Compromise: Capture group doesn't work for .+ or *

[Drache93]

Quick one since we're all ready for the holidays!
Happy to go into more detail when people are free.

Capture groups seem to capture everything BEFORE your target, if using .+ or *. Using OR (#Noun|#Verb)+ works fine.

Quick console example:

> nlp('John always ravenously eats his glue').match('john [.+] eats').tag('test').debug()
=====
  -----
  ｜ 'John'     - MaleName, FirstName, Person, ProperNoun, Singular, Noun, TitleCase, Test
  ｜ 'always'   - Adverb, Test
  ｜ 'ravenously'  - Adverb, Test
  ｜ 'eats'     - PresentTense, Verb, Test

> nlp('John always ravenously eats his glue').match('john [*] eats').tag('test').debug()
=====
  -----
  ｜ 'John'     - MaleName, FirstName, Person, ProperNoun, Singular, Noun, TitleCase, Test
  ｜ 'always'   - Adverb, Test
  ｜ 'ravenously'  - Adverb, Test
  ｜ 'eats'     - PresentTense, Verb, Test

> nlp('John always ravenously eats his glue').match('john [#Adverb+] eats').tag('test').debug()
=====
  -----
  ｜ 'always'   - Adverb, Test
  ｜ 'ravenously'  - Adverb, Test

> nlp('John always ravenously eats his glue').match('john [(#Noun|#Adverb)+] eats').tag('test').debug()
=====
  -----
  ｜ 'always'   - Adverb, Test
  ｜ 'ravenously'  - Adverb, Test

Last two examples are correct and expected behaviour of the first 2.

[spencermountain]

ah, thank you.
Looking into the issue, i guess I see what this comment is for ;).

um, maybe I could ask your advice on something similar that I got stuck on -
It may be the same issue:

// === optional greedy at start ==
nlp(`wayne's world, party time`)
  .match('#Noun+? wayne')
  .debug()
// 'wayne's world, party time'

'aaa'.match(/b?a/)
//a


// ==== greedy-until-anything====
nlp('Toronto Ontario foo')
  .match('#Place+ .')
  .debug()
//'toronto ontario'

'aaab'.match(/a+b/)

I think your examples are the same as the 2nd example, right?
this stuff does my head in.
thanks

[Drache93]

Yeh was wondering about those comments :D Will take a look proper look tomorrow but looks like a bug if I'm understanding the expected behaviour...

This works fine:

nlp(`wayne's world, party time`)
  .match('#Noun+? world') // changed end to 2nd word
  .text()
// wayne's world

Probably escaping too early somewhere...

[Drache93]

You might be able to shed some light here...

test('optional greedy at start - first', function(t) {
  let doc = nlp(`wayne's world, party time`).match('#Noun+? wayne')
  t.equal(doc.text(), 'wayne')

  t.end()
})

Before the result actually get's returned as a doc, the match gets run quite a number of times, with differing result, I'm printing out the sliced terms and the current regs:

result: wayne's [{"tag":"Noun"}]
result: wayne's [{"tag":"Possessive"}]
result: wayne's [{"tag":"Person"}]
result: wayne's [{"tag":"FirstName"}]
result: wayne's world [{"tag":"FirstName"},{"choices":[{"tag":"Singular"},{"tag":"Possessive"}],"operator":"or"}]
result: world [{"end":true,"anything":true}]
result: wayne's world [{"tag":"FirstName"},{"choices":[{"tag":"Noun"},{"tag":"TitleCase"}],"operator":"or"}]
result: wayne's [{"start":true,"tag":"Possessive"}]
result: world [{"tag":"LastName"}]
result: wayne's world [{"choices":[{"tag":"TitleCase"},{"tag":"Singular"}],"operator":"or"},{"optional":true,"tag":"Acronym"},{"tag":"LastName"}]
result: world [{"end":true,"anything":true}]
result: wayne's world party time [{"optional":true,"greedy":true,"tag":"Noun"},{"word":"wayne"}]

Second last line get's us into here https://github.com/spencermountain/compromise/blob/dev/src/Phrase/match/03-tryMatch.js#L97 which is where things go wrong. I don't quite get the logic of continuing if it's optional, but this is only happening because world has "end":true.

But generally, I would expect to see optional/end waaay back at the start.

[spencermountain]

yeah, for cleaner logs i either run nlp.tokenize() or i comment-out this line to reduce the number of match statements run by the tagger.

if you're looking at a pr today/tomorrow, use the dev branch. I've got a bunch of fixes there

[Drache93]

Ah that's fair, I'll take another look. Thanks, not in any rush this year!

[Drache93]

Will take another look at this next!

[spencermountain]

hey @Drache93 I just discovered the .export() format bungles any sentence with a contraction. It's kind of bad. I should have caught this before now.

It'll require a breaking change - the good news though, is that it's an opportunity to change anything you'd like in this release - so free license to you.
Take a crack at anything you want.

[Drache93]

Ah bad luck! Haha that's a dangerous idea! 😂

Would you be up for selecting and or creating some good issues for next version that are big enough? Might as well get other breaking changes and/or just major stuff too! Ooh, I wonder if a hook into the HTML output might be good for creating react content...