Didier Rémy
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Didier Rémy has a total of 4 co-authors. Here are the top co-authors by number of shared publications:
- Clément Blaudeau
INRIA, Paris, France
2 shared publications - Gabriel Radanne
INRIA, Lyon, France
2 shared publications - Gabriel Scherer
Inria IRIF Université Paris Cité Paris France
2 shared publications - Alistair O'BRien
University of Cambridge, United Kingdom
1 shared publication
Latest Publications
- Open Access
- 15 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Programming Languages and Systems
Omnidirectional type inference for ML: principality any way
The Damas-Hindley-Milner (ML) type system owes its success to principality , the property that every well-typed expression has a unique most general type. This makes inference predictable and efficient. Yet, principality is fragile : many extensions of ML—GADTs, higher-rank polymorphism, and static overloading—break it by introducing fragile constructs that resist principal inference. Existing approaches recover principality through directional inference algorithms, which propagate known type information in a fixed (or static ) order ( e.g. as in bidirectional typing) to disambiguate such constructs. However, the rigidity of a static inference order often causes otherwise well-typed programs to be rejected. We propose omnidirectional type inference, where type information flows in a dynamic order. Typing constraints may be solved in any order, suspending when progress requires known type information and resuming once it becomes available, using suspended match constraints . This approach is straightforward for simply typed systems, but extending it to ML is challenging due to let-generalization . Existing ML inference algorithms type let-bindings in a fixed order—type the let-bound expression first, generalize its type, and then type the let-body. To overcome this, we introduce incremental instantiation , allowing partially solved type schemes containing suspended constraints to be instantiated, with a mechanism to incrementally update instances as the scheme is refined. Omnidirectionality provides a general framework for restoring principality in the presence of fragile features. We demonstrate its versatility on two fundamentally different features of OCaml: static overloading of record labels and datatype constructors and semi-explicit first-class polymorphism. In both cases, we obtain a principal type inference algorithm that is more expressive than OCaml’s current typechecker.
- 15 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Programming Languages and Systems
Open AccessOmnidirectional type inference for ML: principality any way
The Damas-Hindley-Milner (ML) type system owes its success to principality , the property that every well-typed expression has a unique most general type. This makes inference predictable and efficient. Yet, principality is fragile : many extensions of ML—GADTs, higher-rank polymorphism, and static overloading—break it by introducing fragile constructs that resist principal inference. Existing approaches recover principality through directional inference algorithms, which propagate known type information in a fixed (or static ) order ( e.g. as in bidirectional typing) to disambiguate such constructs. However, the rigidity of a static inference order often causes otherwise well-typed programs to be rejected. We propose omnidirectional type inference, where type information flows in a dynamic order. Typing constraints may be solved in any order, suspending when progress requires known type information and resuming once it becomes available, using suspended match constraints . This approach is straightforward for simply typed systems, but extending it to ML is challenging due to let-generalization . Existing ML inference algorithms type let-bindings in a fixed order—type the let-bound expression first, generalize its type, and then type the let-body. To overcome this, we introduce incremental instantiation , allowing partially solved type schemes containing suspended constraints to be instantiated, with a mechanism to incrementally update instances as the scheme is refined. Omnidirectionality provides a general framework for restoring principality in the presence of fragile features. We demonstrate its versatility on two fundamentally different features of OCaml: static overloading of record labels and datatype constructors and semi-explicit first-class polymorphism. In both cases, we obtain a principal type inference algorithm that is more expressive than OCaml’s current typechecker.
- Open Access
- 7 January 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Programming Languages
- Vol. 9 (POPL) , 1962-1991
Avoiding Signature Avoidance in ML Modules with Zippers
We present ZipML , a new path-based type system for a fully fledged ML-module language that avoids the signature avoidance problem. This is achieved by introducing floating fields , which act as additional fields of a signature, invisible to the user but still accessible to the typechecker. In practice, they are handled as zippers on signatures, and can be seen as a lightweight extension of existing signatures. Floating fields allow to delay the resolution of instances of the signature avoidance problem as long as possible or desired. Since they do not exist at runtime, they can be simplified along type equivalence, and dropped once they became unreachable. We give rewriting rules for the simplification of floating fields without loss of type-sharing and present an algorithm that implements them. Remaining floating fields may fully disappear at signature ascription, especially in the presence of toplevel interfaces. Residual unavoidable floating fields can be shown to the user as a last resort, improving the quality of error messages. Besides, ZipML implements early and lazy strengthening, as well as lazy inlining of definitions, preventing duplication of signatures inside the typechecker. The correctness of the type system is proved by elaboration into M ω , which has itself been proved sound by translation to F ω . ZipML has been designed to be an improvement over OCaml that could be retrofitted into the existing implementation.
- 7 January 2025
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Programming Languages
- Vol. 9 (POPL) , 1962-1991
Open AccessAvoiding Signature Avoidance in ML Modules with Zippers
We present ZipML , a new path-based type system for a fully fledged ML-module language that avoids the signature avoidance problem. This is achieved by introducing floating fields , which act as additional fields of a signature, invisible to the user but still accessible to the typechecker. In practice, they are handled as zippers on signatures, and can be seen as a lightweight extension of existing signatures. Floating fields allow to delay the resolution of instances of the signature avoidance problem as long as possible or desired. Since they do not exist at runtime, they can be simplified along type equivalence, and dropped once they became unreachable. We give rewriting rules for the simplification of floating fields without loss of type-sharing and present an algorithm that implements them. Remaining floating fields may fully disappear at signature ascription, especially in the presence of toplevel interfaces. Residual unavoidable floating fields can be shown to the user as a last resort, improving the quality of error messages. Besides, ZipML implements early and lazy strengthening, as well as lazy inlining of definitions, preventing duplication of signatures inside the typechecker. The correctness of the type system is proved by elaboration into M ω , which has itself been proved sound by translation to F ω . ZipML has been designed to be an improvement over OCaml that could be retrofitted into the existing implementation.
- Open Access
- 29 April 2024
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Programming Languages
- Vol. 8 (OOPSLA1) , 194-222
Fulfilling OCaml Modules with Transparency
ML modules come as an additional layer on top of a core language to offer large-scale notions of composition and abstraction. They largely contributed to the success of OCaml and SML. While modules are easy to write for common cases, their advanced use may become tricky. Additionally, despite a long line of works, their meta-theory remains difficult to comprehend, with involved soundness proofs. In fact, the module layer of OCaml does not currently have a formal specification and its implementation has some surprising behaviors. Building on previous translations from ML modules to Fω, we propose a type system, called Mω, that covers a large subset of OCaml modules, including both applicative and generative functors, and extended with transparent ascription. This system produces signatures in an OCaml-like syntax extended with Fω quantifiers. We provide a reverse translation from Mω signatures to path-based source signatures along with a characterization of signature avoidance cases, making Mω signatures well suited to serve as a new internal representation for a typechecker. The soundness of the type system is shown by elaboration in Fω. We improve over previous encodings of sealing within applicative functors, by the introduction of transparent existential types, a weaker form of existential types that can be lifted out of universal and arrow types. This shines a new light on the form of abstraction provided by applicative functors and brings their treatment much closer to those of generative functors.
- 29 April 2024
- Published by Association for Computing Machinery (ACM) in Proceedings of the ACM on Programming Languages
- Vol. 8 (OOPSLA1) , 194-222
Open AccessFulfilling OCaml Modules with Transparency
ML modules come as an additional layer on top of a core language to offer large-scale notions of composition and abstraction. They largely contributed to the success of OCaml and SML. While modules are easy to write for common cases, their advanced use may become tricky. Additionally, despite a long line of works, their meta-theory remains difficult to comprehend, with involved soundness proofs. In fact, the module layer of OCaml does not currently have a formal specification and its implementation has some surprising behaviors. Building on previous translations from ML modules to Fω, we propose a type system, called Mω, that covers a large subset of OCaml modules, including both applicative and generative functors, and extended with transparent ascription. This system produces signatures in an OCaml-like syntax extended with Fω quantifiers. We provide a reverse translation from Mω signatures to path-based source signatures along with a characterization of signature avoidance cases, making Mω signatures well suited to serve as a new internal representation for a typechecker. The soundness of the type system is shown by elaboration in Fω. We improve over previous encodings of sealing within applicative functors, by the introduction of transparent existential types, a weaker form of existential types that can be lifted out of universal and arrow types. This shines a new light on the form of abstraction provided by applicative functors and brings their treatment much closer to those of generative functors.