hoisting + memfiles = wierd

[scriptum]

I'm trying to understand how to use term-rewriting macros for performance optimization purposes. One interesting thing is hoisting for regular expressions.

Simple example:

import memfiles, re
proc main()=
  var c = 0
  var f = memfiles.open("in.txt")
  for line in f.lines():
    if line.contains(re"agggtaaa|tttaccct"):
      c += 1
  echo(c)
  f.close()
main()

In this example Regex object created for every line.

Hoisting could improve performance (idea taken from docs):

import memfiles, re

template optRe*{re(pattern, flags)}(pattern: string{lit}, flags: typed): Regex =
  var glRe {.global, gensym.} = re(pattern, flags)
  glRe

proc main()=
  var c = 0
  var f = memfiles.open("in.txt")
  for line in f.lines():
    if line.contains(re"agggtaaa|tttaccct"):
      c += 1
  echo(c)
  f.close()
main()

Now magic. Take a look at generated C code:

NIM_EXTERNC N_NOINLINE(void, hoisting_weirdInit000)(void) {
nimRegisterGlobalMarker(T1021622097_3);
nimRegisterGlobalMarker(T1021622097_5);
    asgnRefNoCycle((void**) (&glre_154025_1021622097), re_152079_2126175263(((NimStringDesc*) &T1021622097_4), 24));
    asgnRefNoCycle((void**) (&glre_154045_1021622097), re_152079_2126175263(((NimStringDesc*) &T1021622097_4), 24));
    main_154006_1021622097();
}

Global variable initialized twice!

But replace memfiles.open to system.open:

NIM_EXTERNC N_NOINLINE(void, hoisting_weirdInit000)(void) {
nimRegisterGlobalMarker(T1021622097_3);
    asgnRefNoCycle((void**) (&glre_154024_1021622097), re_152079_2126175263(((NimStringDesc*) &T1021622097_4), 24));
    main_154006_1021622097();
}

This looks good.

I guess this is because memfiles iterator is more complicated and probable this is not a bug. Is there any workaround?

[Araq]

I don't know of a workaround, but the codegen should not emit duplicate assignments.

[scriptum]

Same thing with lexim project. Seems that memfiles iterator due to two yields not compatible with macros. More precisely, iterator + several yields + macros = double substitution. Probably, iterators need more attention.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think it is still a valid issue, write a comment below; otherwise it will be closed. Thank you for your contributions.