Is it expected behavior that a Sentence can have two fields with the same name?

[BeckySharp]

see title. This is currently possible, if not desired I can add a require or something

[myedibleenso]

Hmm...

So, for example, chunks:

Sentence(
  numTokens=9,
  fields=Seq(
    TokensField(
      name="words",
      tokens=Seq("he", "wore", "an", "itsy", "bitsy", "teeny", "weeny", "yellow", "speedo"),
    ),
    TokensField(
      name="chunks",
      tokens=Seq("O", "O", "O", "B-ADJP", "I-ADJP", "I-ADJP",  "I-ADJP",  "I-ADJP",  "I-ADJP", "O"),
    ),
    TokensField(
      name="chunks",
      tokens=Seq("O", "O", "O", "B-NP", "I-NP", "I-NP",  "I-NP",  "I-NP",  "I-NP", "I-NP"),
    ),
    TokensField(
      name="chunks",
      tokens=Seq("B-NP", "O", "B-NP", "I-NP", "I-NP",  "I-NP",  "I-NP",  "I-NP", "I-NP"),
    ),
    TokensField(
      name="chunks",
      tokens=Seq("O", "B-DP", "I-DP", "I-DP", "I-DP",  "I-DP",  "I-DP",  "I-DP", "I-DP"),
    ),
    TokensField(
      name="chunks",
      tokens=Seq("O", "B-VP", "I-VP", "I-VP", "I-VP",  "I-VP",  "I-VP",  "I-VP", "I-VP"),
    )
  )
)

... is currently possible, right?

[marcovzla]

There is nothing stopping you from making an OdinsonDocument like that, but why would you want to index that? what behavior do you expect?

[BeckySharp]

i would expect odinson to warn me at index time at least that i'm going to be... overwriting?? things? doing something weird/bad or something? idk...
i'm not saying that it should be valid, rather I am asking if we should let people make the "mistake" or if we should issue a warning / error

[myedibleenso]

I may not have understood the original issue. I wanted to share an example of what I thought this issue describes to anchor the discussion.

why would you want to index that? what behavior do you expect?

I'd like to be able to match different types of phrases (chunks in this example) in different rules. This example is meant to illustrate that these overlap or one might subsume another. An alternative to this is to name each of these fields differently (ex. nps, vps, etc.).

I don't want to end up with only the last entry with this name when indexing.

[myedibleenso]

^I haven't tried indexing and querying that example, so I don't know what the current behavior is tbh.

[BeckySharp]

@myedibleenso wanna make a unit test to see what currently happens, and then we can (a) decide what we think the behavior should be (should we allow it? should we disallow it?) and then we can (b) make sure we achieve that behavior?

[myedibleenso]

Whatever behavior we decide we want, I agree that it should be tested.

Sure, I'll create a unit test as a starting point.