alexjercan commited on
Commit
560bdd9
1 Parent(s): 3f802bb
Files changed (1) hide show
  1. README.md +11 -10
README.md CHANGED
@@ -96,19 +96,19 @@ dataset_info:
96
 
97
  # About the Dataset
98
 
99
- The source code used to generate the dataset can be found on
100
  [GitHub](https://github.com/alexjercan/bug-detection/tree/master/bugnet)
101
 
102
- The dataset is based on the [CodeNet project](https://github.com/IBM/Project_CodeNet)
103
- and contains Python and C++ code submissions for online coding competitions. The data
104
- is obtained by selecting consecutive attempts of a single user that resulted in fixing a
105
- buggy submission. Thus the data is represented by code pairs and annotated by the diff
106
- and error of each changed instruction. We have already tokenized all the source code
107
  files and kept the same format as in the original dataset.
108
 
109
- The upgrade made compared to CodeNetPy is that we only keep one line errors.
110
- This means that the task of bug detection and repair will be easier to manage.
111
- We also removed all the files that fail on linters, so that we are focusing only
112
  on bugs that cannot be identified easily.
113
 
114
  The resulting dataset file will be a csv with the following columns:
@@ -124,6 +124,7 @@ The resulting dataset file will be a csv with the following columns:
124
  - `j2`: End of the change in the accepted source (not inclusive; for `delete` we have `j1 == j2`)
125
  - `error`: The error that was obtained running the buggy source code on the input/output examples
126
  - `stderr`: The full output of stderr of running the buggy source code on the input/output examples
 
127
  - `description`: The problem statement in html format
128
  - `input`: The input for the test case
129
- - `output`: The output for the test case
 
96
 
97
  # About the Dataset
98
 
99
+ The source code used to generate the dataset can be found on
100
  [GitHub](https://github.com/alexjercan/bug-detection/tree/master/bugnet)
101
 
102
+ The dataset is based on the [CodeNet project](https://github.com/IBM/Project_CodeNet)
103
+ and contains Python and C++ code submissions for online coding competitions. The data
104
+ is obtained by selecting consecutive attempts of a single user that resulted in fixing a
105
+ buggy submission. Thus the data is represented by code pairs and annotated by the diff
106
+ and error of each changed instruction. We have already tokenized all the source code
107
  files and kept the same format as in the original dataset.
108
 
109
+ The upgrade made compared to CodeNetPy is that we only keep one line errors.
110
+ This means that the task of bug detection and repair will be easier to manage.
111
+ We also removed all the files that fail on linters, so that we are focusing only
112
  on bugs that cannot be identified easily.
113
 
114
  The resulting dataset file will be a csv with the following columns:
 
124
  - `j2`: End of the change in the accepted source (not inclusive; for `delete` we have `j1 == j2`)
125
  - `error`: The error that was obtained running the buggy source code on the input/output examples
126
  - `stderr`: The full output of stderr of running the buggy source code on the input/output examples
127
+ - `stdout`: The full output of stdout of running the buggy source code on the input/output examples
128
  - `description`: The problem statement in html format
129
  - `input`: The input for the test case
130
+ - `output`: The output for the test case