Skip to content

Latest commit

 

History

History
46 lines (34 loc) · 2.22 KB

USERSTUDY.md

File metadata and controls

46 lines (34 loc) · 2.22 KB

TransMap User Study

The details about the user study can be found in the paper, section 6.4. We provide the following files for the user study:

Related Files

  • data/transmap/user_study_cmp/user_data_ro: The randomly selected cases (4 mistakes for long programs and 8 mistakes for short programs) for the user study.
  • data/transmap/user_study_cmp/Dockerfile: The dockerfile for the code-server (the browser version of vscode, exposed at port 8080 of the docker container). Additionally, we installed debuggers and language server plugins to make it a standard IDE environment for coding and debugging Python and JavaScript programs. The control group uses a browser link to this web-based vscode as the only tool to debug the programs.

Below describes the semantic mistakes of selected benchmarks in the user study.

short programs

  • short_GFG_COUNT_POSSIBLE_...
    • mod % operation is semantically different.
  • short_GFG_MINIMUM_COST_...
    • sort() function call is used wrongly.
  • short_GFG_SQUARE_ROOT_OF_AN_INTEGER_1_py
    • Math.floor function call is missing.
  • short_H132
    • wrong loop stucture used: for .. in .. and for .. of ..
  • short_L0885_SpiralMatrixIII_py
    • a < b < c is different in Python and JavaScript.
  • short_L0926_FlipStringtoMonotoneIncreasing_py
    • Array.fill(0) function call is missing.
  • short_L1002_FindCommonCharacters_py
    • wrong loop stucture used: for .. in .. and for .. of ..
  • short_L2171_RemovingMinimumNumberofMagicBeans_py
    • sort() function call is used wrongly.

long programs

  • stringsim0 (mistake 1)
    • missing fill(0) function call
  • stringsim0 (mistake 2)
    • d[x] should be d.at(x) for negative index.
  • stringsim3
    • pop(i) with arguments should be splice(i).
  • stringsim4
    • x.length should be Object.keys(x).length depending on the context.

User Study Results

  • Due to privacy concerns, we will not release the operation logs for each of the human subjects during the user study. We only release the aggregated results and the code producing the results in the Jupyter Notebook at automations/transmap/userstudy_results.ipynb. However, since the raw data is not released, the notebook cannot be executed interactively.