WILDkat: a Workaround for Information Loss by DMCA at KAT, ®107 (thru May 2015) (N=1.1M)

Contains: Plaintext listings providing information about more than a million torrents removed from KAT.


Details:
First Torrent: Torrent: studio ghiblis gedo senki-tales from earthsea 2006[672x432-x264-dts][goat] mkv (#6)
Last Torrent: Torrent: DevilsFilm – Karmen Karma, Danica Dillon – The Destruction of Danica Dillon [HD-720p] (#10,799,962)
Total Count: 1,136,243
True Positives: ~520,000 (torrents removed due to DMCA)
False Positives: ~620,000 (torrents removed, but not due to DMCA)
False Negatives: ~250,000 (torrents removed due to DMCA, but not listed)


Formatting: The original data stems from ~450 historical snapshorts of the original KAT torrent inventory data, which were collated into a single coherent and comprehensive dataset. This listing represents what is in that historical dataset, but not in the current one. Entries thru mid-April 2015 are formatted like so:

|h>5AC665A211C02FD8280ADB50D19F0BF7ACFD42D9|c>Movies |i>10499973|n>Taken 3 L Ora Della Verità 2015 iTALiAN AC3 DUAL EXTENDED 1080p BluRay x264-iG

More recent entries are more informative, taking advantage of a change in KAT's data export format:

|h>78B6D68DDD6A5E406DA2EDFF4820B5E328E37C9E|c>Movies/Other |i>10500122|s> 5,864,636,557|f> 1|u> 0|d> 0|t>2015-04-14/1428948626|n>Godzilla 2014

Three of the fields contain data intrinsic to the torrent itself:
|h>-info_hash
|s>-content size in bytes, formatted for improved legibility, with a 1-TB ceiling
|f>-number of contained files

Two of the remaining fields contain data originating elsewhere but cached by KAT (and consequently last updated some time before the torrent's removal, hence the current values will differ somewhat and may differ significantly):
|u>-number of uploaders (seeders)
|d>-number of downloaders (peers)

The final four fields contain data which, short of other sites obtaining it from here or the uploader re-submitting it elsewhere, is original to KAT:
|c>-major category (/minor subcategory)
|i>-internal identifier
|t>-submission date (yyyy-mm-dd)/timestamp (in "seconds since 1970" POSIX format)
|n>-submitted name


Organization: The torrent contains a big and a small .rar archive.

The big archive is mainly meant for searching: It contains the full dataset in the form of a single ~150-MB plaintext file, sorted by the ID column, with the entries switching from the old 4-field to the new 9-field format at torrent #10,500,000.

The small archive is mainly meant for browsing: It contains only the entries in the new format, in the form of 45 ~1 MB plaintext files, resulting from breaking up the torrents into the nine major categories and then sorting them by five different combinations of two or more columns - by the minor subcategory, then by one of the remaining fields, minus the ones that would be mostly meaningless (info_hash and number of downloaders) or practically redundant (ID).