darchives/hn-data-dumps

hn-data-dumps

Hacker News corpus has a few very nice properties. It is small enough that it can be analyzed on a laptop but at the same time it's big and interesting enough to do some non trivial experiments for learning or otherwise. To do any analysis it would be nice to have a copy of HN corpus. Looking around on the web I did find some efforts to have such a copy but each had something missing. Google BigQuery HN dataset came the closest but looks like it has not been updated in a while.

Luckily HN has a nice FireBase API which updates in real time. So I wrote a (very) small crawler to get all the items starting with id 1 all the way to id 25,562,625 (at the time of this writing).

Once the initial dataset has been crawled, incremental updates are quite cheap. There is a small script which runs once a day to download everything since last sync and then uploads a snapshot to this repo, in case anyone else finds it useful as well.

Running the script

This script has two dependencies, tqdm for reporting progress and aiohttp for async http client support. First step is to install them

pip install tqdm aiohttp

After this running the script running is as simple as

python hn_async2.py

This will start downloading all items sequentially starting with id 1 to the current max from Firebase DB and store them locally in a SQLite database named hn2.db3. Once the initial download is finished, which can take a long time depending on your computer and network, subsequent runs only download the new items created since the last run so they finish quickly.

Getting the stories data

The whole DB is around 13GB uncompressed but its mostly comments. If we limit to titles and urls of stories submitted on HN then its less than 1GB which easily compresses down to 30% of raw size. Daily updated snapshot all HN stories is available here

Once you have downloaded files, decompress the DB

zstd -d hn_stories.db3.zst

and load it in SQLite

sqlite3 hn_stories.db3

Data Schema

The schema of the DB is very simple. It only has one table - hn_stories which contains integer ID of the story and its attributes stored as JSON.

.schema
CREATE TABLE hn_stories(id INT PRIMARY KEY, item_json TEXT);

Here is a sample of rows

┌────┬───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┐
│ id │                                                                                           item_json                                                                                           │
├────┼───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┤
│ 20 │ {"by":"pg","time":1160424038,"title":"Salaries at VC-backed companies","url":"http://avc.blogs.com/a_vc/2006/10/search_by_salar.html"}                                                        │
│ 4  │ {"by":"onebeerdave","time":1160419662,"title":"NYC Developer Dilemma","url":"http://avc.blogs.com/a_vc/2006/10/the_nyc_develop.html"}                                                         │
│ 2  │ {"by":"phyllis","time":1160418628,"title":"A Student's Guide to Startups","url":"http://www.paulgraham.com/mit.html"}                                                                         │
│ 3  │ {"by":"phyllis","time":1160419233,"title":"Woz Interview: the early days of Apple","url":"http://www.foundersatwork.com/stevewozniak.html"}                                                   │
│ 7  │ {"by":"phyllis","time":1160420455,"title":"Sevin Rosen Unfunds - why?","url":"http://featured.gigaom.com/2006/10/09/sevin-rosen-unfunds-why/"}                                                │
│ 21 │ {"by":"sama","time":1160443271,"title":"Best IRR ever?  YouTube 1.65B...","url":"http://www.techcrunch.com/2006/10/09/google-has-acquired-youtube/"}                                          │
│ 9  │ {"by":"askjigga","time":1160421542,"title":"weekendr: social network for the weekend","url":"http://www.weekendr.com/"}                                                                       │
│ 1  │ {"by":"pg","time":1160418111,"title":"Y Combinator","url":"http://ycombinator.com"}                                                                                                           │
│ 5  │ {"by":"perler","time":1160419864,"title":"Google, YouTube acquisition announcement could come tonight","url":"http://www.techcrunch.com/2006/10/09/google-youtube-sign-more-separate-deals/"} │
│ 81 │ {"by":"justin","time":1171869130,"title":"allfreecalls.com shut down by AT&T","url":"http://www.techcrunch.com/2007/02/16/allfreecalls-shut-down/"}                                           │
└────┴───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────┘

Let me know if you find this useful. Thanks!