger" before now.
I have recently been thinking about the iteration speed of my embedded systems development, and I've been able to get some good returns using TCL on the host side and Forth on the target side. I repackaged pforth so I could statically link it, complete with an initial Forth system as text in the binary, and then link in my embedded software as another static library. Then I can use functions from my software at the interpreter, as long as I do the work to expose them. I also redid the way that C functions are exposed to pforth a bit to make this easier- there is nothing wrong with the way its done normally, but I wanted pforth to be part of my program, not the other way around.
What all of this made me think about was that for the first time I was using Forth to solve an engineering problem. Something messy and real-world where I just needed results in the lab. I was curious if I would find Forth useful in this context, or if I would find it a problem.
I'm happy to say that for the problems I actually faced, I found Forth to be perfectly useful and fast. I could add functions to test my hardware, try things out again and again until I got them right (due to the fast iteration speed), and even do the little tasks like print out some memory or compare values within collected data with no problem.
Part of this is being fairly familiar with Forth, and part of it is that Forth is good at the kinds of things I wanted to do- talk to hardware, inspect memory, string together functions in different ways quickly. The other part is just Forth's style of being factored into small pieces (one line usually), and combined upwards into useful functionality. I could break the small problems I faced down quickly, test things out, and build up what I needed. I can't say for certain what I would have found for larger problems, but I didn't actually face those problems, so I won't comment.
The moral of the story here is that I'm very pleased with the results, and I plan on continuing to keep this Forth build of my software for iterating on logic, for talking to hardware, and to ensure that my software is factored well enough to use it as a library if necessary.
While doing all of this I did a survey of language implementations that where available, including TCL, JimTCL, Lua, pforth, libforth, zforth, picol, and more. I hope I can write up some thoughts on all of these some time, as they are all points within this design space with some interesting tradeoffs in completeness, complexity, portability, standard library size, memory use, etc.