Performing code completion using data flow
A corpus of source code from a code database is accessed and a language prediction model is trained based on the corpus of source code. A given program is accessed and a completion of a given line of the given program is predicted by performing inferencing using the language prediction model and at least a portion of the given program. The given line is completed based upon the prediction.
1. A method for generating computer code, the method comprising:
accessing, using a hardware processor, a corpus of source code from a code database;
training, using the hardware processor, a language prediction model based on the corpus of source code, wherein the training of the language prediction model is performed by inputting a trimmed data flow graph;
accessing, using the hardware processor, a given program;
predicting, using the hardware processor, a completion of a given line of the given program by performing inferencing using the language prediction model and at least a portion of the given program; and
completing, using the hardware processor, the given line based upon the prediction.
2. The method of claim 1 , wherein the training of the language prediction model is performed by inputting a trimmed node.
3. The method of claim 1 , wherein the predicting of the completion of the given line further comprises converting a representation of a code snippet to an actual code snippet that conforms to rules of a given compiler.
4. The method of claim 1 , further comprising rendering the completed code into machine-executable form and running the machine-executable code.
5. The method of claim 1 , wherein the predicting operation predicts a plurality of completions of the given line and the method further comprising ranking the predicted completions.
6. The method of claim 1 , further comprising:
generating a data flow graph for each of a plurality of programs in the corpus of source code; and
trimming all edges to a node that represents a method invocation to be predicted, to generate the trimmed data flow graph.
7. The method of claim 6 , further comprising iteratively repeating the trimming operation to generate additional trimmed data flow graphs.
8. The method of claim 6 , wherein the training operation is performed by inputting the trimmed data flow graph and the trimmed node.
9. The method of claim 6 , further comprising generating an abstraction of the generated data flow graph, wherein the trimming of the edges is performed on the abstraction of the generated data flow graph and wherein the training is based on the trimmed abstracted data flow graph.
10. A computer readable medium comprising computer executable instructions which when executed by a computer cause the computer to perform the method of:
accessing a corpus of source code from a code database;
training a language prediction model based on the corpus of source code, wherein the training of the language prediction model is performed by inputting a trimmed data flow graph;
accessing a given program;
predicting a completion of a given line of the given program by performing inferencing using the language prediction model and at least a portion of the given program; and
completing the given line based upon the prediction.
11. The computer readable medium of claim 10 , the method further comprising:
generating a data flow graph for each of a plurality of programs in the corpus of source code; and
trimming all edges to a node that represents a method invocation to be predicted to generate the trimmed data flow graph, wherein the training operation is performed by inputting the trimmed data flow graph and the trimmed node.
12. An apparatus comprising:
a memory; and
at least one processor, coupled to said memory, and operative to perform operations comprising:
accessing a corpus of source code from a code database;
training a language prediction model based on the corpus of source code, wherein the training of the language prediction model is performed by inputting a trimmed data flow graph;
accessing a given program;
predicting a completion of a given line of the given program by performing inferencing using the language prediction model and at least a portion of the given program; and
completing the given line based upon the prediction.
13. The apparatus of claim 12 , wherein the at least one processor is operative to train the language prediction model by inputting a trimmed node.
14. The apparatus of claim 12 , wherein the at least one processor is operative to predict the completion of the given line by converting a representation of a code snippet to an actual code snippet that conforms to rules of a given compiler.
15. The apparatus of claim 12 , wherein the at least one processor is further operative to render the completed code into machine-executable form and run the machine-executable code.
16. The apparatus of claim 12 , wherein the at least one processor is operative to predict a plurality of completions of the given line and wherein the at least one processor is further operative to rank the predicted completions.
17. The apparatus of claim 12 , wherein the at least one processor is further operative to:
generate a data flow graph for each of a plurality of programs in the corpus of source code; and
trim all edges to a node that represents a method invocation to be predicted to generate the trimmed data flow graph.
18. The apparatus of claim 17 , wherein the at least one processor is further operative to iteratively repeat the trimming operation to generate additional trimmed data flow graphs.
19. The apparatus of claim 17 , wherein the at least one processor is operative to train the language prediction model by inputting the trimmed data flow graph and the trimmed node.
20. The apparatus of claim 17 , wherein the at least one processor is further operative to generate an abstraction of the generated data flow graph, and wherein the at least one processor is operative to trim the edges by performing the trimming on the abstraction of the generated data flow graph and to train the language prediction model based on the trimmed abstracted data flow graph.