( )⚙ D1408 treedirstate: use vlqencoding for numbers

This is an archive of the discontinued Mercurial Phabricator instance.

treedirstate: use vlqencoding for numbers
ClosedPublic

Authored by mbthomas on Nov 14 2017, 12:39 PM.
Tags
None
Subscribers

Details

Reviewers
durham
Group Reviewers
Restricted Project
Commits
rFBHGXa636c026aa4f: treedirstate: use vlqencoding for numbers
Summary

Change to use VLQ-encoded numbers for everything in the tree file. Block sizes
remain as u32s so that they can be read by the store in a single read
operation, but everything else is a VLQ as it is generally smaller and more
futureproof.

Diff Detail

Repository
rFBHGX Facebook Mercurial Extensions
Lint
Automatic diff as part of commit; lint not applicable.
Unit
Automatic diff as part of commit; unit tests not applicable.

Event Timeline

mbthomas created this revision.Nov 14 2017, 12:39 PM
Herald added a reviewer: Restricted Project. · View Herald TranscriptNov 14 2017, 12:39 PM
mbthomas updated this revision to Diff 3605.Nov 17 2017, 9:45 AM
mbthomas updated this revision to Diff 3660.Nov 20 2017, 12:21 PM
durham added a subscriber: durham.Nov 21 2017, 12:11 PM
durham added inline comments.
rust/treedirstate/src/tree.rs
86

noob question: why did this change?

durham accepted this revision.Nov 21 2017, 12:11 PM
This revision is now accepted and ready to land.Nov 21 2017, 12:11 PM
mbthomas added inline comments.Nov 21 2017, 12:54 PM
rust/treedirstate/src/tree.rs
86

I'm actually not sure. I think it has something to do with the way read_vlq() expects the reference for the Read trait object, but in trying to explain it I realized I don't understand it. I know the compiler requires it, though.

Would any Rust experts care to weigh in?

mbthomas updated this revision to Diff 3735.Nov 21 2017, 1:36 PM
quark added a subscriber: quark.Nov 23 2017, 5:56 PM
quark added inline comments.
rust/treedirstate/src/tree.rs
86

My understanding is &mut Read as a whole implements Read and is not a typical reference despite having the letter &. So

&mut &mut Read

makes it a reference and works.

I guess &mut Read itself can also be implicitly converted to a reference (how?). If itself is not marked as mut, it cannot be converted to a &mut reference. Hence the need of mut.

mbthomas updated this revision to Diff 3829.Nov 24 2017, 11:53 AM
mbthomas updated this revision to Diff 3846.Nov 24 2017, 3:18 PM
This revision was automatically updated to reflect the committed changes.
jsgf added a subscriber: jsgf.Dec 6 2017, 6:01 PM
jsgf added inline comments.
rust/treedirstate/src/tree.rs
86

Yeah, there's some subtle things going on here.

It's because .read_vlq() is a method on the VLQDecode trait, not Read. But VLQDecode has a blanket implementation for all types which implement Read.

In this case the object which implements Read is r, which has type &mut Read. The .read_vlq() method takes &mut self as its parameter, so it's getting &mut r as the parameter (hence the requirement for mut r).

But &mut &mut Read doesn't implement Read (and therefore VLQDecode) itself, so the Rust compiler's deref coercion kicks in - effectively, it keeps adding * on the front until either the type doesn't implement Deref (or DerefMut in this case), or it gets a type that matches the signature. (Alternatively it will also add at most one &/&mut if needed, which is why you don't need to keep typing (&foo).bar() for bar(&self)).

This means the end result is an effective method call of:

VLQDecode::read_vlq(*&mut &mut r)

(Though TBH I've probably got some details wrong here.)

But echoing a comment I made earlier in the series, I think it would be better to express this as:

fn read(r: &mut Read) -> Result<(Key, NodeEntry<T>)> {
    let mut r = r;

to keep the confusing extra mut out of the function definition.