serp.bytecode.package.html Maven / Gradle / Ivy
Bytecode Manipuation
This package contains a framework for Java bytecode manipulation.
Bytecode manipulation is a powerful tool in the arsenal of the Java
developer. It can be used for tasks from compiling alternative
programming languages to run in a JVM, to creating new classes on the
fly at runtime, to instrumenting classes for performance analysis, to
debugging, to altering or enhancing the capabilities of existing
compiled classes. Traditionally, however, this power has come at a
price: modifying bytecode has required an in-depth knowledge of the
class file structure and has necessitated very low-level programming
techniques. These costs have proven too much for most developers, and
bytecode manipulation has been largely ignored by the mainstream.
The goal of the serp bytecode framework is to tap the full power of
bytecode modification while lowering its associated costs.
The framework provides a set of high-level APIs for manipulating all
aspects of bytecode, from large-scale structures like class member
fields to the individual instructions that comprise the code of
methods. While in order to perform any advanced manipulation, some
understanding of the
class file format and especially of the
JVM instruction set is necessary, the framework makes it as easy
as possible to enter the world of bytecode development.
There are several other excellent bytecode frameworks available. Serp
excels, however, in the following areas:
-
Ease of use. Serp provides very high-level APIs for
all normal bytecode modification functionality. Additionally,
the framework contains a large set of convenience methods to
make code that uses it as clean as possible. From overloading
its methods to prevent you from having to make type
conversions, to making shortcuts for operations like adding
default constructors, serp tries to take the pain out of
bytecode development.
-
Power. Serp does not hide any of the power of
bytecode manipulation behind a limited set of high-level
functions. In addition to its available high-level APIs, which
themselves cover the functionality all but the most advanced
users will ever need, serp gives you direct access to the
low-level details of the class file and constant pool. You
can even switch back and
forth between low-level and high-level operations;
serp maintains complete consistency of the class structure
at all times. A change to a method descriptor in the constant
pool, for example, will immediately change the return values
of all the high-level APIs that describe that method.
-
Constant pool management. In the class file format,
all constant values are stored in a constant pool of shared
entries to minimize the size of class structures. Serp gives
you access to the constant pool directly, but most of you
will never use it; serp's high-level APIs completely
abstract management of the constant pool.
Any time a new constant is needed, serp will automatically add
it to the pool while ensuring that no duplicates ever exist.
Serp also does its best to manipulate the pool so that the
effects of changing a constant are as expected: i.e. changing
one instruction to use the string "bar" instead of "foo"
will not affect other instructions that use the string "foo",
but changing the name of a class field will instantly change
all instructions that reference that field to use the new name.
-
Instruction morphing. Dealing with the individual
instructions that make up method code is the most difficult
part of bytecode manipulation. To facilitate this process,
most serp instruction representations have the ability to
change their underlying low-level opcodes on the fly as the
you modify the parameters of the instruction. For
example, accessing the constant integer value 0 requires the
opcode
iconst0, while accessing the string
constant "foo" requires a different opcode, ldc,
followed by the constant pool index of "foo". In serp, however,
there is only one instruction, constant. This
instruction has setValue methods which use the
given value to automatically determine the correct opcodes and
arguments -- iconst0 for a value of 0 and
ldc plus the proper constant pool index for the
value of "foo".
Serp is not ideally suited to all applications. Here are a few
disadvantages of serp:
-
Speed. Serp is not built for speed. Though there
are plans for performing incremental parsing, serp currently
fully parses class files when a class is loaded, which is a
slow process.
Also, serp's insistence on full-time consistency between the
low and high level class structures slows down both access and
mutator methods. These factors are less of a concern, though,
when creating new classes at runtime (rather than modifying
existing code), or when using serp as part of the compilation
process. Serp excels in both of these scenarios.
-
Memory. Serp's high-level structures for representing
class bytecode are very memory-hungry.
-
Multi-threaded modifications. The serp toolkit is
not threadsafe. Multiple threads cannot safely make
modifications to the same classes the same time.
-
Project-level modifications. Changes made in one
class in a serp project are not yet automatically propogated
to other classes. However, there are plans to implement this,
as well as plans to allow operations to modify bytecode based
on specified patterns, similar to aspect-oriented programming.
The first class that you should study in this package is the
{@link serp.bytecode.Project} type. From there, move onto the
{@link serp.bytecode.BCClass}, and trace its APIs into
{@link serp.bytecode.BCField}s, {@link serp.bytecode.BCMethod}s,
and finally into actual {@link serp.bytecode.Code}.