This reveals another bug I found.

This change should not affect the test at all, but does.

At this point, it works more or less correctly. However, in the next
commit adding another `jj push` breaks this test.

This will make it more similar to the test introduced in the next commit.

If it doesn't exist, generate it from git refs as jj did
before. This should work as long as the last operation wasn't
"undo" or "restore".

This is a temporary behavior designed for people upgrading
from older jj. It should be eventually removed.

- import happens on fetch. Does/should the file be updated? Should export happen on push?
- how does this work if the user does `git push` instead of `jj git push`. It does seem to work :)

The function will be reused by `jj git import` shortly.

This commit reverts 8a440d8 AKA https://github.com/martinvonz/jj/pull/832/commits/4dfd7651798d5cafcdb66e86ab064d6e7dd01ce5

The problem that commit is that the Git repo is, unaffected by `jj undo`. So,
it is important for JJ's view of the git refs to be unaffected by `jj undo`.

Fixes #922. Also creates another bug that is fixed in a follow-up commit.

They seem to be unused, but the new combined error will be used in the next
commit.

Looks like "change_id normal" + color. Picture for `jj obslog`:

https://user-images.githubusercontent.com/4123047/230708271-4108cf5f-255d-419e-bd3e-fc97dc8b8660.png

This should also be occasionally be useful for e.g. `jj log commit_id`.

I also considered the wording `(Was change_id)` or `change_id (old)`, `change_id (obs)`.

The "default" color resets the terminal color to default.

We'll remove All, so it makes sense to not have None either.

New ResolvedExpression enum ensures that the evaluation engine doesn't have
to know the symbol resolution details. In this commit, I've moved Filter
and NotIn resolution to resolve_visibility(). Implicit All/VisibleHeads
resolution will be migrated later.

It's tempting to combine resolve_symbols() and resolve_visibility() to get
rid of panic!()s, but the resolution might have to be two passes to first
resolve&collect explicit commit ids, and then substitute "all()" with
"(:visible_heads())|commit_id|..". It's also possible to apply some tree
transformation after symbol resolution.

I'll make the resolution stage mandatory, and have it return a "resolved"
type. RevsetExpression::evaluate() will be moved to the "resolved" type.

This makes it clear what should be resolved at resolve_symbols(). Symbol
is a bit special while parsing function arguments, but it's no different
than the other unresolved references at expression level.

I'm going to merge unresolved variants as RevsetExpression::CommitRef(_).
This prepares for the change.

It helps while debugging tree substitution.

The current use of `timestamp.ago()` in the default template makes the
tests depend on the current time, which they shouldn't.

It works because a build command is line oriented, but it's technically
wrong to include "\n" in git_hash.

Currently, the hash becomes colored if jj is used to get it at
build time and the user configured `color="always"` in jj's
config.

It's slightly faster, and removes the use of RevsetExpression::descendants()
API.

We wouldn't care for the cost of virtual dispatch at this level, but I
think a concrete struct type is easier to deal with than trait object.

This also removes the use of RevsetExpression::connected() API from the
evaluation engine.

The returned reachable set can be reused to calculate roots() expression.

I'm thinking of transforming RevsetExpression to a enum dedicated for
the evaluation stage. To help the migration, I want to remove the use of
the RevsetExpression builder API from the evaluation engine.

Fewer virtual dispatch is also better.

This changes the version number reported by `jj version` from "0.7.0"
to something like
"0.7.0-24a512683bc921699575b6a953624b05c068d544a". The hash is added
if running in a jj repo or a git repo.

Bumps [github/codeql-action](https://github.com/github/codeql-action) from 2.2.10 to 2.2.11.
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](https://github.com/github/codeql-action/compare/8c8d71dde4abced210732d8486586914b97752e8...d186a2a36cc67bfa1b860e6170d37fb9634742c7)

---
updated-dependencies:
- dependency-name: github/codeql-action
dependency-type: direct:production
update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

This adds `conflict()` revset that selects commits with conflicts. We
may want to extend it later to consider only conflicts at certain
paths.

Bumps [github/codeql-action](https://github.com/github/codeql-action) from 2.2.9 to 2.2.10.
- [Release notes](https://github.com/github/codeql-action/releases)
- [Changelog](https://github.com/github/codeql-action/blob/main/CHANGELOG.md)
- [Commits](https://github.com/github/codeql-action/compare/04df1262e6247151b5ac09cd2c303ac36ad3f62b...8c8d71dde4abced210732d8486586914b97752e8)

---
updated-dependencies:
- dependency-name: github/codeql-action
dependency-type: direct:production
update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

Bumps [EmbarkStudios/cargo-deny-action](https://github.com/EmbarkStudios/cargo-deny-action) from 1.5.0 to 1.5.1.
- [Release notes](https://github.com/EmbarkStudios/cargo-deny-action/releases)
- [Commits](https://github.com/EmbarkStudios/cargo-deny-action/compare/8af37f5d0c77859223226de776036f7f691e1469...3d26fd4e2adf837aee64b12f86d4dabb799fcfa2)

---
updated-dependencies:
- dependency-name: EmbarkStudios/cargo-deny-action
dependency-type: direct:production
update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

With this change, the output looks like this:

```
3% 165.9 Mi at 13.4 MiB/s [█▋ ]
```

Closes #1483.

TreeDiffIterator wouldn't load identical subtrees, but it's up to caller to
optimize out the root tree loading.

Bumps [libc](https://github.com/rust-lang/libc) from 0.2.140 to 0.2.141.
- [Release notes](https://github.com/rust-lang/libc/releases)
- [Commits](https://github.com/rust-lang/libc/compare/0.2.140...0.2.141)

---
updated-dependencies:
- dependency-name: libc
dependency-type: direct:production
update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

Such branches lead to confusing errors on git expoert or push.

This prevents an editor opening, and is useful in scripts/tests.

The `heads()` revset function with one argument is the counterpart to
`roots()`. Without arguments, it returns the visible heads in the
repo, i.e. `heads(all())`. The two use cases are quite different, and
I think it would be good to clarify that the no-arg form returns the
visible heads, so let's split that out to a new `visible_heads()`
function.

This basically removes hidden 'all() &' from union/negation of filters. To
achieve that, I have two options: 1. add separate evaluation path (like the
one this commit introduced), or 2. wrap "all()" revset to override predicate
as Box::new(|_| true) function. I took the former since it's less ad-hoc.

We can add an explicit RevsetExpression node to branch between evaluate()
and evaluate_predicate(), but I don't think it would simplify the
implementation at this point. We might need such node if we want to resolve
"all()" at resolve_symbols(). It might be even better to extract a subset of
RevsetExpression enum, which only contains evaluatable nodes.

The cost of 'all() &' isn't significant for most filters. '~merges()' is
the exception. For jj repo,

revsets/:v0.3.0 & (author(martinvonz) | committer(martinvonz))
--------------------------------------------------------------
base 1.06 11.2±0.04m
new 1.00 10.5±0.05m

revsets/~merges()
-----------------
base 1.69 750.0±8.47µ
new 1.00 444.1±3.50µ

I think this is more readable, and apparently it produces slightly better code
maybe because the compiler can determine that there are no unwanted markers.

Since filter is slow in general, its input set should be minimized. This has
measurable impact on artificial query like '~(v0.4.0..) & author(_)'. If it
were evaluated as a difference of sets, all commits would have to be loaded.

This document is meant to be a record of how we think about Git
submodules (and not a _current_ implementation of submodules). We will
fill it out incrementally as we get a clearer idea of what we want
submodules to look like.

As an initial version, I started with (IMO) the least controversial
points:

- We want to support most workflows Git submodules users are accustomed
to.

- A roadmap that allows us to incrementally roll out Git submodule
functionality (instead of having to boil the ocean).

Bumps [pest_derive](https://github.com/pest-parser/pest) from 2.5.6 to 2.5.7.
- [Release notes](https://github.com/pest-parser/pest/releases)
- [Commits](https://github.com/pest-parser/pest/compare/v2.5.6...v2.5.7)

---
updated-dependencies:
- dependency-name: pest_derive
dependency-type: direct:production
update-type: version-update:semver-patch
...

Signed-off-by: dependabot[bot] <support@github.com>

Bumps [test-case](https://github.com/frondeus/test-case) from 3.0.0 to 3.1.0.
- [Release notes](https://github.com/frondeus/test-case/releases)
- [Changelog](https://github.com/frondeus/test-case/blob/master/CHANGELOG.md)
- [Commits](https://github.com/frondeus/test-case/compare/v3.0.0...v3.1.0)

---
updated-dependencies:
- dependency-name: test-case
dependency-type: direct:production
update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

Bumps [git2](https://github.com/rust-lang/git2-rs) from 0.16.1 to 0.17.0.
- [Release notes](https://github.com/rust-lang/git2-rs/releases)
- [Changelog](https://github.com/rust-lang/git2-rs/blob/master/CHANGELOG.md)
- [Commits](https://github.com/rust-lang/git2-rs/compare/0.16.1...git2-curl-0.17.0)

---
updated-dependencies:
- dependency-name: git2
dependency-type: direct:production
update-type: version-update:semver-minor
...

Signed-off-by: dependabot[bot] <support@github.com>

This follows up on 5c703aeb0397b560a383bdd57a87235834074b64.

The only reason for this change is that, subjectively, the result looks better to me. I'm not sure why, but I couldn't get used to the old symbol in spite of its seeming reasonableness. It felt really bold and heavy.

If people agree, we can wait until we need to update the screenshots for some other reason before merging this. Sorry I didn't figure this out while the discussion about the referenced commit was going on.

I'm not 100% certain how many fonts support each symbol. Please try it out and let me know if it doesn't work for you.

Compare after:


and before:


It no longer handles RevsetResolutionError.

Since resolve_symbols() now removes Present(_) node, it make sense to
handle symbol resolution error there. That's why I added a "pre" callback
to try_transform_expression().

Perhaps, "operation" scope (#1283) can be implemented in a similar way,
(but somehow need to resolve operation id and call repo.reload_at(op).)

This will be a hook for resolve_symbols() to transform Present(_) subtree.

I'm going to parameterize error type of TransformResult, and the result type
will be replaced with Result<TransformedExpression, E>.