Code adaptation through deep learning

A hardware storage device having stored thereon computer executable instructions that are structured to be executable by a processor of a computing device to thereby cause the computing device to perform actions that:

• retrieve a plurality of source code snippets having an undefined variable name;
• extract an extended context from each of the plurality of source code snippets;
• replace the undefined variable name in each extended context of each of the plurality of source code snippets with a select one of a plurality of anonymized values;
• create a training dataset comprising a plurality of training samples, wherein a training sample comprises an extended context of a source code snippet with the undefined variable name replaced by the select one of the plurality of anonymized values;
• and
• train a deep learning model from the training dataset to learn to map a particular anonymized value to a defined variable name,
• wherein the trained deep learning model generates the defined variable name from a target source code program to replace the undefined variable name from an external source code snippet pasted into the target source code program.

A computer-implemented method comprising:

• collecting a plurality of source code snippets having an undefined variable name;
• extracting an extended context from each of the plurality of source code snippets;
• transforming the undefined variable name in each extended context of each of the plurality of source code snippets with a select one of a plurality of anonymized values;
• create a training dataset comprising a plurality of training samples, wherein a training sample comprises an extended context of a source code snippet with the undefined variable name replaced by the select one of the plurality of anonymized values;
• constructing a deep learning model with a plurality of decoder units, wherein each decoder unit is associated with a select one of the plurality of anonymized values;
• training the deep learning model from the training dataset to learn to map a particular anonymized value to a defined variable name, wherein each training sample is feed into each of the plurality of decoder units in parallel for a respective decoder unit to learn a mapping of the select one of the plurality of anonymized values to a corresponding defined variable name,
• wherein the trained deep learning model generates the defined variable name from a target source code program to replace the undefined variable name from an external source code snippet pasted into the target source code program.