1 / 12

Simple Type Inference for Structural Polymorphism

Simple Type Inference for Structural Polymorphism. Jacques Garrigue Type Refinements Seminar. Motivation. HM (X) used to explain structural polymorphism Assume subtyping absent HM(X) is not a simple extension to previous algorithms Constraints occur in all types

cleatus
Download Presentation

Simple Type Inference for Structural Polymorphism

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Simple Type Inference for Structural Polymorphism Jacques Garrigue Type Refinements Seminar

  2. Motivation • HM (X) used to explain structural polymorphism • Assume subtyping absent • HM(X) is not a simple extension to previous algorithms • Constraints occur in all types • Type errors become unreadable

  3. Proposed Solution • Have “local constraints” : constraints on top of classic Hindley-Milner system • Constraints embedded in kinding context for type variables • Less powerful than HM(X)

  4. Type System • Types  ::=  | u | ! • Kind envs K ::=  | K, :: (C,9.R) • Polytypes  ::=  | 8.K . 

  5. Free Variables of Types • FVK(81,,n.K’ . ) = FVK,K’()\{1,,n} • FVK,:(C,9 .R) () = {}[ FVK(R) • FVK(u) =  • FVK(1!2) = FVK(1)[ FVK(2)

  6. Type Substitutions • K `:K’ ( is admissible between K and K’) If for all ::(C,9.R) 2 K, () is ’ • ’::(C’,9’ R’) 2 K’ • C’ ² C •  :  !’ s.t. ( R)µ R exists

  7. Typing Rules K,K0` : K Dom ()½ B K;,x:8 B.Ko. ` x:() K;,x:` e:’ K;` fun x ! e : !’ K;` e1:!’ K;` e2: K;` e1 e2 : ’

  8. Typing Rules (contd.) K;` e1 :  K;, x:` e2 :  K;` let x = e1 in e2 :  K; ` e :  B = FVK() \ FVK () K|B; ` e : 8 B.K|B. Gen

  9. Compare with HM (X) K;` e :  B = (FV () [ FV(KB) ) FV () 9 B.K;` e : 8 B.KB. KBµ K containing all kinds related to B Generalizes over tyvars not “relevant” HMX-Gen

  10. Relevant type variable • Technical definition in paper • Basically, the meaning of the type depends on what kind the variable gets • Recall, kinds include constraints • Useful to understand constraints/ type errors

  11. Benefits • Easier understanding of constraints • e.g. If  FVK(1) [ FVK(2), solving K Æ1 = 2 does not give us more information about 

  12. Conclusions • Less powerful than HM (X) : no subtyping • Gives less precise answers • More practical (??) • Sufficient for most purposes

More Related