Link endRegex to open blocks of the same rule

[TonyGravagno]

Problem:
Right now, an endRegex can fire even if no block from that rule is open, which:

• Allows "stray" ends that close some other rule's block.
• Forces us to make endRegex very conservative, or accept misfolds.

Feature request: per-rule "active block stack"

• Only consider endRegex when there is an open block created by that rule's beginRegex (or middleRegex for transitions).
• Ignore endRegex hits for that rule if its stack is empty.

That would allow us to be much more liberal with various end patterns, because they couldn't kill a block that doesn't belong to them.

[daiyam]

For performance, it's designed to reduce the matching to a single loop/regex (main).

Only consider endRegex when there is an open block created by that rule's beginRegex (or middleRegex for transitions).

It can be achieve with "nested": false.

Ignore endRegex hits for that rule if its stack is empty.

Can you give an example (settings and code)? Thx

[TonyGravagno]

It can be achieve with "nested": false.

With nested:false, processing skips to the end of the block before processing anything else, right? If so then we lose all nested folding for that rule.

[TonyGravagno]

Ignore endRegex hits for that rule if its stack is empty.

Can you give an example (settings and code)? Thx

I didn't word that correctly. My thought is that each time the Begin value of a rule matches, it should push the rule/Begin onto a stack. Each time the End value of a rule matches, it should pop the stack back to the Begin point. If the End value of a rule is triggered and that rule is not already on the stack, then we don't want to terminate that rule - it's the wrong rule, move on to the next rule that matches.
Again, this prevents rule 2 from ending a fold begun by rule 1.

I'm looking at a lot of code and rules right now, and have already addressed these issues with complex regex. I don't like this regex and want to simplify it. If necessary I'll pull an example later - but this scenario is commonly reported so I hope I don't need to contrive an example myself.

Thanks!

[daiyam]

If the End value of a rule is triggered and that rule is not already on the stack, then we don't want to terminate that rule

it's what it currently do.

it's the wrong rule, move on to the next rule that matches.

currently, it goes to the next match...

The extension try to rely on single regex build from the rules on the same level and accepting nesting.
So basically a big regex like begin1|end1|begin2|end2|..., so there is some overlap, only the first regex will be used.
To do what you ask, I will need to be able to detect any collision in the end regexes. Maybe for those of simple equality. But for more complex ones, the detection isn't fast and not 100%.

From what I could understand, you are trying to add folding to js/ts files.
It's not easy to parse it with only regexes (look how syntax highlighting is defined: https://github.com/microsoft/vscode/blob/main/extensions/javascript/syntaxes/JavaScript.tmLanguage.json)

If you could have access to the textmate scopes generated by the syntax highlighting, maybe what you trying to do would have been easier.

PS: I only use it for {{{/}}}...

[daiyam]

Maybe by adding a flag "fallback": true, and by testing only the top item of the stack if it's an end and if it can also match the matched text, then it could be closed correctly...
Maybe, the flag might not be necessary.

Can you try #144?

[TonyGravagno]

I spent my entire day with the extension and new rules. 😁
I see what you did with #144. I will try within the next few days.
I think the flag should be required, either a fallbackFromHere or a fallbackToHere (with shorter names). Otherwise there's a risk of changing expected behavior for someone who has already crafted rules in a specific order.
🙇‍♂️