US20100281470A1

US20100281470A1 - Method for generating interpretable code for storage in a device having limited storage

Info

Publication number: US20100281470A1
Application number: US12/838,542
Authority: US
Inventors: Gregory R. Bentz; John F.A. Dahms; David C. Yach
Original assignee: Research in Motion Ltd
Current assignee: Malikie Innovations Ltd
Priority date: 2002-11-29
Filing date: 2010-07-19
Publication date: 2010-11-04
Also published as: CA2507371C; US7761861B2; AU2002347146A1; EP1909174A1; ES2305316T3; CA2507371A1; DE60226786D1; CN1695117A; US20060020932A1; DE60239024D1; WO2004051468A1; EP1570347B1; HK1081294A1; EP1909174B1; ATE496330T1; EP1570347A1; ATE396452T1; CN100395703C

Abstract

In some embodiments of the present invention, files are generated from .java files or .class files or .jar files. The generated files are directly linkable and interpretable by a Java Virtual Machine. The generated files may be stored in a directly addressable memory of a device. References between .class files may appear in the generated files as hard offsets or symbolic references. The generated files may be grouped so that cross-references between generated files in the same group appear as hard offsets.

Description

CROSS REFERENCE TO RELATED APPLICATIONS

This application is a continuation of U.S. patent application Ser. No. 10/536,745 filed on May 27, 2005, which was the National Stage of International Application No. PCT/CA2002/001841 filed Nov. 29, 2002. The entirety of U.S. patent application Ser. No. 10/536,745 and International Application No. PCT/CA2002/001841 are hereby incorporated by reference.

BACKGROUND OF THE INVENTION

Java source code files (.java files) are compiled into .class files by the Java compiler. These .class files may be read into memory by the Java Virtual Machine (VM) in an in-memory format that is suitable for interpretation by the VM. The .class files are then linked with other .class files that have been similarly read. References between .class files are resolved symbolically, using character strings. These character strings appear in both the .class file that contains the definition and in the .class file that references the definition. Therefore, the presence of references between .class files may increase the size of the .class files.
Java .class files may be archived (and optionally compressed) into a .jar file. However, .jar files are not directly interpretable by the Java VM, and the .class files must be extracted (and decompressed, if applicable) from the .jar file (and read into memory) in order for them to be linked, resolved and interpreted by the Java VM.
Although .jar files comprising archived and compressed .class files are smaller than the .class files themselves (and are therefore more suitable for transmission between communication devices), storage space for the extracted (and decompressed, if applicable).class files needs to be available in the environment where the application is to be executed, so that the Java VM may access the .class files. Consequently, a solution involving .jar files may not represent a savings in storage space. In fact, a solution involving .jar files may require extra storage space, for both the .jar files and the extracted .class files. The extracted .class files need not be retained once they have been loaded into memory, linked and resolved. However, both the .jar file and the in-memory representations of the .class files must be retained. In an environment having limited storage, where storage space is at a premium, it may therefore be preferable to store only the .class files and not to use a solution involving .jar files. However, as explained above, the size of .class files increases when the number of references between .class files increases.
Therefore, it would be beneficial to generate directly interpretable files that are of a smaller size than .class files, while providing a solution for references between .class files.

BRIEF DESCRIPTION OF THE DRAWINGS

The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, both as to organization and method of operation, together with objects, features and advantages thereof, may best be understood by reference to the following detailed description when read with the accompanied drawings in which:

FIGS. 1A and 1B are simplified prior art illustrations of six exemplary .class files;

FIGS. 2A and 2B are simplified illustrations of exemplary .cod files, according to some embodiments of the present invention;

FIG. 3 is a flowchart illustration of a method for generating .cod files, according to some embodiments of the present invention; and

FIG. 4 is a simplified block-diagram illustration of a device having a computing unit and directly addressable memory, according to some embodiments of the present invention.

It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.

DETAILED DESCRIPTION OF THE INVENTION

In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However it will be understood by those of ordinary skill in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures and components have not been described in detail so as not to obscure the present invention.
Java source code files (.java files) are compiled into .class files by the Java compiler. These .class files, once read into memory, may be linked, resolved and interpreted by the Java Virtual Machine (VM). FIGS. 1A and 1B are simplified prior art illustrations of six exemplary .class files—Aa.class file 100A, Bb.class file 100B, Cc.class file 100C, Dd.class file 100D, Ee.class file 100E, and Ff.class file 100F, collectively referred to as file 100.
The structure of the .class file is well-documented and will not be described here in detail. File 100 is illustrated in a simplified format for clarity of explanation, however it will be understood by persons of ordinary skill in the art that this simplified format is not an exact representation of the structure of file 100. Each file 100 comprises a constant pool 102. Constant pool 102 comprises indexed cp_info entries, some of which comprise character strings indicating the names of the class and parent class, the names of methods, the type of methods, the names of fields, the type of fields, etc. referred to within the ClassFile structure. Each file 100 also comprises “byte codes and information structures” 104 regarding the class properties, the methods, fields and attributes of the class, and their types. These structures may point to entries in constant pool 102 using ordinal index numbers.
In Aa.class file 100A, Method G includes a reference to another method, Method K, found in class Bb. In fact, Method G is defined in the .class file as a method_info structure comprising, among other things, indices of entries in constant pool 102A. However, for clarity of description, the definition of Method G and its reference to Method K are shown in FIG. 1A in a simplified format. Therefore, the character strings “K” and “Bb” are included in constant pool 102A. The inclusion of “K” and “Bb” in constant pool 102 A actually appear as part of several cp_info structures. However, for clarity of description, only the character strings “K” and “Bb” are shown in FIG. 1A. Since the argument of Method K is of type integer, the BaseType character “I” is included in constant pool 102A. The symbolic reference to Method K by Method G in the byte codes and information structures 104A comprises the indices of the relevant cp_info structures in constant pool 102A.
Similarly, in Bb.class file 100B, Method K is defined in class Bb, so constant pool 102B includes all the strings “Bb”, “K” and the BaseType character “I” (representing the ‘int’ type).
FIGS. 2A and 2B are simplified illustrations of exemplary .cod files, the generation of which will be described hereinbelow with respect to FIG. 3. .cod files are directly interpretable files that comprise the information of .class files in a manner that requires less storage space than the .class files themselves. .cod files may be “solo” .cod files or “sibling” .cod files, as will be explained in greater detail hereinbelow.
In the example shown in FIG. 2B, solo .cod file 220 is named “EeFf.cod”, to signify that it results from combining the files Ee.class and Ff.class, although the scope of the present invention is not limited in this respect. Solo EeFf.cod file 220 comprises a constant pool 222, one or more fixup tables 228, and “byte codes and information structures” 224 regarding the class properties, the methods, fields and attributes of the class, and their types.
FIG. 3 is a simplified flowchart illustration of a method for the generation of .cod files from a collection of .java files or .class files. The method may be performed by executable software on a general-purpose computer, the software being provided, for example, as part of a software developer's kit (SDK).
A list of .java files or .class files is received (step 300). If the input is .java files, then a Java compiler is applied to the .java files to produce corresponding .class files (step 302). Alternatively, although this is not shown in FIG. 3, one or more .jar files could be received, and information regarding the archived .class files could be extracted therefrom.
The executable software identifies common entries in the constant pools of the .class files (step 304). For example, in FIG. 1B, the parent class name character string “java/lang/Object”, the method name character string “T” and the BaseType character “C” (representing the ‘char’ type) are common to Ee.class file 100E and to Ff.class file 100F.
The executable software identifies cross-references between classes, methods, and fields in the .class files (step 306). For example, in FIG. 1B, Method Ee.T is referenced by Method Ff.U.
The executable software then generates the .cod file by combining elements of the .class files (step 308). Unlike standard compression algorithms that compress data regardless of its content, the combination of elements of the .class files into the .cod file is performed differently based upon the type of element and its content. For example, the .cod file comprises a constant pool whose entries include the information stored in the constant pools of the .class files, but without duplications of redundant entries. In the generated .cod file, the constant pool contains only a single copy of the common entries identified in step 304. For example, in FIG. 2B, constant pool 222 of solo EeFf.cod file comprises only one instance of character strings “java/lang/Object” and “T” and one instance of BaseType character “C”.
The executable software uses hard offsets in the generated solo .cod file for cross-references between classes and methods defined in the .class files being combined. For example, in FIG. 2B, the reference to Method Ee.T by Method Ff.U in the “byte codes and information structures” 224 comprises a hard offset, HOEe.T, specifying the location of the definition of Method Ee.T within solo EeFf.cod file 220. This hard offset does not need to be resolved or put into context by the Java VM at link time.
The above-described use of a hard offset in a .cod file may be contrasted with the use of offsets in Windows™ .DLL files. References in Windows™ .DLL files may be in terms of a symbolic name or an ordinal number. The ordinal number is used as an index into a table of offsets. Embodiments of the present invention use the hard offsets directly in the .cod file, giving a more compact representation. Moreover, Windows™ .DLL files are used for compiled code, while the .cod files are used for interpreted Java.
The executable software uses symbolic references in the generated solo .cod file for cross-references between the classes received in step 300 and other classes. In the present example, class Ee and class Ff extend java.lang.Object. Therefore, constant pools 102E, 102F and 222 each comprise a single instance of class name “java/lang/Object” so that definitions of the classes may refer to this parent class. In solo EeFf.cod file 220, constant pool 222 comprises the string “java/lang/Object”, and the reference to the java.lang.Object class in the definitions of the classes in “byte codes and information structures” 224 is a symbolic reference (using the index of the string in constant pool 222) that needs to be resolved by the Java VM at link time using information stored in fixup table 228.
The executable software may perform additional actions that have not been illustrated in FIG. 3. For example, other space-saving techniques may be used to further reduce the size of the solo .cod files.
It will be appreciated by persons of ordinary skill in the art that when either of the source code files Ee.java and Ff.java are modified, it is necessary to invoke the executable software to generate a new EeFf.cod file, or alternatively to invoke the executable software on the modified .java or .class files along with one or more additional Java or .class files to generate a new EeFfxxx.cod file, where “x” indicates the additional classes.
It will also be appreciated by persons of ordinary skill in the art that if it were possible to generate a single solo .cod file for all the .class files to be used in a particular application, then the Java VM would not need to resolve symbolic references between classes, methods and fields, since all such references would appear in the .cod file as hard offsets. Moreover, such a .cod file might be significantly smaller than the total size of the .class files, since duplications of information in the .class files would be eliminated in the .cod file. Such a single .cod file would also be smaller than the multiple solo .cod files due to the reduction in duplicated information.
However, unlike Windows™ .DLL files, which are relatively unlimited in size, there are sometimes significant limitations on the size of a single .cod file. For example, FIG. 4 shows a device 400 comprising a computing unit 402 able to execute a Java VM and a directly addressable memory 404 having one or more .cod files 406 stored therein. Although the scope of the present invention is not limited in this respect, computing unit 402 may be a general-purpose microprocessor or any computing unit capable of executing a Java VM, and device 400 may be a digital camera, a handheld video recorder, a cellular telephone, a personal digital assistant, a handheld computer, an MP3 player, a CD player, a handheld gaming device, a pager, a two-way pager, etc. Although the scope of the present invention is not limited in this respect, directly addressable memory 404 may be not-OR (NOR) flash memory whose physical layout imposes an upper limit of 64 kiloBytes for the size of each .cod file stored therein.
If an application were to be stored exclusively as solo .cod files in a storage medium that imposes a limit on the size of individual .cod files, then each .cod file would likely comprise many symbolic references to account for the cross-references between classes in that .cod file and classes in other solo .cod files. Since symbolic references require more storage space in the .cod file than hard offsets, the overall size of the solo .cod representation of the application may be quite large.
As an alternative to the exclusive use of solo .cod files in the representation of an application, sibling .cod files may be used when the .cod files are to be stored in a storage medium that imposes a limit on the size of individual .cod files. The software developer may group together .java files or .class files into sibling groups. This grouping may be based on the software developer's knowledge of the intricacy of cross-references between the classes represented in the files. A .cod file may be generated from these .java files or .class files, and if, after a portion of the .java files or .class files have been combined into a .cod file, the size of the .cod file would exceed a predetermined limit if another .java or .class file were to be combined into the .cod file, then one or more sibling .cod files are created for the remaining .java files or .class files in the sibling group. Classes are not split across .cod file boundaries.
It will be appreciated by persons of ordinary skill in the art that for a given set of .class files, packing the .class files into two sibling .cod files will be a more compact representation than packing the .class files into three sibling .cod files. Therefore, it is desirable to pack the .class files in the group into as few .cod files as possible, while maintaining the constraint of individual .cod file size. The problem of how to divide a set of cross-referencing classes among sibling .cod files is similar to the well-known “packing problem” in the art of software engineering. Various techniques to minimize the overall size of the .cod representation of an application may be used.
FIG. 2A is a simplified illustration of two exemplary sibling .cod files according to some embodiments of the present invention. A sibling .cod file 200 is named “AaBb.cod” to signify that it results from combining the files Aa.class and Bb.class, although the scope of the present invention is not limited in this respect. Similarly, a sibling .cod file 210 is named “CcDd.cod” to signify that it results from combining the files Cc.class and Dd.class.
Each sibling .cod file comprises a list of its siblings. For example, a sibling list 206 of AaBb.cod file 200 comprises “CcDd” to indicate that CcDd.cod file 210 is a sibling of AaBb.cod file 200. Similarly, a sibling list 216 of CcDd.cod file 210 comprises “AaBb” to indicate that AaBb.cod file 200 is a sibling of CcDd.cod file 210.
Sibling AaBb.cod file 200 also comprises a constant pool 202, “byte codes and information structures” 204, and one or more fixup tables 208. Similarly, sibling CcDd.cod file 210 also comprises a constant pool 212, “byte codes and information structures” 214, and one or more fixup tables 218. The fixup tables 208, 218 may include indications of the places in the .cod file where resolution work is to be done by the Java VM at link time. The fixup tables may include pointers to the name of another .cod file, if necessary to the name of the class containing the symbol, if necessary to the name of the method or field within that class being referenced, and to method or field typing information.
FIG. 3 is a simplified flowchart illustration of a method for the generation of .cod files from a collection of .java files or .class files. The method may be performed by executable software on a general-purpose computer, the software being provided, for example, as part of a software developer's kit (SDK).
A list of .java files or .class files is received (step 300). If the input is .java files, then a Java compiler is applied to the .java files to produce corresponding .class files (step 302). Alternatively, although this is not shown in FIG. 3, one or more .jar files could be received, and information regarding the archived .class files could be extracted therefrom.
The executable software identifies common entries in the constant pools of the .class files (step 304).
The executable software identifies cross-references between classes, methods, and fields in the .class files (step 306).
The executable software then generates sibling .cod files by combining elements of the .class files (step 308). In the generated sibling .cod file, the constant pool contains only a single copy of the common entries identified in step 304. For example, in FIG. 2A, constant pool 202 of sibling AaBb.cod file 200 comprises only one instance of character strings “Bb” and “K” and one instance of BaseType character “I”. In another example, constant pool 212 of sibling CcDd.cod file 210 comprises only one instance of character strings “Cc” and “P” and one instance of BaseType character “I”.
The executable software uses hard offsets in the generated sibling .cod file for cross-references between classes whose .cod files are in the same sibling group. The executable software uses symbolic references in the generated sibling .cod file for cross-references between classes whose .cod files belong to different sibling groups, and between classes having solo .cod files.
This is better understood by way of examples:
1) As shown in FIG. 1A, Method Aa.G comprises a reference to Method Bb.K. Therefore, as shown in FIG. 2A, this reference appears in the definition of Method Aa.G in “byte codes and information structures” 204 of sibling AaBb.cod file 200 as a hard offset, HOBb.K, specifying the location of the definition of Method Bb.K within sibling AaBb.cod file 200. This hard offset does not need to be resolved or put into context by the Java VM at link time.
2) Similarly, as shown in FIG. 1B, Method Dd.Q comprises a reference to Method Cc.P. Therefore, as shown in FIG. 2A, this reference appears in the definition of Method Dd.Q in “byte codes and information structures” 214 of sibling CcDd.cod file 210 as a hard offset, HOCc.P, specifying the location of the definition of Method Cc.P within sibling CcDd.cod file 210. This hard offset does not need to be resolved or put into context by the Java VM at link time.
3) As shown in FIG. 1A, Method Bb.L comprises a reference to Method Cc.N. Therefore, as shown in FIG. 2A, this reference appears in the definition of Method Bb.L in “byte codes and information structures” 204 of sibling AaBb.cod file 200 as a hard offset, HOCc.N, specifying the location of the definition of Method Cc.N within sibling CcDd.cod file 210. This hard offset needs to be placed into context by the Java VM at link time using information stored in fixup table 208. For example, fixup table 208 may comprise a pointer to the name of sibling CcDd.cod file along with the location of the reference in Method Bb.L to Method Cc.N so that the Java VM knows which sibling .cod file contains Method Cc. N at the specified hard offset.
4) As shown in FIGS. 1A and 1B, Method Cc.M comprises a reference to Method Ff.U. The .cod file for class Ff is outside of the sibling group of CcDd.cod file 210. Therefore, as shown in FIG. 2A, this reference appears in the definition of Method Cc.M in “byte codes and information structures” 214 of sibling CcDd.cod file 210 as a symbolic reference to Method Ff.U and its argument(s). This symbolic reference needs to be resolved by the Java VM at link time using information stored in fixup table 218. For example, fixup table 218 may comprise a pointer to the name of the .cod file for the class where Method Ff.U is defined (e.g. “EeFf.cod”), a pointer to the name of the class where Method Ff.U is defined (e.g. “Ff”), a pointer to the name of the method (e.g. “U”) and a pointer to the name of the method typing information (e.g. “I”).
It will be appreciated by persons of ordinary skill in the art that when either of the source code files Cc.java and Dd.java are modified, it is necessary to invoke the executable software to generate a new CcDd.cod file, or alternatively to invoke the executable software on the modified .java or .class files along with one or more additional .java or .class files to generate a new CcDdxxx.cod file, where “x” indicates the additional classes. If a new CcDd.cod file or a new CcDdxxx.cod file has been generated, then the hard offset HOCc.N appearing in resolved AaBb.cod file 200 no longer accurately describes the location of the byte codes and other information for Method N in the new CcDd.cod file or a new CcDdxxx.cod file.
Accordingly, when a .cod file is updated, all of its sibling .cod files are updated simultaneously. Although the scope of the present invention is not limited in this respect, this is generally done by taking the entire set of .class files and regenerating all the output sibling .cod files. Since some of the classes have changed, the division of classes among the newly generated sibling .cod files may be different from that in the original sibling .cod files.
While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents will now occur to those of ordinary skill in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

Claims

1. A device comprising:

a computing unit configured to execute a Java Virtual Machine; and

a memory unit directly addressable by the computing unit, the memory storing a plurality of sibling files to be linked and interpreted by the Java Virtual Machine, wherein the plurality of sibling files contain at least one class selected from a plurality of class files, wherein the plurality of sibling files comprises fewer files than the plurality of class files, and wherein each of the plurality of sibling files comprises:

a constant pool containing entries from two or more of the plurality of class files without duplication of entries;

a byte code and information structure having combined byte code and information structure entries from the two or more class files; and

a fixup table for providing information to the Java Virtual Machine to resolve at least one entry in the sibling file at link time.

2. The device of claim 1, wherein at least two of the plurality of sibling files belong to a common sibling group, and the at least two sibling files further comprise a sibling list that lists other sibling files in the common sibling group.

3. The device of claim 2, wherein constant pool entries that are cross-references between sibling files in the common sibling group are indicated by hard offsets, and wherein constant pool entries that are references to files that are not part of the common sibling group are indicated using symbolic references.

4. The device of claim 3, wherein the fixup table provides the location of data needed by the Java Virtual Machine to resolve a symbolic reference in at least one of the plurality of sibling files.

5. The device of claim 3, wherein the fixup table provides the location of data of a cross-referenced sibling file needed by the Java Virtual Machine to place one of the hard offsets that corresponds to the cross-referenced sibling file into context at link time.

6. The device of claim 3, wherein at least one of the hard offsets does not need to be resolved or put into context by the Java Virtual Machine at link time.

7. The device of claim 1, wherein the memory imposes an upper limit on the size of each of the sibling files stored therein.

8. A computer-readable storage medium for storing a plurality of sibling files to be linked and interpreted by a Java Virtual Machine, wherein the plurality of sibling files contain at least one class selected from a plurality of class files, wherein the plurality of sibling files comprises fewer files than the plurality of class files, and wherein each of the plurality of sibling files comprises:

9. The computer-readable storage medium of claim 8, wherein at least two of the plurality of sibling files belong to a common sibling group, and the at least two sibling files further comprise a sibling list that lists other sibling files in the common sibling group.

10. The computer-readable storage medium of claim 9, wherein constant pool entries that are cross-references between sibling files in the common sibling group are indicated by hard offsets, and wherein constant pool entries that are references to files that are not part of the common sibling group are indicated using symbolic references.

11. The computer-readable storage medium of claim 10, wherein the fixup table provides the location of data needed by the Java Virtual Machine to resolve a symbolic reference in at least one of the plurality of sibling files.

12. The computer-readable storage medium of claim 10, wherein the fixup table provides the location of data of a cross-referenced sibling file needed by the Java Virtual Machine to place one of the hard offsets that corresponds to the cross-referenced sibling file into context at link time.

13. The computer-readable storage medium of claim 10, wherein at least one of the hard offsets does not need to be resolved or put into context by the Java Virtual Machine at link time.

14. The computer-readable storage medium of claim 8, wherein an upper limit on the size of each of the sibling files is imposed.

15. A memory device comprising:

the memory device storing a plurality of sibling files to be linked and interpreted by a Java Virtual Machine, wherein the plurality of sibling files contain at least one class selected from a plurality of class files, wherein the plurality of sibling files comprises fewer files than the plurality of class files, and wherein each of the plurality of sibling files comprises:

16. The device of claim 15, wherein at least two of the plurality of sibling files belong to a common sibling group, and the at least two sibling files further comprise a sibling list that lists other sibling files in the common sibling group.

17. The device of claim 16, wherein constant pool entries that are cross-references between sibling files in the common sibling group are indicated by hard offsets, and wherein constant pool entries that are references to files that are not part of the common sibling group are indicated using symbolic references.

18. The device of claim 17, wherein the fixup table provides the location of data needed by the Java Virtual Machine to resolve a symbolic reference in at least one of the plurality of sibling files.

19. The device of claim 17, wherein the fixup table provides the location of data of a cross-referenced sibling file needed by the Java Virtual Machine to place one of the hard offsets that corresponds to the cross-referenced sibling file into context at link time.

20. The device of claim 17, wherein at least one of the hard offsets does not need to be resolved or put into context by the Java Virtual Machine at link time.

21. The device of claim 15, wherein the memory device imposes an upper limit on the size of each of the sibling files stored therein.