Processed 172,803,584 links and ranked 8,166,507 pages from Wikipedia, using PageRank algorithm. The top ranked pages, excluding pages that are internal or administrative for Wikipedia, are in rank.txt.

DataProcessor

Process the raw wiki links data obtained using wikicrush with help from Kaiyuan.

1. Scan through indexbi.bin. Create offset -> index HashMap. Store the offset in offset.bin in the order of their indexes for recovery of the article title later.

2. Read indexbi.bin. Only write to the new file links.bin [ pageNum, { pageIdx,linkNum,[ linkedPageIdx, ... ] }, ... ]. Every number is converted from little-endian to big-endian.

PageRank

In the cache-efficient version, the links file is divided into several files, based on Haveliwala's algorithm, to exploit locality.