This avoids https://github.com/rust-lang/futures-rs/issues/2090. I
don't think we need to worry about reading legacy conflicts
asynchronously - async is really only useful for Google's backend
right now, and we don't use the legacy format at Google. In
particular, I don't want `MergedTree::value()` to have to be async.
The commit backend at Google is cloud-based (and so are the other
backends); it reads and writes commits from/to a server, which stores
them in a database. That makes latency much higher than for disk-based
backends. To reduce the latency, we have a local daemon process that
caches and prefetches objects. There are still many cases where
latency is high, such as when diffing two uncached commits. We can
improve that by changing some of our (jj's) algorithms to read many
objects concurrently from the backend. In the case of tree-diffing, we
can fetch one level (depth) of the tree at a time. There are several
ways of doing that:
* Make the backend methods `async`
* Use many threads for reading from the backend
* Add backend methods for batch reading
I don't think we typically need CPU parallelism, so it's wasteful to
have hundreds of threads running in order to fetch hundreds of objects
in parallel (especially when using a synchronous backend like the Git
backend). Batching would work well for the tree-diffing case, but it's
not as composable as `async`. For example, if we wanted to fetch some
commits at the same time as we were doing a diff, it's hard to see how
to do that with batching. Using async seems like our best bet.
I didn't make the backend interface's write functions async because
writes are already async with the daemon we have at Google. That
daemon will hash the object and immediately return, and then send the
object to the server in the background. I think any cloud-based
solution will need a similar daemon process. However, we may need to
reconsider this if/when jj gets used on a server with a custom backend
that writes directly to a database (i.e. no async daemon in between).
I've tried to measure the performance impact. That's the largest
difference I've been able to measure was on `jj diff
--ignore-working-copy -s --from v5.0 --to v6.0` in the Linux repo,
which increases from 749 ms to 773 ms (3.3%). In most cases I've
tested, there's no measurable difference. I've tried diffing from the
root commit, as well as `jj --ignore-working-copy log --no-graph -r
'::v3.0 & author(torvalds)' -T 'commit_id ++ "\n"'` (to test a
commit-heavy load).
When we start writing tree-level conflicts in an existing repo, we
don't want commits that change the format to be non-empty if they
don't change any content. This patch updates `MergeTreeId::eq()` to
consider two resolved trees equal even if only their `MergedTreeId`
variant is different (one is path-level and one is tree-level).
I think I've gone through all places we compare tree ids and checked
that it's safe to compare them this way. One consequence is that
rebasing a commit without changing the parents (typically
auto-rebasing after `jj describe`) will not lead to the tree id
getting upgraded, due to an optimization we have for that case. I
don't think that's serious enough to handle specially; we'll have to
support the old format for existing repos for a while regardless of a
few commits not getting upgraded right away.
The number of failing tests with the config option enabled drop from
108 to 11 with this patch.
We currently represent the root tree id in a commit by `Merge<TreeId>`
plus a boolean `uses_tree_conflict_format`. It's better to use an enum
for that. That makes it harder to forget to check which type of tree
it is, and it makes it impossible to store a legacy tree with multiple
ids (as we could with `uses_tree_conflict_format=false`,
`root_tree=Merge::new(...)`).
Maybe more importantly, we're also going to want to pass around this
information in most places where we currently pass a single `TreeId`,
and passing two separate values would be annoying.
One of the error types that I later created embedded `BackendError`, but `clippy` complained that the size of the type was too large. This helps address that.
Since `Conflict<T>` can also represent a non-conflict state (a single
term), `Merge<T>` seems like better name.
Thanks to @ilyagr for the suggestion in
https://github.com/martinvonz/jj/pull/1774#discussion_r1257547709
Sorry about the churn. It would have been better if I thought of this
name before I introduced `Conflict<T>`.
Tree-level conflicts (#1624) will be stored as multiple trees
associated with a single commit. This patch adds support for that in
`backend::Commit` and in the backends.
When the Git backend writes a tree conflict, it creates a special root
tree for the commit. That tree has only the individual trees from the
conflict as subtrees. That way we prevent the trees from getting
GC'd. We also write the tree ids to the extra metadata table
(i.e. outside of the Git repo) so we don't need to load the tree
object to determine if there are conflicts.
I also added new flag to `backend::Commit` indicating whether the
commit is a new-style commit (with support for tree-level
conflicts). That will help with the migration. We will remove it once
we no longer care about old repos. When the flag is set, we know that
a commit with a single tree cannot have conflicts. When the flag is
not set, it's an old-style commit where we have to walk the whole tree
to find conflicts.
Errors that may occur while loading backend would vary per backends, and
it's unlikely that these errors could be mapped to BackendError variants
other than BackendError::Other. So let's extract Other(_) of that kind as
a separate type to clarify there would be no other error variants.
Perhaps, Backend/Error will be renamed to CommitBackend/Error or
CommitStore/Error?, whereas I think BackendInit/LoadError can be shared
among store factories.
This is much simpler and I was slightly surprised that it doesn't have
much impact on performance. I tried `jj --ignore-working-copy diff -s
--from root --to v5.15` in the Linux kernel repo, and there was
perhaps a 1.5% slowdown (508 ms -> 515 ms). In more normal cases (like
diffing a single commit against its parent), I couldn't measure any
difference at all.
I don't think we'll want to record a label for each term, because such
labels would get stale, and it seems hard to make them make sense
after transferring a remote to another repo. I think we'll probably
want to infer labels on demand instead (#1176).
The internal backend at Google doesn't let you write any value you
want for in the committer field. The `Store` type still caches the
value it attempted to write, which gets a little weird when the
written value is not what we tried to write. We should use the value
the backend actually wrote. However, we don't know if the backend
changed anything without reading the value back, which is often
wasteful. This commit changes the API to return the written value.
I only changed the signature of `write_commit()` for now. Maybe we
should make a similar change to `write_tree()`.
This has several advantages:
* Makes it possible to downcast to non-Git custom backends (might be
useful at Google, but we haven't needed it yet)
* Lets us access more specific functionality on the `GitBackend`,
making it possible to access the `git2::Repository` without
creating a copy of it.
* Removes the dependency on Git from the backend
It took a while before I realized that conflicts could be modeled as
simple algebraic expressions with positive and negative terms (they
were modeled as recursive 3-way conflicts initially). We've been
thinking of them that way for a while now, so let's make the
`ConflictPart` name match that model.
Our internal backend at Google uses a 32-byte change id, so I'd like
to make the backend able to decide the length. To start with, let's
make the backend able to decide what the root change id should
be. That's consistent with how we already let the backend decide what
the root commit id should be.
The function is currently only about the length of commit IDs, so
let's clarify that. I'm going to add another function for the length
of change IDs next. I don't know if we're going to care about lengths
of other hashes in the future. We might even be able to remove the
current restriction that all commit IDs and all change IDs have the
same length.
This helps us to migrate commit_id index to ReadonlyIndex. For large
repositories, this also reduces initialization cost, but that's not the main
intent of this change.
https://github.com/martinvonz/jj/pull/1041#issuecomment-1399225876
common_hex_len() and iter_half_bytes() are added to backend.rs since more
call sites will be added to index.rs, and I feel index.rs isn't a good place
to host this kind of utility functions.
I made it a free function. Alternatively, the root id could be instantiated
by and obtained through backend, but I don't think we'll need such level of
abstraction.
I'm going to add a workaround for shortest prefix calculation of the root ids,
where this function will be used.
Dereferencing `self` as `*self` in order to perform patten-matching
using `ref` is unnecessary and will be done automatically by the
compiler (match ergonomics, introduced in Rust 1.26).
I needed this in the course of debugging an error. Before this commit, the error looked like this:
```
Error: Unexpected error from backend: Object not found
```
After this commit, it looks like this:
```
Error: Unexpected error from backend: Object with CommitId 8f59646bc9bb6bb44b5624f1248f4a708f37003c not found: object not found - no match for id (8f59646bc9bb6bb44b5624f1248f4a708f37003c); class=Odb (9); code=NotFound (-3)
```
Strictly speaking, we could rely on e.g. `git2::Oid::from_str` to produce an error, but I figure that having an explicit error for a mismatching hash length might demystify some error condition in the future, since commit IDs and change IDs and potentially other backends' IDs may have different lengths, so this could flag a mismatch earlier/more obviously.
It's unlikely we'll need to customize these impls per type, so let's ensure
that these newtypes have identical implementations. This commit also adds
from_hex() to FileId, SymlinkId, and ConflictId.
Let's acknowledge everyone's contributions by replacing "Google LLC"
in the copyright header by "The Jujutsu Authors". If I understand
correctly, it won't have any legal effect, but maybe it still helps
reduce concerns from contributors (though I haven't heard any
concerns).
Google employees can read about Google's policy at
go/releasing/contributions#copyright.
There are no "non-normal" files, so "normal" is not needed. We have
symlinks and conflicts, but they are not files, so I think just "file"
is unambiguous.
I left `testutils::write_normal_file()` because there it's used to
mean "not executable file" (there's also a `write_executable_file()`).
I left `working_copy::FileType::Normal` since renaming `Normal` there
to `File` would also suggest we should rename `FileType`, and I don't
know what would be a better name for that type.
Decouples view/operation IDs from serialized forms, which are not
necessarily stable. Not breaking as these IDs are persistent, never
recomputed or used for integrity checking.
I was reading a draft of "Git Rev News: Edition 91" [1] where Peff
mentions some unfinished patches to allow negative timestamps in
Git. So I figured I should add support for that before I forget. I
haven't checked if libgit2 supports it, so it might be that our Git
backend still doesn't support it after this patch.
[1] https://github.com/git/git.github.io/blob/master/rev_news/drafts/edition-91.md
We currently determine if the repo uses the Git backend or the local
backend by checking for presence of a `.jj/repo/store/git_target`
file. To make it easier to add out-of-tree backends, let's instead add
a file that indicates which backend to use.
I had made the backends unaware of the virtual root commit because
they don't need to know about it, and we could avoid some duplicated
code by putting that in `Store` instead. However, as we saw in
b21a123bc894, the root commit being virtual has some user-visible
effects (they can't create a merge with the root and some other
commit). So I'm thinking that we may want to make the root commit an
actual commit, depending on which backend is used. Specificially, when
using the Git backend, we cannot record the root commit as an actual
parent since Git would fail when trying to look it up. Backends that
don't need compatibility can make the root commit an actual commit,
however.
This commit therefore makes the backends aware of the root commit. It
makes it remain a virtual commit in the Git backend, and makes it an
actual commit in the `LocalBackend`.
This commit breaks any existing repos using the `LocalBackend`, but
there shouldn't be any such repos other than for testing.
