20230110 modern lambda

[prozacchiwawa]

Create a new form

(lambda ((& captures ...) arguments) (code))

Which captures the items in captures from the environment and creates runnable code taking (arguments) as arguments.

In this code:

(mod (add-number L)

  (include *standard-cl-21*)

  (defun map (F L)
    (if L
      (c (a F (list (f L))) (map F (r L)))
      ()
      )
    )

  (map
    (lambda ((& add-number) number) (+ add-number number))
    L
    )
  )

A runnable (with the 'a' operator) lambda is emitted by the lambda form, which captures add-number from the environment. When run by 'map' with another argument (here filling 'number'), it can use both the captured arguments and the arguments given when it is run.

Run with (5 (1 2 3 4)) gives (6 7 8 9)

You can introduce bindings for capture using let:

(mod (add-number L)

  (include *standard-cl-21*)

  (defun map (F L)
    (if L
      (c (a F (list (f L))) (map F (r L)))
      ()
      )
    )

  (map
    (let ((A (* add-number 2)))
      (lambda ((& A) number) (+ A number))
      )
    L
    )
  )

Run with (5 (1 2 3 4)) gives (11 12 13 14)

[prozacchiwawa]

based on use, a bug fix might be needed. looking at a test case.

[prozacchiwawa]

not a bug, added some test cases.

[arvidn]

the 1s and 2s here refer to the quote- and apply- opcodes, right?
if so, it might be easier to understand by using #q and #a instead

[prozacchiwawa]

Admittedly, I find the mixing more confusing, but will change that.

[arvidn]

perhaps I misunderstand what this means, but I would hope that the compiler would catch my mistake if I misspell a variable name in the lambda body. having an explicit capture seems like an opportunity for us to be stricter about this

[prozacchiwawa]

As we discussed earlier, that wasn't a change I felt i could make at this point. Since it's been very contentious I will speed through the changes needed to be fully strict.

[prozacchiwawa]

This upcoming pr allows us to get to strictness in macro output since it allows precise control of the form the output takes, as well as other improvements.

https://github.com/Chia-Network/clvm_tools_rs/pull/134

It's a precondition for being fully strict.

[prozacchiwawa]

https://github.com/Chia-Network/clvm_tools_rs/pull/176/files

[arvidn]

I appreciate these tests. I think it would make sense to have negative tests as well. it's important compilation fails when it should

[prozacchiwawa]

It's been a bit hard because not much actually is meant to fail in chialisp and I often discover things people want that are questionable or should fail to me, but I will set those bounds as well as I can.

[prozacchiwawa]

now that other required features have been merged, the test suite for chia-gaming is now in here as a fairly comprehensive regression test.