How to best match $( ... ) in a string

How to best match $( ... ) in a string

[Sebastian Gniazdowski]

Hello,
the problem is possible quoting, e.g. $( echo \) ). Has anyone a
pattern that would handle some sort of quoting?

-- 
Best regards,
Sebastian Gniazdowski

Re: How to best match $( ... ) in a string

[Joey Pabalinas]

[-- Attachment #1: Type: text/plain, Size: 349 bytes --]

On Fri, Jun 01, 2018 at 10:44:58AM +0200, Sebastian Gniazdowski wrote:
> Hello,
> the problem is possible quoting, e.g. $( echo \) ). Has anyone a
> pattern that would handle some sort of quoting?

You mean like a regex/glob to match the entire "$(...)" construct with those
possible weird \) and such inside?

-- 
Cheers,
Joey Pabalinas

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: How to best match $( ... ) in a string

[Sebastian Gniazdowski]

On 1 June 2018 at 10:49, Joey Pabalinas <joeypabalinas@gmail.com> wrote:
> On Fri, Jun 01, 2018 at 10:44:58AM +0200, Sebastian Gniazdowski wrote:
>> Hello,
>> the problem is possible quoting, e.g. $( echo \) ). Has anyone a
>> pattern that would handle some sort of quoting?
>
> You mean like a regex/glob to match the entire "$(...)" construct with those
> possible weird \) and such inside?

Yes, to not stop matching on \), etc.

-- 
Best regards,
Sebastian Gniazdowski

Re: How to best match $( ... ) in a string

[Peter Stephenson]

On Fri, 1 Jun 2018 10:44:58 +0200
Sebastian Gniazdowski <sgniazdowski@gmail.com> wrote:
> the problem is possible quoting, e.g. $( echo \) ). Has anyone a
> pattern that would handle some sort of quoting?

In general, this isn't possible.  Apart from nested unquoted
$(...), which is also valid, the real killer is

echo $(case foo in
  foo) echo This does work
  ;;
  esac)

This required us to rewrite our internal parsing completely and we're
still shaking the odd bug out of the result.

If you ignore that case, it's possible character by character with
a bit of extra state for quotes, nested parentheses, etc. (as that's
what zsh did for two decades) but you're going to need some incredibly
sophisticated regular expression involving recursion to replace that.

pws

Re: How to best match $( ... ) in a string

[Sebastian Gniazdowski]

On 1 June 2018 at 11:50, Peter Stephenson <p.stephenson@samsung.com> wrote:
> On Fri, 1 Jun 2018 10:44:58 +0200
> Sebastian Gniazdowski <sgniazdowski@gmail.com> wrote:
>> the problem is possible quoting, e.g. $( echo \) ). Has anyone a
>> pattern that would handle some sort of quoting?
> (...)
> If you ignore that case, it's possible character by character with
> a bit of extra state for quotes, nested parentheses, etc. (as that's
> what zsh did for two decades) but you're going to need some incredibly
> sophisticated regular expression involving recursion to replace that.

Yes I should state that I'm aware it is not possible. I look for a
best possible solution. Fast-syntax-highlighting now supports
colorizing of command substitution $( ... ), it is called recursively
on it. The point is that 90% of uses will be colored good with my
current pattern:

inputs=( ${(0)${(S)__buf[1,110]//(#b)*\$\((?#)([^\\\"]\)|(#e))/${mbegin[1]};${match[1]}${match[2]%\)}${__nul}}%$__nul*}
)

The meaningful bit in above is: \$\((?#)([^\\\"]\)|(#e)). Match $(,
then look for unquoted (not \), not ")) closing ) or for end of whole
string. I utilize (S) non-greedy matching, and the (0)/$nul trick to
handle multiple substitutions in one command line.The effect is
already quite nice: http://psprint.blinkenshell.org/cmdsubst.png

So the point is, the glob can be hacky and imperfect, if it holds some
valuable solutions then syntax-highlighting user is at better
position. Any uplifts to my current pattern are welcomed.

-- 
Best regards,
Sebastian Gniazdowski