MCPcopy Create free account
hub / github.com/atomicdotdev/atomic / similarity_bps

Function similarity_bps

atomic-cli/src/commands/git/parallel.rs:9467–9496  ·  view source on GitHub ↗

Approximate byte similarity in basis points (`0..=10_000`) (CB-9C). Import rename detection is heuristic evidence, so the score documents the basis rather than claiming a metric: byte-identical content scores full marks, differing content uses a common prefix+suffix ratio over the longer byte sequence. The multiply runs in widened `u64` (review CB-9C R4): a `u32` shared-bytes × 10,000 product ove

(old_bytes: &[u8], new_bytes: &[u8])

Source from the content-addressed store, hash-verified

9465/// silent wrap in release — for supported inputs above ~429 KB, which the
9466/// larger-file corpus requirement makes a normal case, not an exception.
9467fn similarity_bps(old_bytes: &[u8], new_bytes: &[u8]) -> u16 {
9468 if old_bytes == new_bytes {
9469 return 10_000;
9470 }
9471 if old_bytes.is_empty() && new_bytes.is_empty() {
9472 return 10_000;
9473 }
9474 let common_prefix = old_bytes
9475 .iter()
9476 .zip(new_bytes.iter())
9477 .take_while(|(a, b)| a == b)
9478 .count();
9479 let common_suffix = old_bytes[common_prefix..]
9480 .iter()
9481 .rev()
9482 .zip(new_bytes[common_prefix.min(new_bytes.len())..].iter().rev())
9483 .take_while(|(a, b)| a == b)
9484 .count();
9485 let denominator = old_bytes.len().max(new_bytes.len()) as u64;
9486 let shared = (common_prefix + common_suffix).min(denominator as usize) as u64;
9487
9488 if denominator == 0 {
9489 10_000
9490 } else {
9491 shared
9492 .checked_mul(10_000)
9493 .map(|scaled| (scaled / denominator).min(9_999) as u16)
9494 .unwrap_or(10_000)
9495 }
9496}
9497
9498/// The graph-backed materialization (kind + mode) one canonical Git mode
9499/// requires (CB-9C). Symlinks project the platform-independent `0o777` the

Calls 4

is_emptyMethod · 0.45
countMethod · 0.45
iterMethod · 0.45
lenMethod · 0.45