Utilities used by the Deep Program Understanding team
Clone or download


DPU Utilities

Build Status

This contains a set of utilities used across projects of the DPU team.


pip install dpu-utils

Overview of Utilities:

Generic Utilities:

  • dpu_utils.utils.RichPath a convenient way of using both paths and Azure paths in your code.
  • dpu_utils.utils.*Iterator iterator wrappers that can parallelize their iteration in other threads/processes.
  • dpu_utils.utils.{load,save}_json[l]_gz convenience methods for loading .json[l].gz from the filesystem.
  • dpu_utils.utils.git_tag_run that tags the current working directory git the state of the code.
  • dpu_utils.utils.run_and_debug when an exception happens, start a debug session. Usually a wrapper of __main__.
  • dpu_utils.utils.ChunkWriter that helps writing chunks to the output.

TensorFlow Utilities:

  • dpu_utils.tfutils.GradRatioLoggingOptimizer a wrapper around optimizers that logs the ratios of grad norms to parameter norms.
  • dpu_utils.tfutils.unsorted_segment_logsumexp
  • dpu_utils.tfutils.unsorted_segment_log_softmax
  • dpu_utils.tfutils.TFVariableSaver save TF variables in an object that can be pickled.

General Machine Learning Utilities:

  • dpu_utils.mlutils.CharTensorizer for character-level tensorization.
  • dpu_utils.mlutils.Vocabulary a str to int vocabulary for machine learning models

TensorFlow Models:

  • dpu_utils.tfmodels.SparseGGNN a sparse GGNN implementation.
  • dpu_utils.tfmodels.AsyncGGNN an asynchronous GGNN implementation.

Code-related Utilities

  • dpu_utils.codeutils.split_identifier_into_parts() split identifiers into subtokens on CamelCase and snake_case.
  • dpu_utils.codeutils.{Lattice, CSharpLattice} represent lattices and some useful operations in Python.
  • dpu_utils.codeutils.get_language_keywords() that retrieves the keyword tokens for many programming languages.
  • dpu_utils.codeutils.deduplication.DuplicateDetector that detects duplicates in codebases.

Command-line tools

Approximate Duplicate Code Detection

You can use the deduplicationcli command to detect duplicates in pre-processed source code, by invoking

 $ deduplicationcli DATA_PATH OUT_JSON

where DATA_PATH is a file containing tokenized .jsonl.gz files and OUT_JSON is the target output file. For more options look at --help.

An exact (but usually slower) version of this can be found here along with code to tokenize Java, C#, Python and JavaScript into the relevant formats.


Run the unit tests

python setup.py test

Generate code coverage reports

# pip install coverage
coverage run --source dpu_utils/ setup.py test && \
  coverage html

The resulting HTML file will be in htmlcov/index.html.


This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.