use data instead of fun so pattern-matching on def's is possible

[odanoburu]

are there any drawbacks to these changes? (e.g., performance-wise?) I found that several constructs were already declared as data, but some were not.

[krangelov]

The definition of the language demands that all functions defined as `data` must be in the same module as the `cat` definition for their result type. This is kind of like in Haskell. The RGL has one Cat module which defines all categories and then the functions are spread in several modules. This means that using `data` in the RGL is against the language specification. However, I am not sure that the compiler checks that restriction. Best Regards, Krasimir

…

On 13 August 2018 at 21:45, bruno cuconato ***@***.***> wrote: are there any drawbacks to these changes? (e.g., performance-wise?) I found that several constructs were already declared as data, but some were not. ------------------------------ You can view, comment on, or merge this pull request online at: #8 Commit Summary - - use data instead of fun so pattern-matching on def's is possible File Changes - *M* src/abstract/Adjective.gf <https://github.com/GrammaticalFramework/gf-rgl/pull/8/files#diff-0> (3) - *M* src/abstract/Conjunction.gf <https://github.com/GrammaticalFramework/gf-rgl/pull/8/files#diff-1> (2) - *M* src/abstract/Noun.gf <https://github.com/GrammaticalFramework/gf-rgl/pull/8/files#diff-2> (4) - *M* src/abstract/Sentence.gf <https://github.com/GrammaticalFramework/gf-rgl/pull/8/files#diff-3> (2) - *M* src/abstract/Structural.gf <https://github.com/GrammaticalFramework/gf-rgl/pull/8/files#diff-4> (2) - *M* src/abstract/Tense.gf <https://github.com/GrammaticalFramework/gf-rgl/pull/8/files#diff-5> (2) Patch Links: - https://github.com/GrammaticalFramework/gf-rgl/pull/8.patch - https://github.com/GrammaticalFramework/gf-rgl/pull/8.diff — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ATBZZGhmbpl4gtO6P2UW7rLdLSLNVKXnks5uQddrgaJpZM4V7HPg> .

[odanoburu]

@krangelov thanks for the answer!

when you say "in the same module" you mean literally in the same file, or would an extension (**) of that file also work? if the later is fine, this is why this PR seems work alright (not only the CI build, but I've been using this patch now). if so, can we merge this? (as I said, these are not the only data declarations, the RGL is already full of them!)

However, I am not sure that the compiler checks that restriction.

it does, I tried adding data the grammar I was developing and got an error!

[krangelov]

No extension is not an option. If you allow data types to be extended then we should also make it possible to extend def rules. This would be a nice feature but it is not a trivial extension.

…

On 15 August 2018 at 13:50, bruno cuconato ***@***.***> wrote: @krangelov <https://github.com/krangelov> thanks for the answer! when you say "in the same module" you mean literally in the same file, or would an extension (**) of that file also work? if the later is fine, this is why this PR seems work alright (not only the CI build, but I've been using this patch now). if so, can we merge this? (as I said, these are not the only data declarations, the RGL is already full of them!) However, I am not sure that the compiler checks that restriction. it does, I tried adding data the grammar I was developing and got an error! — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ATBZZAQzXs01CA8yW_tqG0Xx7xbnGaEnks5uRAsUgaJpZM4V7HPg> .

[odanoburu]

sorry, I wasn't very clear: I meant extended modules, not extended types! i.e., the RGL has several cases of data declarations that occur in modules that extend the Cat module. I'm merely adding some more; is that ok? if not, I guess I'm misunderstanding something -- can you provide an example?

[odanoburu]

misclicked the close button, sorry.

[inariksit]

@krangelov So what was the conclusion, does "against the language specification" mean that we should not merge this, or do we not care?

[krangelov]

I think it is better not to merge it. Anyway if you want to pattern match on trees it is so far simpler to do it outside GF.

…

On 21 August 2018 at 13:42, Inari Listenmaa ***@***.***> wrote: @krangelov <https://github.com/krangelov> So what was the conclusion, does "against the language specification" mean that we should not merge this, or do we not care? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#8 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ATBZZI6nIxDqJav06rjpPYOXvn3bQvimks5uS_I-gaJpZM4V7HPg> .

[odanoburu]

@krangelov should I make a PR removing the existing instances of this problem?

Anyway if you want to pattern match on trees it is so far simpler to do it outside GF.

yes, but at the cost of another dependency..

[hleiss]

My opinion on this old question is:
of course, tree transformations ought to be definable in GF itself, as part of the RGL; so most fun's should be data's.

a) one might want to define a tree normalization nf : trees -> trees, e.g. to transform trees with possessive genitive attribute, PossNP cn np, to trees with attributive possessive pronoun, DetCN (DetQuant (PossPron pron) num) cn, in case np is a (UsePron pron).
b) a good translation between languages will need tree transformations, e.g. Ancient Greek has 11 participles, while Eng or Ger etc. have only 2. So, to express the intended meaning properly in Eng or Ger, many Greek participle constructions need to be transformed to relative or adverbial clauses in Eng or Ger etc. (and reversely, perhaps).

The data constructors in the RGL are in Noun, Numeral, Sentence, Verb with result types like Card, Digit, Numeral Cl, VP etc. that are defined in Cat. So, clearly the compiler does not enforce the restriction of C.3.6 mentioned by Krasimir.
(On the other hand, C.3.6 says that (instead of data f : A -> C) we can say fun f : A -> C ; data C = f , while the compiler says this is conflicting information on f.)