Mercurial > public > mercurial-scm > hg
view rust/hg-core/src/revlog/path_encode.rs @ 48178:f12a19d03d2c
fix: reduce number of tool executions
By grouping together (path, ctx) pairs according to the inputs they would
provide to fixer tools, we can deduplicate executions of fixer tools to
significantly reduce the amount of time spent running slow tools.
This change does not handle clean files in the working copy, which could still
be deduplicated against the files in the checked out commit. It's a little
harder to do that because the filerev is not available in the workingfilectx
(and it doesn't exist for added files).
Anecdotally, this change makes some real uses cases at Google 10x faster. I
think we were originally hesitant to do this because the benefits weren't
obvious, and implementing it efficiently is kind of tricky. If we simply
memoized the formatter execution function, we would be keeping tons of file
content in memory.
Also included is a regression test for a corner case that I broke with my first
attempt at optimizing this code.
Differential Revision: https://phab.mercurial-scm.org/D11280
author | Danny Hooper <hooper@google.com> |
---|---|
date | Thu, 02 Sep 2021 14:08:45 -0700 |
parents | fad504cfc94b |
children | e98fd81bb151 |
line wrap: on
line source
use sha1::{Digest, Sha1}; #[derive(PartialEq, Debug)] #[allow(non_camel_case_types)] enum path_state { START, /* first byte of a path component */ A, /* "AUX" */ AU, THIRD, /* third of a 3-byte sequence, e.g. "AUX", "NUL" */ C, /* "CON" or "COMn" */ CO, COMLPT, /* "COM" or "LPT" */ COMLPTn, L, LP, N, NU, P, /* "PRN" */ PR, LDOT, /* leading '.' */ DOT, /* '.' in a non-leading position */ H, /* ".h" */ HGDI, /* ".hg", ".d", or ".i" */ SPACE, DEFAULT, /* byte of a path component after the first */ } /* state machine for dir-encoding */ #[allow(non_camel_case_types)] enum dir_state { DDOT, DH, DHGDI, DDEFAULT, } fn inset(bitset: &[u32; 8], c: u8) -> bool { bitset[(c as usize) >> 5] & (1 << (c & 31)) != 0 } fn charcopy(dest: Option<&mut [u8]>, destlen: &mut usize, c: u8) { if let Some(slice) = dest { slice[*destlen] = c } *destlen += 1 } fn memcopy(dest: Option<&mut [u8]>, destlen: &mut usize, src: &[u8]) { if let Some(slice) = dest { slice[*destlen..*destlen + src.len()].copy_from_slice(src) } *destlen += src.len(); } fn rewrap_option<'a, 'b: 'a>( x: &'a mut Option<&'b mut [u8]>, ) -> Option<&'a mut [u8]> { match x { None => None, Some(y) => Some(y), } } fn hexencode<'a>(mut dest: Option<&'a mut [u8]>, destlen: &mut usize, c: u8) { let hexdigit = b"0123456789abcdef"; charcopy( rewrap_option(&mut dest), destlen, hexdigit[(c as usize) >> 4], ); charcopy(dest, destlen, hexdigit[(c as usize) & 15]); } /* 3-byte escape: tilde followed by two hex digits */ fn escape3(mut dest: Option<&mut [u8]>, destlen: &mut usize, c: u8) { charcopy(rewrap_option(&mut dest), destlen, b'~'); hexencode(dest, destlen, c); } fn encode_dir(mut dest: Option<&mut [u8]>, src: &[u8]) -> usize { let mut state = dir_state::DDEFAULT; let mut i = 0; let mut destlen = 0; while i < src.len() { match state { dir_state::DDOT => match src[i] { b'd' | b'i' => { state = dir_state::DHGDI; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'h' => { state = dir_state::DH; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } _ => { state = dir_state::DDEFAULT; } }, dir_state::DH => { if src[i] == b'g' { state = dir_state::DHGDI; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = dir_state::DDEFAULT; } } dir_state::DHGDI => { if src[i] == b'/' { memcopy(rewrap_option(&mut dest), &mut destlen, b".hg"); charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } state = dir_state::DDEFAULT; } dir_state::DDEFAULT => { if src[i] == b'.' { state = dir_state::DDOT } charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } } } destlen } fn _encode( twobytes: &[u32; 8], onebyte: &[u32; 8], mut dest: Option<&mut [u8]>, src: &[u8], encodedir: bool, ) -> usize { let mut state = path_state::START; let mut i = 0; let mut destlen = 0; let len = src.len(); while i < len { match state { path_state::START => match src[i] { b'/' => { charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'.' => { state = path_state::LDOT; escape3(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b' ' => { state = path_state::DEFAULT; escape3(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'a' => { state = path_state::A; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'c' => { state = path_state::C; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'l' => { state = path_state::L; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'n' => { state = path_state::N; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'p' => { state = path_state::P; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } _ => { state = path_state::DEFAULT; } }, path_state::A => { if src[i] == b'u' { state = path_state::AU; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = path_state::DEFAULT; } } path_state::AU => { if src[i] == b'x' { state = path_state::THIRD; i += 1; } else { state = path_state::DEFAULT; } } path_state::THIRD => { state = path_state::DEFAULT; match src[i] { b'.' | b'/' | b'\0' => escape3( rewrap_option(&mut dest), &mut destlen, src[i - 1], ), _ => i -= 1, } } path_state::C => { if src[i] == b'o' { state = path_state::CO; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = path_state::DEFAULT; } } path_state::CO => { if src[i] == b'm' { state = path_state::COMLPT; i += 1; } else if src[i] == b'n' { state = path_state::THIRD; i += 1; } else { state = path_state::DEFAULT; } } path_state::COMLPT => { if src[i] >= b'1' && src[i] <= b'9' { state = path_state::COMLPTn; i += 1; } else { state = path_state::DEFAULT; charcopy( rewrap_option(&mut dest), &mut destlen, src[i - 1], ); } } path_state::COMLPTn => { state = path_state::DEFAULT; match src[i] { b'.' | b'/' | b'\0' => { escape3( rewrap_option(&mut dest), &mut destlen, src[i - 2], ); charcopy( rewrap_option(&mut dest), &mut destlen, src[i - 1], ); } _ => { memcopy( rewrap_option(&mut dest), &mut destlen, &src[i - 2..i], ); } } } path_state::L => { if src[i] == b'p' { state = path_state::LP; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = path_state::DEFAULT; } } path_state::LP => { if src[i] == b't' { state = path_state::COMLPT; i += 1; } else { state = path_state::DEFAULT; } } path_state::N => { if src[i] == b'u' { state = path_state::NU; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = path_state::DEFAULT; } } path_state::NU => { if src[i] == b'l' { state = path_state::THIRD; i += 1; } else { state = path_state::DEFAULT; } } path_state::P => { if src[i] == b'r' { state = path_state::PR; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = path_state::DEFAULT; } } path_state::PR => { if src[i] == b'n' { state = path_state::THIRD; i += 1; } else { state = path_state::DEFAULT; } } path_state::LDOT => match src[i] { b'd' | b'i' => { state = path_state::HGDI; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'h' => { state = path_state::H; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } _ => { state = path_state::DEFAULT; } }, path_state::DOT => match src[i] { b'/' | b'\0' => { state = path_state::START; memcopy(rewrap_option(&mut dest), &mut destlen, b"~2e"); charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'd' | b'i' => { state = path_state::HGDI; charcopy(rewrap_option(&mut dest), &mut destlen, b'.'); charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } b'h' => { state = path_state::H; memcopy(rewrap_option(&mut dest), &mut destlen, b".h"); i += 1; } _ => { state = path_state::DEFAULT; charcopy(rewrap_option(&mut dest), &mut destlen, b'.'); } }, path_state::H => { if src[i] == b'g' { state = path_state::HGDI; charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } else { state = path_state::DEFAULT; } } path_state::HGDI => { if src[i] == b'/' { state = path_state::START; if encodedir { memcopy( rewrap_option(&mut dest), &mut destlen, b".hg", ); } charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1 } else { state = path_state::DEFAULT; } } path_state::SPACE => match src[i] { b'/' | b'\0' => { state = path_state::START; memcopy(rewrap_option(&mut dest), &mut destlen, b"~20"); charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } _ => { state = path_state::DEFAULT; charcopy(rewrap_option(&mut dest), &mut destlen, b' '); } }, path_state::DEFAULT => { while i != len && inset(onebyte, src[i]) { charcopy(rewrap_option(&mut dest), &mut destlen, src[i]); i += 1; } if i == len { break; } match src[i] { b'.' => { state = path_state::DOT; i += 1 } b' ' => { state = path_state::SPACE; i += 1 } b'/' => { state = path_state::START; charcopy(rewrap_option(&mut dest), &mut destlen, b'/'); i += 1; } _ => { if inset(onebyte, src[i]) { loop { charcopy( rewrap_option(&mut dest), &mut destlen, src[i], ); i += 1; if !(i < len && inset(onebyte, src[i])) { break; } } } else if inset(twobytes, src[i]) { let c = src[i]; i += 1; charcopy( rewrap_option(&mut dest), &mut destlen, b'_', ); charcopy( rewrap_option(&mut dest), &mut destlen, if c == b'_' { b'_' } else { c + 32 }, ); } else { escape3( rewrap_option(&mut dest), &mut destlen, src[i], ); i += 1; } } } } } } match state { path_state::START => (), path_state::A => (), path_state::AU => (), path_state::THIRD => { escape3(rewrap_option(&mut dest), &mut destlen, src[i - 1]) } path_state::C => (), path_state::CO => (), path_state::COMLPT => { charcopy(rewrap_option(&mut dest), &mut destlen, src[i - 1]) } path_state::COMLPTn => { escape3(rewrap_option(&mut dest), &mut destlen, src[i - 2]); charcopy(rewrap_option(&mut dest), &mut destlen, src[i - 1]); } path_state::L => (), path_state::LP => (), path_state::N => (), path_state::NU => (), path_state::P => (), path_state::PR => (), path_state::LDOT => (), path_state::DOT => { memcopy(rewrap_option(&mut dest), &mut destlen, b"~2e"); } path_state::H => (), path_state::HGDI => (), path_state::SPACE => { memcopy(rewrap_option(&mut dest), &mut destlen, b"~20"); } path_state::DEFAULT => (), }; destlen } fn basic_encode(dest: Option<&mut [u8]>, src: &[u8]) -> usize { let twobytes: [u32; 8] = [0, 0, 0x87ff_fffe, 0, 0, 0, 0, 0]; let onebyte: [u32; 8] = [1, 0x2bff_3bfa, 0x6800_0001, 0x2fff_ffff, 0, 0, 0, 0]; _encode(&twobytes, &onebyte, dest, src, true) } const MAXSTOREPATHLEN: usize = 120; fn lower_encode(mut dest: Option<&mut [u8]>, src: &[u8]) -> usize { let onebyte: [u32; 8] = [1, 0x2bff_fbfb, 0xe800_0001, 0x2fff_ffff, 0, 0, 0, 0]; let lower: [u32; 8] = [0, 0, 0x07ff_fffe, 0, 0, 0, 0, 0]; let mut destlen = 0; for c in src { if inset(&onebyte, *c) { charcopy(rewrap_option(&mut dest), &mut destlen, *c) } else if inset(&lower, *c) { charcopy(rewrap_option(&mut dest), &mut destlen, *c + 32) } else { escape3(rewrap_option(&mut dest), &mut destlen, *c) } } destlen } fn aux_encode(dest: Option<&mut [u8]>, src: &[u8]) -> usize { let twobytes = [0; 8]; let onebyte: [u32; 8] = [!0, 0xffff_3ffe, !0, !0, !0, !0, !0, !0]; _encode(&twobytes, &onebyte, dest, src, false) } fn hash_mangle(src: &[u8], sha: &[u8]) -> Vec<u8> { let dirprefixlen = 8; let maxshortdirslen = 68; let mut destlen = 0; let last_slash = src.iter().rposition(|b| *b == b'/'); let last_dot: Option<usize> = { let s = last_slash.unwrap_or(0); src[s..] .iter() .rposition(|b| *b == b'.') .and_then(|i| Some(i + s)) }; let mut dest = vec![0; MAXSTOREPATHLEN]; memcopy(Some(&mut dest), &mut destlen, b"dh/"); { let mut first = true; for slice in src[..last_slash.unwrap_or_else(|| src.len())] .split(|b| *b == b'/') { let slice = &slice[..std::cmp::min(slice.len(), dirprefixlen)]; if destlen + (slice.len() + if first { 0 } else { 1 }) > maxshortdirslen + 3 { break; } else { if !first { charcopy(Some(&mut dest), &mut destlen, b'/') }; memcopy(Some(&mut dest), &mut destlen, slice); if dest[destlen - 1] == b'.' || dest[destlen - 1] == b' ' { dest[destlen - 1] = b'_' } } first = false; } if !first { charcopy(Some(&mut dest), &mut destlen, b'/'); } } let used = destlen + 40 + { if let Some(l) = last_dot { src.len() - l } else { 0 } }; if MAXSTOREPATHLEN > used { let slop = MAXSTOREPATHLEN - used; let basenamelen = match last_slash { Some(l) => src.len() - l - 1, None => src.len(), }; let basenamelen = std::cmp::min(basenamelen, slop); if basenamelen > 0 { let start = match last_slash { Some(l) => l + 1, None => 0, }; memcopy( Some(&mut dest), &mut destlen, &src[start..][..basenamelen], ) } } for c in sha { hexencode(Some(&mut dest), &mut destlen, *c); } if let Some(l) = last_dot { memcopy(Some(&mut dest), &mut destlen, &src[l..]); } if destlen == dest.len() { dest } else { // sometimes the path are shorter than MAXSTOREPATHLEN dest[..destlen].to_vec() } } const MAXENCODE: usize = 4096 * 4; fn hash_encode(src: &[u8]) -> Vec<u8> { let dired = &mut [0; MAXENCODE]; let lowered = &mut [0; MAXENCODE]; let auxed = &mut [0; MAXENCODE]; let baselen = (src.len() - 5) * 3; if baselen >= MAXENCODE { panic!("path_encode::hash_encore: string too long: {}", baselen) }; let dirlen = encode_dir(Some(&mut dired[..]), src); let sha = Sha1::digest(&dired[..dirlen]); let lowerlen = lower_encode(Some(&mut lowered[..]), &dired[..dirlen][5..]); let auxlen = aux_encode(Some(&mut auxed[..]), &lowered[..lowerlen]); hash_mangle(&auxed[..auxlen], &sha) } pub fn path_encode(path: &[u8]) -> Vec<u8> { let newlen = if path.len() <= MAXSTOREPATHLEN { basic_encode(None, path) } else { MAXSTOREPATHLEN + 1 }; if newlen <= MAXSTOREPATHLEN { if newlen == path.len() { path.to_vec() } else { let mut res = vec![0; newlen]; basic_encode(Some(&mut res), path); res } } else { hash_encode(&path) } }