Issues when using empty aliases with certain builtins

[Crestwave]

$ alias a=
$ a declare var=0 # Same result with local, but export works fine. This also fails when aliased to
   declare var=0
                 ^
[ expansion of alias 'a' ]:1: Invalid variable name ''
$ a (( var = 0 ))
  a (( var = 0 ))
    ^~
'/home/user/.config/oil/oshrc':13: Unexpected word while parsing command
line
$ a var=0
$ printf '%s\n' "$a"

$ a f() { :; }
  a f() { :; }
     ^
'/home/user/.config/oil/oshrc':16: Unexpected word while parsing command
line
$ a case a in a) : ;; esac
   case a in a
               ^
[ expansion of alias 'a' ]:1: Expected word type <Id_t Right_CasePat 94>, got <Id_t Word_Compound 199>

[andychu]

The inconsistency between export and declare/local will be fixed with #236 (dynamic assignments).

The other bugs are I guess because OSH has a different parsing/evaluation model of aliases. They seem somewhat similar to cases 33-34 here:

http://www.oilshell.org/release/0.6.pre18/test/spec.wwz/alias.html

where I have alias LEFT='{' and alias LEFT='(', which works in most shells but not OSH.

That is basically because of the design philosophy here:

http://www.oilshell.org/blog/2019/02/07.html

Are you using this for something? If not, I'm inclined to leave this as is, unless there it moves something here forward:

https://github.com/oilshell/oil/wiki/Shell-Programs-That-Run-Under-OSH

All the shell parsers I looked at are littered with special cases regarding aliases. In contrast, OSH uses the parser "from the outside", which is more understandable. It has a pretty clear evaluation model, whereas in most shells you have to enumerate all the special cases to know what it does.

---

Another case where this is true is a slight incompatibility with backticks, which somebody on help-bash@ is trying (and failing) to get an explanation of right now... OSH has a pretty compatible and explainable rule that runs virtualenv, but it disagrees on a corner case where other shells agree.

That is case #25 here:

http://www.oilshell.org/release/0.6.pre18/test/spec.wwz/command-sub.html

However shells stop agreeing on case #26 :)

[andychu]

I just committed a spec test that records this incompatibility. It's unfortunate that all shells do agree on this and OSH doesn't, but like I said, I'm going to put it low on the pile unless real world use cases show up.

I also started documenting some more of the known incompatibilities in doc/osh-manual.md.

[Crestwave]

Are you using this for something? If not, I'm inclined to leave this as is, unless there it moves something here forward:

https://github.com/oilshell/oil/wiki/Shell-Programs-That-Run-Under-OSH

Nope; I found it while playing with https://github.com/rdebath/Brainfuck/blob/master/extras/bf2bash.sh, which is honestly quite horribly written/hacky in my opinion.

[andychu]

OK since I documented this incompatibility in the spec tests, I'm closing this... it may or may not come up again in the future :-/