Expansions

[amaanq]

@ObserverOfTime

[amaanq]

${!x//c/x} is tough

[ObserverOfTime]

${!x//c/x} is tough

I don't think that's valid...

[amaanq]

it's in the examples corpus (bash repo)

Fixing some things breaks others, and vice versa 😁

[ObserverOfTime]

Fixing some things breaks others, and vice versa 😁

The circle of life.

[ObserverOfTime]

Maybe you could make expansions start with optional('!') with the exception of ${#foo}, ${!foo*} & ${!foo@}.

[amaanq]

${!##} is doing my head in

[amaanq]

after a few painful hours I think I have a complete solution, would appreciate you trying it out @ObserverOfTime

[ObserverOfTime]

Expansions appear to be fine now but there are a few arithmetic issues left.

• Operators in C-style binary expressions are not operator fields.
• The unary expression operators "-" "+" "~" cannot be captured.
  • Maybe unary and postfix operators should also be fields.
• test_operator seems to have higher precedence over negative variables.

echo $((- f))
echo $((-f))

Parse tree

0:0  - 2:0     program
0:0  - 0:13      command
0:0  - 0:4         name: command_name
0:0  - 0:4           word `echo`
0:5  - 0:13        argument: arithmetic_expansion
0:5  - 0:8           "$(("
0:8  - 0:11          unary_expression
0:10 - 0:11            variable_name `f`
0:11 - 0:13          "))"
0:13 - 1:0       "\n"
1:0  - 1:12      command
1:0  - 1:4         name: command_name
1:0  - 1:4           word `echo`
1:5  - 1:12        argument: arithmetic_expansion
1:5  - 1:8           "$(("
1:8  - 1:10          test_operator `-f`
1:10 - 1:12          "))"
1:12 - 2:0       "\n"

[amaanq]

are test operators allowed in arithmetic expansions? (I'm guessing no)

[ObserverOfTime]

Not as far as I can tell.

echo $((0 -eq 0))
# bash: 0 -eq 0: syntax error in expression (error token is "0")

I assume it parses 0 - eq as a binary expression so the error is on the leftover 0.

[amaanq]

all operators are fielded & should be exposed (incl. unary ops)

[amaanq]

Hmm interesting, token(prec(choice(...))) prevents querying the operators, but not choice(...token(prec())) (no need for aliasing)

@ahlinc

[ObserverOfTime]

I still can't seem to capture unary operators.

[amaanq]

got a sample query code? this works for me on this branch:

(unary_expression "+" @op)

[ObserverOfTime]

Does it work with x=$((+y)) ?

[amaanq]

damn, the query compiles but doesn't seem to execute properly, I'll bring back the aliases

[ObserverOfTime]

I will once again advocate for a prec.lexical function that is intuitive and works as expected.

[amaanq]

Nope, this was my fault. I forgot I had _arithmetic_unary_expression too

I would say the current mechanism works as expected (after the bug fix yesterday), but it's definitely not intuitive.

[amaanq]

try now @ObserverOfTime

[ObserverOfTime]

No field. :^)
But it does work.

[amaanq]

would you consider ternaries as candidates to be fielded as operator?

[ObserverOfTime]

would you consider ternaries as candidates to be fielded as operator?

It's fine either way.

[amaanq]

Alright, I don't really care to - so this should be good as is hopefully

[ObserverOfTime]

Yeah, finally bash is usable.

[amaanq]

Yeah it was pretty "eh" the few times I used it in the past :^) (and arithmetic expansions broke everything)

But writing the parser this language definitely sucks

[ObserverOfTime]

But writing the parser this language definitely sucks

What do you mean this language sucks? 💢

[amaanq]

pfft...it won't suck all that much compared to zsh 😬 🤢

[ObserverOfTime]

Or powershell.

[amaanq]

as if I'd want to touch Windows 😁

[verhovsky]

@amaanq powershell actually runs on Linux and macOS, it's even open source under the MIT license.

https://formulae.brew.sh/cask/powershell

https://github.com/PowerShell/PowerShell

[amaanq]

wow, I never knew, that's cool but I wonder if people actually use it on a non-Windows OS (except for cross-platform testing)