File size: 7,103 Bytes
0c23011
 
7b35f39
 
 
 
 
 
 
 
 
aafae8e
1a4e37f
7b35f39
eb1fd0b
 
 
1a4e37f
 
 
 
 
 
eb1fd0b
 
fd6a622
eb1fd0b
 
b60e0cd
 
 
 
 
 
 
 
1a4e37f
 
 
b60e0cd
 
 
1a4e37f
 
 
b60e0cd
1a4e37f
 
 
 
 
 
b60e0cd
1a4e37f
 
eb1fd0b
7a94e88
b60e0cd
eb1fd0b
7e1d1fd
 
 
 
 
 
 
 
 
 
 
 
 
 
1a4e37f
eb1fd0b
d24a678
 
eb1fd0b
d24a678
 
d799677
eb1fd0b
7a94e88
 
eb1fd0b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0aaf80f
7a94e88
eb1fd0b
 
 
7a94e88
1a4e37f
eb1fd0b
7a94e88
eb1fd0b
 
 
 
 
 
 
 
1c0a594
eb1fd0b
 
d24a678
eb1fd0b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1c0a594
eb1fd0b
b60e0cd
 
 
 
eb1fd0b
b60e0cd
eb1fd0b
 
 
 
 
b531147
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
166
167
168
169
170
---
license: other
language:
- code
- en
task_categories:
- question-answering
- text-generation
- text2text-generation
tags:
- code
viewer: true
pretty_name: StackOverflow Posts Markdown
size_categories:
- 10M<n<100M
---

# StackOverflow Posts Markdown

![StackOverflow Logo](https://stackoverflow.design/assets/img/logos/so/logo-stackoverflow.png)


## Dataset Summary

This dataset contains all posts submitted to StackOverflow before the 14th of June 2023 formatted as **Markdown text**.<br>
The dataset contains ~60 Million posts, totaling ~35GB in size and ~65 billion characters of text.<br>
The data is sourced from [Internet Archive StackExchange Data Dump](https://archive.org/download/stackexchange).

## Dataset Structure

Each record corresponds to one post of a particular type.
Original ordering from the data dump is not exactly preserved due to parallelism in the script used to process the data dump.
The markdown content of each post is contained in the `Body` field. The license for a particular post is contained in the `ContentLicense` field.


### Data Fields
```typescript
{
  Id: long,
  PostTypeId: long, // 1=Question, 2=Answer, 3=Orphaned tag wiki, 4=Tag wiki excerpt, 5=Tag wiki, 6=Moderator nomination, 7=Wiki Placeholder, 8=Privilige Wiki
  AcceptedAnswerId: long | null, // only present if PostTypeId=1
  ParentId: long | null, // only present if PostTypeId=2
  Score: long,
  ViewCount: long | null,
  Body: string | null,
  Title: string | null,
  ContentLicense: string | null,
  FavoriteCount: long | null,
  CreationDate: string | null,
  LastActivityDate: string | null,
  LastEditDate: string | null,
  LastEditorUserId: long | null,
  OwnerUserId: long | null,
  Tags: array<string> | null
}
```
Also consider the [StackExchange Datadump Schema Documentation](https://meta.stackexchange.com/questions/2677/database-schema-documentation-for-the-public-data-dump-and-sede), as all fields
have analogs in the original dump format.

## How to use?

```python
from datasets import load_dataset

# predownload full dataset
ds = load_dataset('mikex86/stackoverflow-posts', split='train')

# dataset streaming (will only download the data as needed)
ds = load_dataset('mikex86/stackoverflow-posts', split='train', streaming=True)

for sample in iter(ds): print(sample["Body"])
```

## How is the text stored?

The original Data Dump formats the "Body" field as HTML, using tags such as `<code>`, `<h1>`, `<ul>`, etc.
This HTML format has been converted to Markdown.

### Markdown format

For reference, [this post on StackOverflow](https://stackoverflow.com/questions/53253940/make-react-useeffect-hook-not-run-on-initial-render) is formatted as follows:

#### Title: Make React useEffect hook not run on initial render

```markdown
According to the docs:

​> `componentDidUpdate()` is invoked immediately after updating occurs. This method is not called for the initial render.

We can use the new `useEffect()` hook to simulate `componentDidUpdate()`, but it seems like `useEffect()` is being ran after every render, even the first time. How do I get it to not run on initial render?

As you can see in the example below, `componentDidUpdateFunction` is printed during the initial render but `componentDidUpdateClass` was not printed during the initial render.

​`​`​`
function ComponentDidUpdateFunction() {
  const [count, setCount] = React.useState(0);
  React.useEffect(() => {
    console.log(""componentDidUpdateFunction"");
  });

  return (
    <div>
      <p>componentDidUpdateFunction: {count} times</p>
      <button
        onClick={() => {
          setCount(count + 1);
        }}
      >
        Click Me
      </button>
    </div>
  );
}
​`​`​`

rest of the post omitted for brevity

```


## Details on the HTML to Markdown conversion

Using Jsoup, the original Body field was converted into a Jsoup Document. The child **nodes** (has special meaning in context of Jsoup) of this document were recursively traversed in a depth-first order.

Jsoup defines `.text()` as follows:
> ... the normalized, combined text of this element and all its children. Whitespace is normalized and trimmed. For example, given HTML <code>&lt;p&gt;Hello &lt;b&gt;there&lt;/b&gt; now! &lt;/p&gt;<code>, p.text() returns "Hello there now!"

Jsoup defines a `Node` as follows:
> The base, abstract Node model. Elements, Documents, Comments etc are all Node instances.

Additionally the existence of the `TextNode` should be noted, which represents floating text inside an HTML document that is not itself an HTML element.
Thus this text tag `<p>Hello<code>World</code></p>` would have two Jsoup child nodes `TextNode(value="Hello")` and `Element(tag="code", value="World")`.
The value `field` of a `TextNode` contains the free standing text without any further treatment (no whitespace stripping, etc.)

### Traversing Rules

- When ecountering a html tag for which a rule exists, children are not further traversed, **unless explicitly stated otherwise**.
- When encountering an `<a>` tag, `[${element.text()}](${element.attr("href")})` is emitted.

- When encountering an `<h1>` tag, `\n# ${element.text()}\n\n` is emitted.
- When encountering an `<h2>` tag, `\n## ${element.text()}\n\n` is emitted.
- When encountering an `<h3>` tag, `\n### ${element.text()}\n\n` is emitted.
- When encountering an `<h4>` tag, `\n#### ${element.text()}\n\n` is emitted.
- When encountering an `<h5>` tag, `\n##### ${element.text()}\n\n` is emitted.
- When encountering an `<h6>` tag, `\n###### ${element.text()}\n\n` is emitted.

- When encountering a `<code>` tag, `` `${element.text()}` ``is emitted
- When encountering a `<pre>` tag and said element **has** a `<code>` child tag, `` ​`​`​`\n${element.text()}`\n​`​`​`\n`` is emitted.
- When encountering a `<pre>` tag and said element **does not** have a `<code>` child tag, **children are traversed further**.
- When encountering an `<li>` tag, `- ` is emitted and **children are traversed further**.
- When encountering a `<blockquote>` tag, `> ` is emitted and **children are traversed further**.
- When encountering an `<hr>` tag, `\n---\n\n` is emitted
- When encountering an `<img>` tag, `![${element.attr("alt")}](${element.attr("src")})` is emitted.
- When encountering a `<table>` tag
  - `\n| ` is emitted
  - For each element of `element.select("th")`
    - `${element.text()} | ` is emitted
  - After the loop `\n| ` is emitted
  - For each element of `element.select("th")`
    - For each character of the `th.text()`
      - `-` is emitted
    - After the loop over each character of th ` | ` is emitted
  - `\n` is emitted
  - For each element of `element.select("tr")` with more than one children of tag type `td`
    - `| ` is emitted
    - For each element of `element.select("td")`
      - `${td.text()} | ` is emitted
    - After the loop over `<td>` elements, `\n` is emitted
  - After the loop over `<tr>` elements, `\n` is emitted
- When encountering a jsoup `TextNode`, `${node.attr(node.nodeName())}` (which is equivalent to accessing the private field `node.value`) is emitted.