SequenceR data new

Dataset of single line patches for SequenceR. Obtained by preprocessing and merging datasets CodRep and the one used for Learning Bug-Fixing Patches in the Wild via Neural Machine Translation.

Preprocessing steps will be located at https://github.com/KTH/chai/tree/master/src/embedding-work

The structure of the  tar file is as follows:

sequencer-data.tar.gz
├── test
│   ├── meta.tsv
│   ├── 1
│   │   ├── buggy_file.java
│   │   └── fixed_line.txt
│   ├── 2
│   │   ├── buggy_file.java
│   │   └── fixed_line.txt
│   ├── 3
│   │   ├── ...
├── train
│   ├── meta.tsv
│   ├── 1
│   │   ├── ...
│   │   ...
 

.java files contain the abstracted source code (see SequenceR paper), .txt files contain the fixed line, and the meta.tsv files contain a mapping between each dir (1, 2, ...) and a unique id of the example, and the line number for the buggy line.