Bruce, K.B., Cardelli, L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Ito, T. (eds.): Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol. 1281. Springer-Verlag, Berlin Heidelberg New York (1997) 415--438  Home/Search   Document Details and Download   Summary   Related Articles   Check

....studied. The starting point of this thesis is a functional object based calculus, the Lambda Calculus of Objects of Fisher, Honsell, Mitchell [FHM94] formally introduced in Chapter 2 (from now on we will refer to it as LCO) Objectbased feautures are satisfactory modeled in a functional context [BCP97], and most of the calculi presented here are functional. More difficulties arise for class based languages in a functional context, so two imperative class based calculi are proposed. However, both of them, even though very different from each other, have strongly got benefits from the calculus in ....

....by Bruce in [Bru94] to give direct interpretations for class based object calculi. A number of type theoretic encodings for the so called object based calculi have then been formulated by Pierce and Turner [PT94] and Abadi, Cardelli and Viswanathan [AC96a, ACV96] Bruce, Pierce and Cardelli [BCP97], and recently by Crary [Cra98] These interpretations apply to a rich variety of object calculi with primitives of object formation, message send and (func147 tional) method override: they succeed in validating the operational semantics of these calculi as well as the expected subtyping relations ....

[Article contains additional citation context not shown here]

K.B. Bruce, L. Cardelli, and B. Pierce. Comparing Object Encodings. In Proc. of TACS'97, volume 1281 of LNCS, pages 415--438. Springer-Verlag, 1997.

....of the generator model were later proposed by Bruce [Bru94] to give interpretations of class based object calculi. A number of encodings for object based calculi have then been formulated by Pierce and Turner [PT94] Abadi, Cardelli and Viswanathan [AC96,ACV96,Vis98] Bruce, Pierce and Cardelli [BCP97], and by Crary [Cra98] These interpretations apply to a rich variety of object calculi with primitives of object formation, message send and (functional) method override: they succeed in validating the operational semantics of these calculi as well as the expected subtyping relations. None of ....

....methods. 7 Related Work The idea to split methods into different components is inspired by the object encoding of [ACV96] That translation applies only to nonextensible objects, which are encoded by a combined use of recursive and bounded existential types, subsequently named ORBE encoding [BCP97]. Our translation, instead, uses a combination of recursion and universal quantification to render MyType polymorphism. We are then able to obtain a corresponding translation for nonextensible objects with essentially equivalent results as [ACV96] A variant of the ORBE encoding that does not use ....

K. Bruce, L. Cardelli, and B. Pierce. Comparing Object Encodings. In Proc. of TACS'97, volume 1281 of Lecture Notes in Computer Science, pages 415-- 438. Springer-Verlag, 1997.

....to object oriented languages [7, 12, 18, 32] Last year s FOOL workshop even featured a panel discussion on typed intermediate languages. These intermediate languages are typically based on typed # calculi. There is significant precedent for encoding objectoriented languages in typed # calculi[2,4,5,6,9],butthis domain type preserving compilation imposes several new requirements and allows us to reject a few traditional assumptions. The intermediate language must provide extremely simple primitives (that correspond, e.g. to at most several machine instructions) so that our encodings are ....

....polymorphism in Objective ML for both class types and type inference on unordered records. Our calculus is explicitly typed, but we use ordered rows to represent the open type of self. Our object representation is superficially similar to several of the classic encodings in F# based languages [5, 26]. As in the Abadi, Cardelli, and Viswanathan encoding [2] method invocation uses self application; however, we hide the actual class of the receiver using existential quantification over row variables instead of splitting the object into a known interface and a hidden implementation. This allows ....

Kim Bruce, Luca Cardelli, and Benjamin Pierce. Comparing object encodings. In Proc. Int'l Symp. on Theoretical Aspects of Computer Software (TACS),Sendai, Japan, September 1997. To appear in Information and Computation.

....any object which has an appropriate method of the correct arity. Thus an object of one type may stand for an object of another as the receiver of an invocation. A number of approaches to the typing and subtyping of object systems are considered in the literature. Four approaches are compared in [BCP97]. One uses recursion to deal with interface types and is similar to our approach to typing objects. The other approaches use existential types or a combination of existential and recursive types. The approach to subtyping object languages using existential types to bind object self references ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Software (TACS), Sendai, Japan, September 1997. An earlier version was presented as an invited lecture at the Third International Workshop on Foundations of Object Oriented Languages (FOOL 3), July 1996.

....related direction is that of object encodings which might be thought of as syntactic presentations of semantics. Pierce and Turner [54] study the encoding of objects as abstract types, which bears some similarity to the parametricity semantics in this paper. More recent work along these lines is [12]. Fisher and Mitchell [21, 20] also relate classes to data abstraction, though this seems to be at a di erent level than that discussed here. All this work is usually carried out in a functional setting for objects, but some of the ideas deal with state. The major developments in the research ....

....is added to the state set in this fashion. Thus, every opening of the abstract type gives rise to a new instance with its own state component that is independent of all other state components. Remark. In comparing this operation with the object encoding proposed by Pierce, Turner and others [54, 12], we note that they treat objects as abstract types whereas we treat classes as abstract types. Our objects correspond to opened abstract types whose representation types are merged into the global state set. Sending a message to the object merely involves selecting a component of its method ....

K. C. Bruce, L. Cardelli, and B. C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Software, volume 1281 of LNCS, pages 415-438. Springer-Verlag, Berlin, 1997.

....principal type inference algorithm. While Nishimura s type system gives important and elegant foundations for dynamic messages, the typing discipline that results from his system seems to depart signi cantly from the foundational models of objects and object types found in the previous literature [AC96, Bru94, BSvG95, PT94, Mic90, FHM94, FM95, BCP97, CHC90]. In his second order system, no type is provided that directly represents the structure of objects and messages. Instead, the type information is expressed at the kind level: an object type has the form t : where t is a type variable that indexes the kind which, in turn, encodes all the ....

K. Bruce, L. Cardelli, and B. Pierce. Comparing Object Encodings. In Proc. of TACS'97, volume 1281 of Lecture Notes in Computer Science, pages 415-438. Springer-Verlag, 1997.

....theusual word spacing. Second and subsequentlv3N are indented. Footnotesshoul end with a period. Al x Hofmann et al. references if the pa is not completely filled. AnexaFY3 is given a the end of this informa x sheet. For cita;Dx in the textpleaD usesqua; bra kets as consecutive numbers: 1] [2], 3] 2.8 Page Numberk9 and Running Heads Your pa er should show no printed pa numbers; theyae aey cay by the volume editor. Indica the ordering of yourpaD by numbering the sheets in pencila the bottom of the reverse side. Do not set runningheain 2.9 Prc ting Quality For reproduction we ....

Bruce, K.B.,CardelxN L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Ito, T. (eds.):Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol 1281. Springer-Verlin Berln Heidel erg New York (1997) 415-- 438

....applications, it is also crucial to investigate the introduction of imperative features in our calculus. Finally we would like to nd an equational theory dealing with the objects of Obj , and to study possible representation of our typing discipline in the framework theory of F : see [9]) ....

K. Bruce, L Cardelli, and C. B. Pierce. Comparing Object Encoding. In Proc. of TACS, Lecture Notes in Computer Science. Springer-Verlag, 1997.

....are declared and provide alias protection for object oriented languages. It is nonsensical to infer which objects are representation; such information is part of the programmer s intention. Nevertheless, the underlying techniques are similar, modulo encoding objects into a functional language [9]. To be precise, rep corresponds to letregion, and context parameters correspond to region variables. Regions also include read and write effects, which we do not. To our knowledge, no existing formal system statically models object ownership and statically prevents object access from outside an ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing Object Encodings. In Theoretical Aspects of Computer Software (TACS'97), LNCS 1281, pages 415--438, 1997.

....type system is based on polymorphic recursively constrained types, for which they have a sound type inferencing algorithm. The main advantage of this approach is the extreme exibility a orded by recursively constrained types. However, inferred types are large and dicult to read. Bruce et al. [11] show how the main approaches to modeling objects can be seen in a uni ed framework. The state of the art in modeling classes is not as well established. We hope that this work might be a step in this direction. 8 Conclusions and Future Work We have presented a core calculus of classes and ....

Bruce, K. B., L. Cardelli and B. C. Pierce, Comparing object encodings, in: Proc. TACS '97 (1997), pp. 415-438.

....Our purpose in this paper is to introduce an encoding of Mitchell al. calculus, extended with subtyping [14] into a calculus with extensible records equipped with a rich but fairly standard type system, that has been recognized as a suitable framework for studying the typing of objects, see [18, 7]. We show in particular how the typing rules for prototypes can be derived in a natural way. The idea of the encoding is very simple. To explain it, let us rst come back to the diOEculty of typing method specialization for prototypes: there are two seemingly opposite requirements for types in this ....

....However, using the rules of instanciation, subsumption and generalization, one can achieve the same eoeect. That is, the following inference is valid: Gamma M : 8t :oe) Gamma 0 ; Gamma oe oe 0 Gamma M : 8t 0 :oe 0 ) 0) Our type system is quite close to the one of [7] (see also [18] for a complete description of the same system without recursion) with some dioeerences: we do not use maximal types ( nor existentially quantied types, but we use extensible records rather than just records, and recursion at any kind. Our system is also very close, apart from ....

[Article contains additional citation context not shown here]

K. Bruce, L. Cardelli, B. Pierce, Comparing object encodings, TACS'97, Lecture Notes in Comput. Sci. 1281 (1997) 415-438.

.... the correctness of the abstract type based approach of Haskell [9] as well as designing a COM interface for Moby, a ML like language supporting class based object oriented programming [5] The encodings we consider are similar in spirit to the encodings for object oriented programming such as [2]. This paper is structured as follows. In Section 2, we informally review the basics of COM, at the level required by this paper. In Section 3, we present the basic encoding of COM into F : In Section 4, we discuss issues of composition, namely containment and aggregation. Section 5 adds ....

K. B. Bruce, L. Cardelli, and B. C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Science (TACS'97), 1997.

....system is based on polymorphic recursively constrained types, for which they have a sound type inferencing algorithm. The main advantage of this approach is the extreme flexibility afforded by recursively constrained types. However, inferred types are large and difficult to read. Bruce et al. [16] show how the main approaches to modeling objects can be seen in a unified framework. The state of the art in modeling classes is not as well established. We hope that this work might be a step in this direction. To the best of our knowledge, there are not many formal settings in which ....

K. B. Bruce, L. Cardelli, and B. C. Pierce. Comparing object encodings. In Proc. TACS '97, pages 415--438. LNCS 1281, Springer-Verlag, 1997.

....have a sound type inferencing algorithm. The main advantage of this approach is the extreme flexibility afforded by recursively constrained types. Currently, the main problem is that inferred types are large and difficult to read. Work in this area is in progress. There is an interesting paper [BCP97] which shows how the main approaches to modeling objects can be seen in a unified framework. The state of the art in modeling classes is not as well established. We hope that our core calculus might be a step in this direction. 7 Conclusions and Future Work We have presented a core calculus of ....

K. Bruce, L. Cardelli and B. C. Pierce. Comparing Objects Encodings. In TACS '97, LNCS 1281, 415--438, Springer-Verlag, 1997.

.... see e.g. 5, 17, 21, 6, 12, 8, 18] There is a whole body of research on encodings of classes using recursive or existential types, in a suitably rich polymorphic type theory (like F : or F : Four such (functional) encodings are formulated and compared in a common notational framework in [4]. But they all use quanti cation or recursion over type variables, which is not To appear in: Theorem Proving in Higher Order Logics (TPHOLs2000) Springer LNCS, 2000. available in the higher order logic (the logics of pvs and isabelle hol) that will be used here. Quanti cation and or ....

.... ; 1 ; n . New types or type constructors can be introduced via de nitions, as in: 1 Alternatively, one can combine these n functions into elements of a trait type [ f1 : Self 1 ; fn : Self n ] like in [1, xx8.5. 2] 2 As a side remark, all the encodings discussed in [4] implicitly also use coalgebras. 3 Translating to both pvs and isabelle hol o ers the veri er a choice which proof tool to use. type theory lift[ TYPE def = f bot : unit j up : g where unit is the empty labeled product type [ This lift type constructor adds a bottom element to an ....

K.B. Bruce, L. Cardelli, and B.C. Pierce. Comparing object encodings. In M. Abadi and T. Ito, editors, Theoretical Aspects of Computer Software, number 1281 in Lect. Notes Comp. Sci., pages 415-438. Springer, Berlin, 1997.

.... see e.g. 5, 15, 6, 11, 8, 16] There is a whole body of research on encodings of classes using recursive or existential types, in a suitably rich polymorphic type theory (like F : or F : Four such (functional) encodings are formulated and compared in a common notational framework in [4]. But they all use quanti cation or recursion over type variables, which is not available in the higher order logic (comparable to the logics of pvs and isabelle hol) that will be used here. The setting of the encoding in [16] is higher order logic with extensible records . This framework is ....

....for a term N : f lab 1 : 1 j : j lab n : n g in this coproduct type, together with n 1 Alternatively, one can combine these n functions into elements of a trait type [ f1 : Self 1 ; fn : Self n ] like in [1, xx8.5. 2] 2 As a side remark, all the encodings discussed in [4] implicitly also use coalgebras. 3 terms L i (x i ) possibly containing a free variable x i : i , there is a case term CASES N OF f lab 1 x 1 7 L 1 (x 1 ) j : j lab n xn 7 Ln (x n ) g of type . These introduction and elimination constructions for exponents and labeled ....

K.B. Bruce, L. Cardelli, and B.C. Pierce. Comparing object encodings. In M. Abadi and T. Ito, editors, Theoretical Aspects of Computer Software, number 1281 in Lect. Notes Comp. Sci., pages 415-438. Springer, Berlin, 1997.

....principal type inference algorithm. While Nishimura s type system gives important and elegant foundations for dynamic messages, the typing discipline that results from his system seems to depart significantly from the foundational models of objects and object types found in the previous literature [AC96, Bru94, BSvG95, PT94, Mic90, FHM94, FM95, BCP97, CHC90]. In his second order system, no type is provided that directly represents the structure of objects and messages. Instead, the type information is expressed at the kind level: an object type has the form t : where t is a type variable that indexes the kind which, in turn, encodes all the ....

K. Bruce, L. Cardelli, and B. Pierce. Comparing Object Encodings. In Proc. of TACS'97, volume 1281 of Lecture Notes in Computer Science, pages 415--438. Springer-Verlag, 1997.

....small print and placed at the end of your contribution. Please do not insert a pagebreak before the list of references if the page is not completely filled. An example is given at the end of this information sheet. For citations in the text please use square brackets and consecutive numbers: 1] [2], 3] 2.8 Page Numbering and Running Heads Your paper should show no printed page numbers; they are allocated by the volume editor. Indicate the ordering of your pages by numbering the sheets in pencil at the bottom of the reverse side. Do not set running heads. 2 The footnote numeral ....

Bruce, K.B., Cardelli, L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Ito, T. (eds.): Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol. 1281. Springer-Verlag, Berlin Heidelberg New York (1997) 415-- 438

....that a translation of the full LOOM language would necessarily involve some inconvenient workarounds. Of more immediate concern, however, was the fact that the type system of FLINT did not include existential types. Existentials are an integral part of most of the well known encodings of objects [BCP96] where they serve to hide the types of instance variables. CHAPTER 3. OUR INTERMEDIATE LANGUAGE IL 23 A translation from an intermediate language with existential types to FLINT could have been implemented, but would have required more work than time allowed. Interprocedural dataflow analysis ....

....types. In an effort to simplify the semantic processing of IL programs we omitted records in favor of simpler ordered tuples with unlabeled fields; it did not seem necessary to support full fledged record types when unlabeled finite products would do just as well. In most encodings for objects, BCP96] though, record types are used to describe collections of methods or instance variables. At first glance it does not seem that there is a substantial difference between records and tuples for this 32 CHAPTER 4. THE SUBTYPE ENHANCED INTERMEDIATE LANGUAGE SIL 33 : j fl 1 : 1 ; l ....

[Article contains additional citation context not shown here]

Kim B. Bruce, Luca Cardelli, and Benjamin Pierce. Comparing object encodings. In Proceedings of the Third Workshop on Foundations of Object Oriented Languages, 1996. 66 BIBLIOGRAPHY 67

.... 1 p. We call this view external because it characterizes an object behaviour in terms of a pattern of interactions with it, instead of in terms of a hidden internal state. Similarly, we call the packages view internal. These encodings were compared recently by Bruce, Cardelli and Pierce [3]. They considered two other encodings as well, and made their comparison along a number of dimensions, including suitability for inheritance, method update, and binary methods. For our present concerns, however, one of their observations is especially important. They remarked that, if recursion is ....

....however, one of their observations is especially important. They remarked that, if recursion is present in the language, the two encodings give different results. And one certainly does want recursion, particularly since this is the means of modelling the self construct [30] The authors of [3] seem annoyed. The reason could be that computational intuition suggests that the internal and external views, given by the package and recursive record encodings, should amount two ways of describing the same thing. So the disagreement of the two encodings is disappointing. However, as with the ....

K.B. Bruce, L. Cardelli, and B.C. Pierce. Comparing object encodings. In Invited lecture at Third Workshop on Foundations of Object Oriented Languages (FOOL 3), July 1996. Available electronically in informal workshop proceedings.

....[9] types. This is important so that TAL may serve as a target language for compilers of high level languages with these features, of course, but even monomorphic features of high level languages often implicitly require type quantification. For example, existential types arise in objects [1] and closures [8] and universal types arise in stacks [11] For simplicity, universal quantification is limited to code, since that is where the intended use of universal types lies. The introduction form (and the type) allows for the simultaneous abstraction of a series of type variables ff 1 ; ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Software, Sendai, Japan, September 1997.

....we are at a loss to see how to allow non monotonic type constructors. However, this is not a limit of our approach, but just of its current formalizations. Chapter 7 Comparing object encodings To be written. Untill then the reader can refer to the excellent paper of Bruce, Cardelli, and Pierce [BCP97] Chapter 8 Calculi of objects (draft) We want to conclude just hinting at the fact that there is a new tendency in the type theoretic research on object oriented programming. Instead of trying to explain object oriented programming in term of existing constructs (e.g. records or overloaded ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Software (TACS), Sendai, Japan, 1997.

....types leads to the definition of a non conservative extension of the system. This problem is strictly related to the undecidability of subtyping in F [Pie94, Ghe93a] Other papers where second order recursive types are studied don t deal with the subtype checking problem in much depth. In [BCP97], a recursive extension of system F : Car90, CL91] is defined, but only as a tool to compare different models of object oriented languages. Hence, that paper doesn t consider the algorithmical aspects of the subtyping problem, and the defined system is far less powerful than our extension of ....

Kim B. Bruce, Luca Cardelli and Benjamin C. Pierce. Comparing object encodings. In Proceedings of Theoretical Aspects of Computer Software, Sendai, Japan, pages 415-438. Springer-Verlag, September 1997. LNCS 1281.

....Type class can be eliminated by explicit dictionary passing, done by the type checker in the front end. Lazy evaluation requires the use of FLINT primitives, delay and force, to make the evaluation explicit. Translating Java into FLINT is less trivial, but it boils down to what kind of encodings [5] we use to model the Java objects. We believe that FLINT is a sufficiently rich intermediate language that can be used to handle many interesting application languages. While building a new front end will not be completely trivial, it is definitely much easier than translating into C or building ....

K. B. Bruce, L. Cardelli, and B. C. Pierce. Comparing object encodings. In Proc. Third Workshop on Foundations of Object Oriented Languages, July 1996.

....and is not assigned a number in the decimal system of headings. The list should be set in small print and placed at the end of your contribution. An example is given at the end of this information sheet. For citations in the text please use square brackets and consecutive numbers: 1] [2], 3] 2.8 Page Numbering and Running Heads Your paper should show no printed page numbers; they are allocated by the volume editor. Indicate the ordering of your pages by numbering the sheets in pencil at the bottom of the reverse side. Do not set running heads. 2.9 Printing Quality For ....

Bruce, K.B., Cardelli, L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Takayasu, I. (eds.): Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol. 1281. Springer-Verlag, Berlin Heidelberg New York (1997) 415--438

....applications, it is also crucial to investigate the introduction of imperative features in our calculus. Finally we would like to find an equational theory dealing with the objects of Obj , and to study possible representation of our typing discipline in the framework theory of F : see [9]) ....

K. Bruce, L Cardelli, and C. B. Pierce. Comparing Object Encoding. In Proc. of TACS, Lecture Notes in Computer Science. Springer-Verlag, 1997.

....Work Our source language can encode a variety of constructs, including sums, source level existentials, lists, trees, etc. so we can already compile a fully featured programming language. Furthermore, it appears that adding refs, higher order type constructors, recursive types, or objects [10] in the style of Pierce and Turner [35] Bruce [9] and or Cardelli [11] is fairly straightforward. As we use a CPS based compiler, support for first class continuations and exceptions is also straightforward. However, extending TAL so that it can handle a stack based model where continuation ....

K. B. Bruce, L. Cardelli, and B. C. Pierce. Comparing object encodings. In Third Workshop on Foundations of Object Oriented Languages, July 1996.

....recursion through an object s self parameter explicit as an additional argument to each method leads to function types that are recursive in both covariant and contravariant positions, and hence permit no subtyping. More sophisticated type systems that can express the necessary subtyping exist [2, 5, 16], but these type systems require more complex encodings of objects and classes. Object calculi that keep self recursion implicit [1, 5] are more complex than record calculi and do not expose representations in a manner suitable for an intermediate language. Rather than devise an unwieldy IL and ....

....in both covariant and contravariant positions, and hence permit no subtyping. More sophisticated type systems that can express the necessary subtyping exist [2, 5, 16] but these type systems require more complex encodings of objects and classes. Object calculi that keep self recursion implicit [1, 5] are more complex than record calculi and do not expose representations in a manner suitable for an intermediate language. Rather than devise an unwieldy IL and translation, we take a more pragmatic approach. We assume that a Java program is first type checked by the Java type checker before it is ....

[Article contains additional citation context not shown here]

Bruce, K. B., Cardelli, L., and Pierce, B. C. Comparing Object Encodings. In Theoretical Aspects of Computer Software (TACS), Sendai, Japan (Sept. 1997).

.... be modelled by values of existential types; similarly, type theoretic research into the foundations of object oriented programming has made use of existential types, together with various combinations of function, record, and recursive types, to model objects and classes: see the recent paper by Bruce, Cardelli, and Pierce (1997) for a useful survey. To establish the properties of such type theoretic interpretations of information hiding requires a theory of semantic equivalence for elements of existential type. In this respect, the use of relations between types has proved very useful. Study of relational properties of ....

....in this paper. However, that language is limited in two important respects: it lacks recursively defined types and subtyping. These features are needed for applications to reasoning about object oriented programs for example, for proving properties of the object encodings considered by Bruce, Cardelli, and Pierce (1997). Recursively defined types provide a severe technical challenge for the method of operationally based logical relations, since one looses the ability to define the relation by induction on the structure of types. One way round this is to develop syntactical versions of the use of projections ....

Bruce, K. B., L. Cardelli, and B. C. Pierce (1997). Comparing object encodings.

....object from being first class. 10 However, in an impredicative type theory, Pierce and Turner s object encoding can be used quite satisfactorily. In a type theory that additionally supplies recursive types (with negative occurrences) a variety of other object encodings become possible as well [15, 13, 3, 31, 14]. Alternatively, the object encoding of Hickey [50] works entirely within the existing Nuprl type theory and shows promise of being extendable to a practical object system. 3.6 Conclusions I have shown how to give a type theoretic semantics to an expressive programming calculus that supports ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Software, Sendai, Japan, September 1997.

....suppose O is fj 1 : 1 ; n : n jg. Then O is interpreted as: O def = 9ff: ff f 1 : ff 1 ; n : ff n g I will refer to this encoding as the OEI encoding, for objects using existential and intersection types, following the terminology of Bruce et al. [7]. In the remainder of this paper, I will explore the expressiveness of this encoding by showing how it deals with various issues in object oriented programming. The OEI encoding will not prove to be sufficient to support self types or method update, but in Section 3 I will introduce a similar ....

....[5, 12] 4 Comparisons The OREI encoding discussed in this paper is closely related to three other abstraction oriented object encodings: the OE encoding of Pierce, Turner and Hoffman [22, 16] the ORE encoding of Bruce [6] and the ORBE encoding of Abadi, et al. discussed previously. Bruce et al. [7] cast each of these encodings in a common framework, and explore the interrelations between them. The OREI encoding and each of the three encodings from Bruce et al. encode object types in ways that appear fairly similar. Figure 2 summarizes the encodings of object types. However, these syntactic ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Theoretical Aspects of Computer Software, Sendai, Japan, September 1997.

No context found.

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Proceedings of TACS '97 (Theoretical Aspects of Computer Science), Springer LNCS 1281, pages 415--438, 1997.

....an explanation for the type checking rules for LOOM. Later in this paper we will generalize this semantics in order to derive type checking rules for an extension of LOOM with constructs similar to virtual types. The translation presented here is an extension of the ORE semantics presented in [BCP97] The translation differs from the one in that paper by supporting an imperative rather than a purely functional language, and by including a translation for classes as well as objects. 3.1 Objects Let I, M: Type Type represent the types of the records of instance variables and methods ....

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Proceedings of TACS '97 (Theoretical Aspects of Computer Science), Springer LNCS 1281, pages 415--438, 1997.

No context found.

Bruce, K.B., Cardelli, L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Ito, T. (eds.): Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol. 1281. Springer-Verlag, Berlin Heidelberg New York (1997) 415--438

No context found.

Bruce, K.B., Cardelli, L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Ito, T. (eds.): Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol. 1281. Springer-Verlag, Berlin Heidelberg New York (1997) 415--438

No context found.

K.B. Bruce, L. Cardelli, and B. Pierce. Comparing Object Encodings. In Proc. of TACS'97, volume 1281 of LNCS, pages 415--438. Springer-Verlag, 1997.

No context found.

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Proceedings of the Conference on Theoretical Aspects of Computer Software (TACS), 1997.

No context found.

Kim B. Bruce, Luca Cardelli, and Benjamin C. Pierce. Comparing object encodings. In Proceedings of the Conference on Theoretical Aspects of Computer Software (TACS), 1997.

No context found.

K.B. Bruce, L. Cardelli, B.C. Pierce, "Comparing object encodings", Proc. of TACS'97, LNCS 1281, 1997, 415-438.

No context found.

Bruce, K.C., Cardelli, L., , Pierce, B.C.: Comparing object encodings. In: Theoretical Aspects of Computer Software. Volume 1281 of Lecture Notes in Computer Science. Springer, Berlin, Heidelberg, New York (1997) 415--438

No context found.

Bruce, K.B., Cardelli, L., Pierce, B.C.: Comparing Object Encodings. In: Abadi, M., Ito, T. (eds.): Theoretical Aspects of Computer Software. Lecture Notes in Computer Science, Vol. 1281. Springer-Verlag, Berlin Heidelberg New York (1997) 415--438

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC