const folding happens too early in semantic phase so it's invisible to term rewriting macros and astToStr

[timotheecour]

import typetraits, strformat

block:
  template foo(cond: typed): string =
    astToStr(cond)

  block:
    const a = 1
    # a + 10 is const folded before astToStr sees it
    doAssert foo(a + 10) == "11"

  block:
    let a = 1
    doAssert foo(a + 2) == "a + 2"

block:
  template optMul{`*`(a, 2)}(a: int): int =
    1234

  # 3 * 2 is const folded before term rewriting sees it
  doAssert 3 * 2 == 6

  # term rewriting is activated
  let x = 3
  doAssert x * 2 == 1234

note:

for the TRM case, this is the root cause of what was observed by @kaushalmodi in https://scripter.co/notes/nim/#term-rewriting-macros "echo "Calculating 3 * 2:" # Why isn't term rewriting activated here?"

note:

in particular it affects doAssert: doAssert a == 2 will just show false failed as the expression obtained from astToStr instead of the more informative a == 2 failed

[Araq]

For the TR macros that's not a bug, TR macros are not about random rewritings, they are about optmizations and what has been reduced by the compiler doesn't require further optimization.

[timotheecour]

after #9332 , doAssert a == 2 will now render as a == 2 instead of false

[timotheecour]

TR macros are not about random rewritings

well I was hoping it could be useful for "random rewritings" as you said (with caveat that any tool can be misused); a number of use cases come to mind, notably for debugging, eg warning when a certain pattern is found during compilation