Discovered pattern that ideally parses any quoting. Brackets highlighting in FSH.

Discovered pattern that ideally parses any quoting. Brackets highlighting in FSH.

[Sebastian Gniazdowski]

Hello,
not only (z) can handle quoting, turns out regular pattern with while
loop can do it too:

    while [[ $_mybuf = (#b)[^"{}()[]\\\"'"]#((["({[]})\"'"])|[\\](*))(*) ]]; do
        [[ -n "${match[3]}" ]] && {
            __idx+=${mbegin[1]}+1
            _mybuf="${match[3]:1}" # also skip 1 quoted char
        } || {

the main point is [\\](*). Parentheses are placed in ${match[3]} and
allow to restart processing skipping e.g. a quoted backslash.

The rest is matching braces – it's an example from
Fast-Syntax-Highlighting, it now has ideal brackets highlighting – AND
ALSO ", '. Simple handling of those two quotings allows to switch mode
"in quoting" "outside quoting". That's all. The effect:

https://raw.githubusercontent.com/zdharma/fast-syntax-highlighting/master/images/brackets.gif

-- 
Sebastian Gniazdowski
News: https://twitter.com/ZdharmaI
IRC: https://kiwiirc.com/client/chat.freenode.net:+6697/#zplugin

Re: Discovered pattern that ideally parses any quoting. Brackets highlighting in FSH.

[Sebastian Gniazdowski]

The brackets in the pattern obfuscate a little the idea, so here's
minimal code that matches ABC instead, together with full ", '
handling.

_mybuf="A\AB'A'C";
while  [[ $_mybuf = (#b)[^ABC\"\'\\]#(([ABC\"\'])|[\\](*))(*) ]]; do
    [[ -n "${match[3]}" ]] && {
        __idx+=${mbegin[1]}+1
        _mybuf="${match[3]:1}" # also skip 1 quoted char
    } || {
        [[ -z "$__quoting" ]] && {
            if [[ "${match[1]}" = [ABC] ]]; then
                echo "Got ${match[1]}"
            fi
        }
        [[ "${match[1]}" = \" && "$__quoting" != \' ]] && { [[
"$__quoting" = '"' ]] && __quoting="" || __quoting='"'; }
        [[ "${match[1]}" = \' && "$__quoting" != \" ]] && { [[
"$__quoting" = "'" ]] && __quoting="" || __quoting="'"; }
        _mybuf="${match[4]}" # The last (*) in the pattern, i.e.
remaining string, to become next input to while
   }
done

output:
Got A
Got B
Got C

On 30 July 2018 at 09:09, Sebastian Gniazdowski <sgniazdowski@gmail.com> wrote:
> Hello,
> not only (z) can handle quoting, turns out regular pattern with while
> loop can do it too:
>
>     while [[ $_mybuf = (#b)[^"{}()[]\\\"'"]#((["({[]})\"'"])|[\\](*))(*) ]]; do
>         [[ -n "${match[3]}" ]] && {
>             __idx+=${mbegin[1]}+1
>             _mybuf="${match[3]:1}" # also skip 1 quoted char
>         } || {
>
> the main point is [\\](*). Parentheses are placed in ${match[3]} and
> allow to restart processing skipping e.g. a quoted backslash.
>
> The rest is matching braces – it's an example from
> Fast-Syntax-Highlighting, it now has ideal brackets highlighting – AND
> ALSO ", '. Simple handling of those two quotings allows to switch mode
> "in quoting" "outside quoting". That's all. The effect:
>
> https://raw.githubusercontent.com/zdharma/fast-syntax-highlighting/master/images/brackets.gif
>
> --
> Sebastian Gniazdowski
> News: https://twitter.com/ZdharmaI
> IRC: https://kiwiirc.com/client/chat.freenode.net:+6697/#zplugin



-- 
Sebastian Gniazdowski
News: https://twitter.com/ZdharmaI
IRC: https://kiwiirc.com/client/chat.freenode.net:+6697/#zplugin

Re: Discovered pattern that ideally parses any quoting. Brackets highlighting in FSH.

[Sebastian Gniazdowski]

There was __idx assign missing, and I fine-tuned readibility, so final version:

_mybuf="A\AB'A'C" #input
while  [[ $_mybuf = (#b)[^ABC\"\'\\]#(([ABC\"\'])|[\\](*))(*) ]]; do
    [[ -n "${match[3]}" ]] && {
        __idx+=${mbegin[1]}+1
        _mybuf="${match[3]:1}" # also skip 1 quoted char
    } || {
        __idx+=${mbegin[1]}
        [[ -z "$__quoting" ]] && {
            if [[ "${match[1]}" = [ABC] ]]; then
                echo "Got ${match[1]}"
            fi
        }

        # Toggle quoting
        [[ "${match[1]}" = \" && "$__quoting" != \' ]] && {
                [[ "$__quoting" = \" ]] && __quoting= || __quoting=\";
        }
        [[ "${match[1]}" = \' && "$__quoting" != \" ]] && {
                [[ "$__quoting" = \' ]] && __quoting= || __quoting=\';
        }

        # The last (*) in the pattern, i.e. remaining
        # string, to become next input to while
        _mybuf="${match[4]}"
   }
done

Re: Discovered pattern that ideally parses any quoting. Brackets highlighting in FSH.

[Bart Schaefer]

On Mon, Jul 30, 2018 at 12:09 AM, Sebastian Gniazdowski
<sgniazdowski@gmail.com> wrote:
>
> not only (z) can handle quoting, turns out regular pattern with while
> loop can do it too:

Yes, this is actually quite commonly done when writing ad-hoc parsers
in perl via variations on

while (s/^(the-leading-part)//g) {
  do something with $1;
}

Usually you need at least a counter to for example keep track of the
depth of parenthesis nesting, but maybe that part is being handled
outside the code snippet you demonstrate in this thread.

Re: Discovered pattern that ideally parses any quoting. Brackets highlighting in FSH.

[Sebastian Gniazdowski]

On 1 August 2018 at 20:18, Bart Schaefer <schaefer@brasslantern.com> wrote:
> On Mon, Jul 30, 2018 at 12:09 AM, Sebastian Gniazdowski
> <sgniazdowski@gmail.com> wrote:
>>
>> not only (z) can handle quoting, turns out regular pattern with while
>> loop can do it too:
>
> Yes, this is actually quite commonly done when writing ad-hoc parsers
> in perl via variations on
>
> while (s/^(the-leading-part)//g) {
>   do something with $1;
> }

I think my code is different. It matches distinct chars (not sure if
words are possible, sometimes  I think yes, sometimes it seems a no
go, I would have to write it) that we are interested in at the same
time handling backslash, " and '. So user can query for set of chars,
with their positions in buffer, without bothering with "are
backslashes even or odd" and other such problems. It turns out this
comes down to handling backslash, because of obvious "\\\A" problems
(parity of the slash), but also \" – it is also backslash that
controls whether double-quoting ends or begins. This is the [\\](*)
part of the pattern and restarting with ${match[3]}.

I realized that \', i.e. backslash and single quote, aren't special
within single quote. So the code needs update. But it is totally
possible, in the ${match[3]} block (the first one) one has to check if
state==in-single-quote, and decide to skip next char accordingly.

-- 
Sebastian Gniazdowski
News: https://twitter.com/ZdharmaI
IRC: https://kiwiirc.com/client/chat.freenode.net:+6697/#zplugin