Fixing log -- filters, along with performance improvements

pmeunier

Dec 7, 2021, 2:05 PM

6U42MTEZTINWUU2KLJLQW33ZBZXKTHPTF6TMEG56TO642CQDZQMQC

Dependencies

• [2] 6F6AAHK4 Simplifying pijul::commands::log, and fixing Broken Pipe errors
• [3] SMMBFECL Converting to the new patch format "online"
• [4] OU6JOR3C Add path filtering for log, add json output for log
• [5] F2S6XETO Fixing log --hash-only
• [6] SXEYMYF7 Fixing the bad changes in history (unfortunately, by rebooting).
• [7] L4JXJHWX pijul/*: reorganize imports and remove extern crate

Change contents

• replacement in pijul/src/commands/log.rs at line 2
  [3.694]→[3.0:42](∅→∅)
  − use std::collections::{HashMap, HashSet};
  [3.694]

  [3.42]
  + use std::collections::HashMap;
• replacement in pijul/src/commands/log.rs at line 128
  [3.1009]→[2.2805:2851](∅→∅)
  − ) -> Result<Vec<libpijul::Inode>, Error<E>> {
  [3.1009]

  [3.1060]
  + ) -> Result<

  + Vec<(

  + libpijul::Inode,

  + Option<libpijul::pristine::Position<libpijul::ChangeId>>,

  + )>,

  + Error<E>,

  + > {
• replacement in pijul/src/commands/log.rs at line 158
  [2.3222]→[2.3222:3297](∅→∅)
  − Ok(Some(s)) => inodes.push(libpijul::fs::find_inode(txn, s)?),
  [2.3222]

  [3.2759]
  + Ok(Some(s)) => {

  + let inode = libpijul::fs::find_inode(txn, s)?;

  + let inode_position = txn.get_inodes(&inode, None)?;

  + inodes.push((inode, inode_position.cloned()))

  + }
• edit in pijul/src/commands/log.rs at line 169
  [3.2841]→[3.2841:3001](∅→∅),[3.3001]→[2.3298:3355](∅→∅),[2.3355]→[3.3070:3111](∅→∅),[3.3070]→[3.3070:3111](∅→∅),[3.3111]→[2.3356:3405](∅→∅),[2.3405]→[3.3165:3417](∅→∅),[3.3165]→[3.3165:3417](∅→∅),[3.3417]→[2.3406:3436](∅→∅),[2.3436]→[3.3489:3787](∅→∅),[3.3489]→[3.3489:3787](∅→∅),[3.3787]→[2.3437:3573](∅→∅),[2.3573]→[3.4336:4511](∅→∅),[3.4336]→[3.4336:4511](∅→∅)
  − /// Given a list of path filters which represent the files/directories for which

  − /// the user wants to see the logs, find the subset of relevant change hashes.

  − fn filtered_hashes<E: std::error::Error>(

  − txn: &Txn,

  − path: &Path,

  − filters: &[String],

  − ) -> Result<HashSet<libpijul::Hash>, Error<E>> {

  − let inodes = get_inodes(txn, path, filters)?;

  − let mut hashes = HashSet::<libpijul::Hash>::new();

  − for inode in inodes {

  − // The Position<ChangeId> for the file Inode.

  − let inode_position = match txn.get_inodes(&inode, None)? {

  − None => continue,

  − Some(p) => p,

  − };

  − for pair in txn.iter_touched(inode_position)? {

  − let (position, touched_change_id) = pair?;

  − // Push iff the file ChangeId for this element matches that of the file Inode

  − if &position.change == &inode_position.change {

  − let ser_h = txn.get_external(touched_change_id)?.unwrap();

  − hashes.insert(libpijul::Hash::from(*ser_h));

  − } else {

  − // We've gone past the relevant subset of changes in the iterator.

  − break;

  − }

  − }

  − }

  − Ok(hashes)

  − }

  −
• edit in pijul/src/commands/log.rs at line 312
  [3.10433]→[3.10433:10702](∅→∅)
  − // If the user applied path filters, figure out what change hashes

  − // are to be logged.

  − let mut requested_hashes = filtered_hashes(

  − &self.txn,

  − self.repo_path.as_ref(),

  − self.cmd.filters.as_slice(),

  − )?;
• replacement in pijul/src/commands/log.rs at line 313
  [3.10703]→[3.10703:11547](∅→∅),[3.11547]→[2.3650:3695](∅→∅),[2.3695]→[3.11678:12071](∅→∅),[3.11678]→[3.11678:12071](∅→∅)
  − // Get the (Hash, Merkle) pairs for the portion of reverse_log

  − // that are between offset and limit.

  − let hs = self

  − .txn

  − .reverse_log(&*self.channel_ref.read(), None)?

  − .skip(self.offset)

  − .take(self.limit)

  − .map(|res| {

  − res.map(|(_, (ser_h, ser_m))| {

  − (libpijul::Hash::from(ser_h), libpijul::Merkle::from(ser_m))

  − })

  − });

  −

  − for pr in hs {

  − let (h, mrk) = pr?;

  − if (self.cmd.filters.is_empty()) || requested_hashes.remove(&h) {

  − // If there were no path filters applied, OR is this was one of the hashes

  − // marked by the file filters that were applied

  − let entry = self.mk_log_entry(&mut authors, &mut id_path, h, Some(mrk))?;

  − f(entry).map_err(Error::E)?;

  − } else if requested_hashes.is_empty() {

  − // If the user applied path filters, but the relevant change hashes

  − // have been exhausted, we can break early.

  − break;

  − } else {

  − // The user applied path filters; this wasn't a hit, but

  − // there are still hits to be logged.

  − continue;
  [3.10703]

  [3.135674]
  + let inodes = get_inodes(&self.txn, &self.repo_path, &self.cmd.filters)?;

  + let mut offset = self.offset;

  + let mut limit = self.limit;

  + for pr in self.txn.reverse_log(&*self.channel_ref.read(), None)? {

  + let (_, (h, mrk)) = pr?;

  + let cid = self.txn.get_internal(h)?.unwrap();

  + let mut is_in_filters = inodes.is_empty();

  + for (_, position) in inodes.iter() {

  + if let Some(position) = position {

  + is_in_filters = self.txn.get_touched_files(position, Some(cid))?.is_some();

  + if is_in_filters {

  + break;

  + }

  + }

  + }

  + if is_in_filters {

  + if offset == 0 && limit > 0 {

  + // If there were no path filters applied, OR is this was one of the hashes

  + // marked by the file filters that were applied

  + let entry =

  + self.mk_log_entry(&mut authors, &mut id_path, h.into(), Some(mrk.into()))?;

  + f(entry).map_err(Error::E)?;

  + limit -= 1

  + } else if limit > 0 {

  + offset -= 1

  + } else {

  + break;

  + }