File size: 4,027 Bytes
55dff4e
 
 
a344a1a
55dff4e
 
 
 
 
 
23a91b1
 
 
 
d53ce63
d2cb6f7
4ed4f59
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2b035e4
4ed4f59
 
 
e439957
4ed4f59
de5560e
 
e439957
de5560e
e439957
de5560e
e439957
 
 
 
 
 
 
 
 
 
 
 
 
de5560e
e439957
de5560e
e439957
de5560e
 
e439957
de5560e
e439957
 
 
 
 
 
 
 
 
 
 
 
 
 
de5560e
 
e439957
de5560e
 
 
6fa3289
de5560e
2b035e4
 
 
 
de5560e
2b035e4
e439957
 
1ab9161
 
 
e439957
 
17ab3f0
d53ce63
e439957
d53ce63
 
 
 
17ab3f0
 
 
e439957
 
d53ce63
 
 
e439957
 
17ab3f0
d53ce63
 
e439957
 
17ab3f0
d53ce63
 
e439957
 
d53ce63
 
 
e439957
 
d53ce63
e439957
973e469
 
 
e439957
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
---
task_categories:
- summarization
- text-generation
language:
- en
tags:
- code
size_categories:
- 10K<n<100K
---

# Overview

This dataset contains 34,000+ rows of code-docstring-ast data along with additional metadata. Data was gathered from various Python libraries and frameworks and their
publicly available GitHub repos. This dataset was created for the purpose of training the [CodeT5+](https://arxiv.org/abs/2305.07922) transformer on AST-enhanced code-to-doc tasks.

# Sources

The dataset was gathered from various GitHub repos sampled from [this repo by Vinta.](https://github.com/vinta/awesome-python)

The 26 repos are:
- matplotlib
- pytorch
- cryptography
- django
- prospector
- scikit-learn
- pandas
- numpy
- uvicorn
- feincms
- algorithms
- scrapy
- authlib
- seaborn
- coconut
- tensorflow
- flexx
- salmon
- mongo-python-driver
- virtualenv
- sphinx
- schema
- kornia
- scipy
- cherrypy
- pygame

Sampling was at random; I simply browsed through each category from Vinta's list and chose one from a random interesting category. 

# Dataset Instance

An instance of the dataset is as follows:

```
{
  <library> : <The library from which the source code came from>,
  <name> : <The name of the function/class/method>,
  <source_code> : <The raw source code itself stripped of its docstrings and comments>,
  <docstring> : <The corresponding docstring of the code>,
  <type> : <Whether it's a function, method, or class>,
  <file_path> : <The relative path of the file containing the function>,
  <ast_data> : <A flattened representation of the parsed AST. For more info about this, see section below>
}
```

# The AST Data

An extractor only focuses on specific nodes relevant for docstring generation denoted by this set:

```
KEEP_NODES = {
        'FunctionDef', 'AsyncFunctionDef', 'ClassDef',
        'arguments', 'arg', 'Return',
        'If', 'For', 'While', 'Try', 'With',
        'Assign', 'Call',
        'Raise', 'ExceptHandler',
        'decorator', 'bases',
        'Compare', 'BoolOp'
    }
```

Everything else is discarded. For example

**Source Code**
```
def tox_append_version_info() -> str: return '[toxfile]'
```

**Resulting AST Dictionary**
```
"ast_data": {
      "type": "FunctionDef",
      "children": [
        {
          "type": "arguments",
          "args": []
        },
        {
          "type": "Return",
          "has_value": true
        }
      ],
      "name": "tox_append_version_info"
    }
```

This dictionary is then flattened via a helper function which would then look something like `FunctionDef name:tox_append_version_info arguments Return return:yes`.

# Preprocessing

The dataset generally follows [CodeBERT's code2nl](https://github.com/microsoft/CodeBERT/tree/master/CodeBERT/code2nl) dataset cleaning standards which are as follows:

- Removed comments from the code
- Removed examples where code cannot be parsed into an AST
- Remove examples that documents contain special tokens (e.g. <img ...> or https:...)
- Remove examples that documents are not English

Furthermore, the following cleaning steps specific to this dataset were applied:
- Removed examples where, using CodeT5+'s tokenizer, the combined tokens of the source_code + ast_data is > 512
- Removed examples where, using CodeT5+'s tokenizer, the docstrings are > 512

# Final Statistics

```
{
  "original_samples": 128880,
  "processed_samples": 34537,
  "filter_stats": {
    "success": 34537,
    "non_english": 1202,
    "docstring_too_long": 3848,
    "input_too_long": 8366,
    "docstring_too_short": 74013,
    "error": 0,
    "error: unhashable type: 'list'": 6914
  },
  "split_sizes": {
    "train": 24175,
    "val": 5180,
    "test": 5182
  },
  "input_token_stats": {
    "min": 16,
    "max": 505,
    "avg": 164.071
  },
  "target_token_stats": {
    "min": 4,
    "max": 254,
    "avg": 52.758
  },
  "type_distribution": {
    "method": 12682,
    "function": 8733,
    "class": 2760
  }
}

```

# NOTE

This dataset is *imperfect*. Use under your own volition.