eval evil?

[coderofsalvation]

So from what i've understood is that the compiler is using eval, to stay flexible concerning bash vs sh output.
So powscript generally produces this type of output:

set?() {
  {
    local s;
    [ $# -ge 1 ] && s="${1}"
  }
  eval "{ return \"\$(( \${$s+-1}+1 ))\"; }"
}

@fcard: Might there be a way to get rid of the eval you think?

Could we get away with this kind of output in most cases? :

set?() {
  {
    local s;
    [ $# -ge 1 ] && s="${1}"
  }
  return "$(( ${$s+-1}+1 ))"
}

That way powscript can also be attractive to people who dismiss the usage of eval (because of security reasons).
Looking at all my powscript-output, I haven't really seen a case where wrapping shellscript in an eval-call was necessary.
WDYT?

[fcard]

Unfortunately we can't do${$s} (gives a "bad substitution" error), which is exactly why we need eval when dealing with names and arrays, otherwise I'd agree with you that we should avoid eval.

Since we have expand for these cases, we can make it more secure, and recommend it over directly using eval whenever possible.

[coderofsalvation]

i see.
Maybe a silly question, but why would we want to do ${$s} if we can do ${s}?
And if the powscript transpiler can output any text, shouldn't it just be able to output something like this:

set?(){
  {
    local s;
    [ $# -ge 1 ] && s="${1}"
  }
  return $?           # will return resultcode of last executed command 
}

or ideally:

set?(){
  [ $# -ge 1 ] && return 0 || return 1
}

I'm hoping that we've just discovered a lowhanging fruit here :)

[fcard]

Your version would print "true" in the below code:

unset s

if set? s
  echo "true"

We should probably add tests to the std lib soon, so we can experiment better :P

The reason we need an equivalent to ${$s} is because we're dealing with names (or references) instead of strings.

[coderofsalvation]

Afaik in shellscript reference-notation (without $) is only used in (undoing) declarations & assignments, not in if-statements etc.
As confusing as that already is, maybe we should stick to that? (with function arguments as exception?)
Or are there other cases which require us to stick with eval?

unset s

if set? $s
   echo "true" # will not print

Let me know what u think.

[fcard]

But that will print false below.

s=""

if set? $s
  echo true
else
  echo false

There is no meaning in set? without references.
There are many uses in references, they are normally just too annoying in shell scripting to use :P

[fcard]

Actually, that will print true, I forgot about auto quoting. However, that means that

unset s

if set? $s
   echo "true"

Will print true with your version.

[coderofsalvation]

hm, im on the bus right now, so i'll have to digest these truths about shellscript :) On Sep 4, 2018 5:30 PM, "fcard" <notifications@github.com> wrote: But that will print false below. s="" if set? $s echo true else echo false There is no meaning in set? without references. There are many uses in references, they are normally just too annoying in shell script to use :P — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#59 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAK_ZLzl3ESjgafrCvoF0EbdE7q6MYkTks5uXpysgaJpZM4WXDHU> .