cdx-cc-news

CDXj indexes for Common Crawl News Dataset. See also News Dataset Announcement

I could not find indexes to Common Crawl News, so I decided to create some.

Directory Layout

├── cdxj
│   └── cc-news_CCYY_MM.cdxj.zst
└── data
    └── year=CCYY
        └── month=MM
            └── data_N.parquet

Basic Schema

• surt sort friendly URL optimized for btree lookup

• capture_time 14 decimal digits of time CCYYMMDDHHMMSSmmm

• json_record

{"url": "...",
 "mime": "text/html",
 "status": "200",
 "digest": "sha1:...",
 "length": "<bytes>",
 "offset": "<bytes>", 
 "filename": "CC-NEWS-20260712132848-00092.warc.gz"
}

com,example)/ 20170730223850 {"url": "http://example.com/", "mime": "text/html", "status": "200", "digest": "G7HRM7BGOKSKMSXZAHMUQTTV53QOFSMK", "length": "1219", "offset": "771", "filename": "example-20170730223917.warc.gz"}

This should give you enough info to do a range request on common crawl WARC files on s3 to pull the archived news url.

erata: CDXJ recommendation seems to suggest no spaces in the json payload? cdxj-indexer leaves spaces in the embedded json.

note: surt, capture_time is not always unique! Scripts in this repo in some cases (such as sqlite) only take the first occurance.

Scripts

A retrospective conversion was run on a c7g.2xlarge.

aws s3 ls s3://commoncrawl/crawl-data/CC-NEWS/ --recursive | grep warc.paths.gz > warc_paths.txt
# then I edited warc_paths.txt using search and replace in vi to get a list of s3://commoncrawl/crawl-data/CC-NEWS/... urls
cat warc_paths.txt | xargs -I {} uv run index_month.py {}

• ./index_month.py process a warc.paths.gz for a month (run in cdxj).
• ./parquet_from_cdxj.py create parquet files (run in root, writes to data)
• ./sqlite_from_cdxj.py create sqlite vertical partation index of a month

sqlite_from_cdxj.py output is not in the dataset, but that command takes local paths or URLs to the data on hugging face.

uv run sqlite_from_cdxj.py "https://huggingface.co/datasets/brian-learns/cdx-cc-news/resolve/main/cdxj/cc-news_2016_08.cdxj.zst?download=true"

Citations:

• Common Crawl Foundation (2026). news-crawl: News crawling with StormCrawler [Computer software]. GitHub. https://github.com/commoncrawl/news-crawl

• Tange, O. (2026). GNU Parallel 20260622 [Computer software]. Zenodo. https://doi.org/10.5281/zenodo.20959929

• Facebook AI (2026). Zstandard compression tool [Computer software]. https://facebook.github.io/zstd/

• Webrecorder (2026). cdxj-indexer [Computer software]. GitHub. https://github.com/webrecorder/cdxj-indexer

• Astral Software (2026). uv: An extremely fast Python package installer and resolver (Version 0.2.0) [Computer software]. GitHub. https://docs.astral.sh/uv/

• Internet Archive (2026). surt: Sort-friendly URI Reordering Transform [Computer software]. https://github.com/internetarchive/surt

• Webrecorder (2022). Crawl Index JSON (CDXJ) [Technical Recomendation] https://specs.webrecorder.net/cdxj/0.1.0/

• duckdb / sqlite