Is EML practical to use?

Short answer: it depends on what "use" means.

As a daily calculator, no

Nobody is going to replace sin(x) with a depth-12 EML tree on a pocket calculator. The representation is:

• Numerically less stable (deep exp/ln compositions amplify floating-point error)
• Slower to evaluate than a lookup table
• Harder for humans to read

As a compiler target, yes

If your output is machine code that will be executed, EML is attractive:

• One instruction type. Simpler CPU, simpler ASIC, simpler FPGA bitstream.
• Fewer edge cases in the compiler back-end.
• Potential for specialized analog hardware where a single physical element implements the entire operator.

As a theoretical tool, yes

The EML framing gives new answers to:

• "What is the minimal basis for this class of functions?"
• "How do we measure formula complexity?"
• "Can we compress a search space for formula discovery?"

Those are the settings where EML genuinely changes what you can do, not "use it instead of your calculator" but "use it when what you care about is structure."

The trade-off in one line

EML wins on uniformity and loses on efficiency. Any application that values the former over the latter is a candidate.