Transforming a Java program using a symbolic description language model
Techniques for transforming Java source code using a symbolic description language are disclosed, including: obtaining a set of Java source code corresponding to a Java program; generating a symbolic description language (SDL) model of the set of Java source code; generating, based on the SDL model, a transformed program including at least one transformation relative to the Java program.
1 . One or more non-transitory computer-readable media storing instructions that, when executed by one or more processors, cause performance of operations comprising:
obtaining a set of Java source code corresponding to a Java program;
generating a symbolic description language (SDL) model of the set of Java source code, wherein the SDL model represents the set of Java source code using a schema comprising:
an operation comprising a name, zero or more operands, an operation result, zero or more attributes, and a body;
the body comprising one or more blocks; and
each block in the one or more blocks comprising, respectively, one or more respective operations;
generating, based on the SDL model, a transformed program comprising at least one transformation relative to the Java program, wherein generating the transformed program comprises one or more of:
a) iterating, by a machine learning model, over transformations of the SDL model until a threshold condition is satisfied; or
b) compiling, by a compiler configured to generate executable code based on SDL models, the transformed program; and
performing, at runtime by a runtime environment configured to execute the transformed program, a reflection operation using at least part of the SDL model.
2 . The one or more non-transitory computer-readable media of claim 1 , wherein generating the transformed program comprises:
generating, based on the SDL model, a differentiated SDL model;
generating, based on the differentiated SDL model, a differentiated program.
3 . The one or more non-transitory computer-readable media of claim 1 , wherein generating the transformed program comprises:
generating, based on the SDL model, an optimized SDL model;
generating, based on the optimized SDL model, an optimized program.
4 . The one or more non-transitory computer-readable media of claim 1 , wherein generating the transformed program comprises:
generating, based on the SDL model, a program that (a) performs one or more functions defined in the set of Java source code and (b) is written in a language other than Java.
5 . The one or more non-transitory computer-readable media of claim 4 , wherein the language other than Java is a domain-specific language.
6 . The one or more non-transitory computer-readable media of claim 1 , the operations further comprising:
executing the transformed program.
7 . A system comprising: at least one device comprising one or more hardware processors, the system being configured to perform operations comprising:
obtaining a set of Java source code corresponding to a Java program;
generating a symbolic description language (SDL) model of the set of Java source code, wherein the SDL model represents the set of Java source code using a schema comprising:
an operation comprising a name, zero or more operands, an operation result, zero or more attributes, and a body;
the body comprising one or more blocks; and
each block in the one or more blocks comprising, respectively, one or more respective operations;
generating, based on the SDL model, a transformed program comprising at least one transformation relative to the Java program, wherein generating the transformed program comprises one or more of:
a) iterating, by a machine learning model, over transformations of the SDL model until a threshold condition is satisfied; or
b) compiling, by a compiler configured to generate executable code based on SDL models, the transformed program; and
performing, at runtime by a runtime environment configured to execute the transformed program, a reflection operation using at least part of the SDL model.
8 . The system of claim 7 , wherein generating the transformed program comprises:
generating, based on the SDL model, a differentiated SDL model;
generating, based on the differentiated SDL model, a differentiated program.
9 . The system of claim 7 , wherein generating the transformed program comprises:
generating, based on the SDL model, an optimized SDL model;
generating, based on the optimized SDL model, an optimized program.
10 . The system of claim 7 , wherein generating the transformed program comprises: generating, based on the SDL model, a program that (a) performs one or more functions defined in the set of Java source code and (b) is written in a language other than Java.
11 . A method comprising:
obtaining a set of Java source code corresponding to a Java program;
generating a symbolic description language (SDL) model of the set of Java source code, wherein the SDL model represents the set of Java source code using a schema comprising:
an operation comprising a name, zero or more operands, an operation result, zero or more attributes, and a body;
the body comprising one or more blocks; and
each block in the one or more blocks comprising, respectively, one or more respective operations;
generating, based on the SDL model, a transformed program comprising at least one transformation relative to the Java program, wherein generating the transformed program comprises one or more of:
a) iterating, by a machine learning model, over transformations of the SDL model until a threshold condition is satisfied; or
b) compiling, by a compiler configured to generate executable code based on SDL models, the transformed program;
performing, at runtime by a runtime environment configured to execute the transformed program, a reflection operation using at least part of the SDL model; and
wherein the method is performed by at least one device comprising one or more hardware processors.
12 . The method of claim 11 , wherein the transformed program is a transformed Java program.
13 . The method of claim 11 , wherein generating the transformed program comprises:
generating, based on the SDL model, a differentiated SDL model;
generating, based on the differentiated SDL model, a differentiated program.
14 . The method of claim 11 , wherein generating the transformed program comprises:
generating, based on the SDL model, an optimized SDL model;
generating, based on the optimized SDL model, an optimized program.
15 . The method of claim 11 , wherein generating the transformed program comprises:
generating, based on the SDL model, a program that (a) performs one or more functions defined in the set of Java source code and (b) is written in a language other than Java.
16 . The method of claim 15 , wherein the language other than Java is a domain-specific language.
17 . The method of claim 11 , wherein iterating, by the machine learning model, over transformations of the SDL model until the threshold condition is satisfied comprises:
executing, by the machine learning model, a gradient descent algorithm that starts with the SDL model and iterates over the transformations of the SDL model until the threshold condition is satisfied.
18 . The method of claim 11 , further comprising:
extracting an SDL representation of the machine learning model from the SDL model;
generating the machine learning model based on the SDL representation of the machine learning model.
19 . The method of claim 11 , wherein compiling the transformed program comprises:
generating a lowered SDL model, at least by transforming an SDL representation of a high-level operation into a plurality of SDL representations of lower-level operations;
compiling the lowered SDL model.