metadata

license: cc-by-4.0
pretty_name: DyPyBench Functions
tags:
  - code
dataset_info:
  features:
    - name: nwo
      dtype: string
    - name: sha
      dtype: string
    - name: path
      dtype: string
    - name: identifier
      dtype: string
    - name: parameters
      dtype: string
    - name: return_statement
      dtype: string
    - name: docstring
      dtype: string
    - name: docstring_summary
      dtype: string
    - name: func_begin
      dtype: int64
    - name: func_end
      dtype: int64
    - name: function
      dtype: string
    - name: url
      dtype: string
    - name: project
      dtype: int64
    - name: executed_lines
      sequence: int64
    - name: executed_lines_pc
      dtype: float64
    - name: missing_lines
      sequence: int64
    - name: missing_lines_pc
      dtype: float64
    - name: covered
      dtype: bool
    - name: filecoverage
      dtype: float64
    - name: function_lines
      dtype: int64
    - name: mccabe
      dtype: int64
    - name: coverage
      dtype: float64
    - name: docstring_lines
      dtype: int64
    - name: function_nodoc
      dtype: string
    - name: id
      dtype: int64
  splits:
    - name: train
      num_bytes: 22383711
      num_examples: 11168
  download_size: 6805239
  dataset_size: 22383711
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*

DyPyBench Functions Datasets

DyPyBench is a dataset constructed by Piyush Krishan Bajaj at the Software Lab, Institute of Software Engineering, University of Stuttgart. It contains 50 open source projects from GitHub. We used Nathan Cooper's function_parser tool, based off GitHub's CodeSearchNet function_parser, to extract all functions from all the projects, excluding library functions in the virtualenv. We also ran all tests in DyPyBench and produced a coverage report in JSON. Not all projects resulted in a coverage report due to project specific coverage report settings.

The columns provided are as follows:

Column	Type	Notes
id	Int64	Unique id of the function
project	Int64	DyPyBench project id
nwo	string	Project name in repo/project format
sha	string	SHA commit hash
url	string	GitHub URL to function lines at commit
path	string	Path of file containing function relative to project root
func_begin	Int64	Begin of function line number in source file
func_end	Int64	End of function line number in source file
function_lines	Int64	Function line count
identifier	string	Function identifier
parameters	string	Function parameters
function	string	Source code of function including docstring
function_nodoc	string	Source code of function without docstring
docstring	string	Function docstring
docstring_lines	Int64	Line count of docstring
docstring_summary	string	Function docstring summary
return_statement	string	Function return statement
filecoverage	Float64	If coverage available, coverage percentage of file function is from
executed_lines	array[int]	If coverage available, executed lines relative to function lines (i.e. [0,1,2,...])
executed_lines_pc	Float64	If coverage available, executed line count over total function line count
missing_lines	array[int]	If coverage available, missing (unexecuted) lines relative to function lines (i.e. [0,1,2,...])
missing_lines_pc	Float64	If coverage available, missing line count over total function line count
covered	boolean	True if all lines executed and/or no lines missing
mccabe	Int64	McCabe complexity of function
coverage	Float64	Function coverage percentage (1-missing lines %)

Note: Missing/executed lines purposefully exclude lines skipped by pytest due to configuration e.g. line level # pragma: no cover.