pack-priv Class SimilarityRenameDetector

File size threshold (in bytes) for detecting renames. Files larger than this size will not be processed for renames.

Number of bits we need to express an index into src or dst list.

This must be 28, giving us a limit of 2^28 entries in either list, which is an insane limit of 536,870,912 file names being considered in a single rename pass. The other 8 bits are used to store the score, while staying under 127 so the long doesn't go negative.

All destinations to consider looking for a rename.

A destination is typically an ChangeType#ADD, as the name has just come into existence, and we want to discover where its initial content came from.

Matrix of all examined file pairs, and their scores.

The upper 8 bits of each long stores the score, but the score is bounded to be in the range (0, 128] so that the highest bit is never set, and all entries are therefore positive.

List indexes to an element of srcs and dsts are encoded as the lower two groups of 28 bits, respectively, but the encoding is inverted, so that 0 is expressed as (1 << 28) - 1. This sorts lower list indices later in the matrix, giving precedence to files whose names sort earlier in the tree.

Score a pair must exceed to be considered a rename.

Skip content renames for binary files.

All sources to consider for copies or renames.

A source is typically a ChangeType#DELETE change, but could be another type when trying to perform copy detection concurrently with rename detection.

Set if any SimilarityIndex.TableFullException occurs.