This PR fixes several inlay hint race conditions that could result in a
violation of the monotonic progress assumption, introduced in #7149.
Specifically:
- In rare circumstances, it could happen that stateful LSP requests were
executed out-of-order with their `didChange` handlers, as both requests
and the `didChange` handlers waited on `lake setup-file` to complete,
with the latter running those handlers in a dedicated task afterwards.
This meant that a request could be added to the stateful LSP handler
request queue before the corresponding `didChange` call that actually
came before it. This PR resolves this issue by folding the task that
waits for `lake setup-file` into the `RequestContext`, which ensures
that we only need to wait for it when actually executing the request
handler.
- While #7164 fixed the monotonic progress assertion violation that was
caused by `$/cancelRequest`, it did not account for our internal notion
of silent request cancellation in stateful LSP requests, which we use to
cancel the inlay hint edit delay when VS Code fails to emit a
`$/cancelRequest` notification. This issue is resolved by always
producing the full finished prefix of the command snapshot queue, even
on cancellation. Additionally, this also fixes an issue where in the
same circumstances, the language server could produce an empty inlay
hint response when a request was cancelled by our internal notion of
silent request cancellation.
- For clients that use `fullChange` `didChange` notifications (e.g. not
VS Code), we would get several aspects of stateful LSP request
`didChange` state handling wrong, which is also addressed by this PR.
Before/after:
```
make -C build/release test ARGS="-j$(nproc) -R interactive" 208.10s user 20.93s system 1982% cpu 11.552 total
make -C build/release test ARGS="-j$(nproc) -R interactive" 87.22s user 22.58s system 1454% cpu 7.548 total
```
This PR fixes an `Elab.async` regression where elaboration tasks are
cancelled on document edit even though their result may be reused in the
new document version, reporting an incomplete result.
While this PR fixes the functional regression, it does so as an
over-approximation by never cancelling such tasks. A follow-up PR will
implement the correct behavior of only cancelling the tasks that are not
reused.
This PR changes the server to run `lake setup-file` on Lake
configuration files (e.g., `lakefile.lean`).
This is needed to support Lake passing the server its own Lake plugin to
load when elaborating the configuration file.
This PR ensures that all tasks in the language server either use
dedicated tasks or reuse an existing thread from the thread pool. This
ensures that elaboration tasks cannot prevent language server tasks from
being scheduled. This is especially important with parallelism right
around the corner and elaboration becoming more likely to starve the
language server of computation, which could drive up language server
latencies significantly on machines with few cores.
Specifically, all language server tasks are refactored to use a new thin
`ServerTask` API wrapper with a single "costly" vs "cheap" dimension,
where costly tasks are always scheduled as dedicated tasks, and cheap
tasks are always made to either run on the calling thread or to reuse
the thread of the task being mapped on by using the `sync` flag.
ProofWidgets4 adaption PR:
https://github.com/leanprover-community/ProofWidgets4/pull/106
### Other changes
- This PR makes several tasks dedicated that weren't dedicated before,
and uses `sync := true` for some others. The rules for this are
described in the module docstring of `ServerTask.lean`.
- Most notably, the reporting task in the file worker was *not* a
dedicated task before this PR, which could easily lead to thread pool
starvation on successive changes. It also did not support cancellation.
This PR ensures that it does.
### Breaking changes
- `RequestTask` and the request-oriented snapshot API are refactored to
use `ServerTask` instead of `Task`. All functions in `Task` have close
analogues in `ServerTask`, and functions on `RequestTask` now need to
distinguish between whether a `map` or a `bind` is cheap or costly. This
affects all downstream users of `RequestM`, e.g. tools that extend the
language server with their own requests, or some users of the RPC
mechanism.
- The following unused functions of the `AsyncList` API have been
deleted: `append`, `unfoldAsync`, `getAll`, `waitHead?`, `cancel`
This PR adds a fast path to the inlay hint request that makes it re-use
already computed inlay hints from previous requests instead of
re-computing them. This is necessary because for some reason VS Code
emits an inlay hint request for every line you scroll, so we need to be
able to respond to these requests against the same document state
quickly. Otherwise, every single scrolled line would result in a request
that can take a few dozen ms to be responded to in long files, putting
unnecessary pressure on the CPU.
It also filters the result set by the inlay hints that have been
requested.
This PR significantly improves the performance of auto-completion by
optimizing individual requests by a factor of ~2 and by giving language
clients like VS Code the opportunity to reuse the state of previous
completion requests, thus greatly reducing the latency for the
auto-completion list to update when adding more characters to an
identifier.
In my testing:
- The latency of completing `C` in a file with `import Mathlib` was
reduced from ~1650ms to ~800ms
- The latency of completing `Cat` in a file with `import Mathlib` was
reduced from ~800ms to ~430ms
- The latency of completing dot notation was mostly unaffected
- Successive completions are now practically instant, e.g. if we were to
complete `C` and then type it out to `Cat`, before it would take roughly
~1650ms + ~800ms, whereas now there is only a significant latency for
completing `C` (~800ms) and the completion list is updated practically
instantly when typing out `Cat`.
<details>
<summary>(Video) Auto-completion latency before this PR</summary>
</details>
<details>
<summary>(Video) Auto-completion latency after this PR</summary>
</details>
In detail, this PR makes the following changes:
- Set `isIncomplete` to `false` in non-synthetic completion responses so
that the client can re-use these completion states.
- Replace the server side fuzzy matching with a simple and fast check
that all characters in the identifier thus far are present in the same
order in the declaration to match against. There are some examples where
the simple and fast check yields a completion item that the fuzzy
matching would filter, but since VS Code filters the completion items
with its own fuzzy matching after that anyways, these extra completion
items are never actually displayed to the user.
- Remove all notions of scoring and sorting completion items from the
language server. We now rely entirely on the client to sort the
completion items as it sees fit. In my testing, the only significant
change as a result of this is that while the language server would
sometimes penalize namespaces with lots of components, VS Code instead
uses a strictly alphabetic order. Even before this change, we never
actually really prioritized local variables over global variables, so
the penalty wasn't very helpful in practice. We might add some small
form of local variable prioritization in the future, though.
- Remove the empty completion list hack that was introduced in #1885. It
does not appear to be necessary anymore.
This PR moves away from using `List.get` / `List.get?` / `List.get!` and
`Array.get!`, in favour of using the `GetElem` mediated getters. In
particular it deprecates `List.get?`, `List.get!` and `Array.get?`. Also
adds `Array.back`, taking a proof, matching `List.getLast`.
This PR adds language server support for request cancellation to the
following expensive requests: Code actions, auto-completion, document
symbols, folding ranges and semantic highlighting. This means that when
the client informs the language server that a request is stale (e.g.
because it belongs to a previous state of the document), the language
server will now prematurely cancel the computation of the response in
order to reduce the CPU load for requests that will be discarded by the
client anyways.
This PR fixes a bug where the goal state selection would sometimes
select incomplete incremental snapshots on whitespace, leading to an
incorrect "no goals" response. Fixes#6594, a regression that was
originally introduced in 4.11.0 by #4727.
The fundamental cause of #6594 was that the snapshot selection would
always select the first snapshot with a range that contains the cursor
position. For tactics, whitespace had to be included in this range.
However, in the test case of #6594, this meant that the snapshot
selection would also sometimes pick a snapshot before the cursor that
still contains the cursor in its whitespace, but which also does not
necessarily contain all the information needed to produce a correct goal
state. Specifically, at the `InfoTree`-level, when the cursor is in
whitespace, we distinguish competing goal states by their level of
indentation. The snapshot selection did not have access to this
information, so it necessarily had to do the wrong thing in some cases.
This PR fixes the issue by adjusting the snapshot selection for goals to
explicitly account for whitespace and indentation, and refactoring the
language processor architecture to thread enough information through to
the snapshot selection so that it can decide which snapshots to use
without having to force too many tasks, which would destroy
incrementality in goal state requests.
Specifically, this PR makes the following adjustments:
- Refactor `SnapshotTask` to contain both a `Syntax` and a `Range`.
Before, `SnapshotTask`s had a single range that was used both for
displaying file progress information and for selecting snapshots in
server requests. For most snapshots, this range did not include
whitespace, though for tactics it did. Now, the `reportingRange` field
of `SnapshotTask` is intended exclusively for reporting file progress
information, and the `Syntax` is used for selecting snapshots in server
requests. Importantly, the `Syntax` contains the full range information
of the snapshot, i.e. its regular range and its range including
whitespace.
- Adjust all call-sites of `SnapshotTask` to produce a reasonable
`Syntax`.
- Adjust the goal snapshot selection to account for whitespace and
indentation, as the `InfoTree` goal selection does.
- Fix a bug in the snapshot tree tracing that would cause it to render
the `Info` of a snapshot at the wrong location when `trace.Elab.info`
was also set.
This PR is based on #6329.
This PR enables the language server to present multiple disjoint line
ranges as being worked on. Even before parallelism lands, we make use of
this feature to show post-elaboration tasks such as kernel checking on
the first line of a declaration to distinguish them from the final
tactic step.
This PR fixes a bug where both the inlay hint change invalidation logic
and the inlay hint edit delay logic were broken in untitled files.
Thanks to @Julian for spotting this!
This PR implements a number of refinements for the auto-implicit inlay
hints implemented in #6768.
Specifically:
- In #6768, there was a bug where the inlay hint edit delay could
accumulate on successive edits, which meant that it could sometimes take
much longer for inlay hints to show up. This PR implements the basic
infrastructure for request cancellation and implements request
cancellation for semantic tokens and inlay hints to resolve the issue.
With this edit delay bug fixed, it made more sense to increase the edit
delay slightly from 2000ms to 3000ms.
- In #6768, we applied the edit delay to every single inlay hint request
in order to reduce the amount of inlay hint flickering. This meant that
the edit delay also had a significant effect on how far inlay hints
would lag behind the file progress bar. This PR adjusts the edit delay
logic so that it only affects requests sent directly after a
corresponding `didChange` notification. Once the edit delay is used up,
all further semantic token requests are responded to without delay, so
that the only latency that affects how far the inlay hints lag behind
the progress bar is how often we emit refresh requests and how long VS
Code takes to respond to them.
- For inlay hints, refresh requests are now emitted 500ms after a
response to an inlay hint request, not 2000ms, which means that after
the edit delay, inlay hints should only lag behind the progress bar by
about up to 500ms. This is justifiable for inlay hints because the
response should be much smaller than e.g. is the case for semantic
tokens.
- In #6768, 'Restart File' did not prompt a refresh, but it does now.
- VS Code does not immediately remove old inlay hints from the document
when they are applied. In #6768, this meant that inlay hints would
linger around for a bit once applied. To mitigate this issue, this PR
adjusts the inlay hint edit delay logic to identify edits sent from the
client as being inlay hint applications, and sets the edit delay to 0ms
for the inlay hint requests following it. This means that inlay hints
are now applied immediately.
- In #6768, hovering over single-letter auto-implicit inlay hints was a
bit finicky because VS Code uses the regular cursor icon on inlay hints,
not the thin text cursor icon, which means that it is easy to put the
cursor in the wrong spot. We now add the separation character (` ` or
`{`) preceding an auto-implicit to the hover range as well, which makes
hovering over inlay hints much smoother.
This PR adds support for plugins to the frontend and server.
Implementation-wise, this adds a `plugins` argument to `runFrontend`,
`processHeader`, amd `importModules`, a `plugins` field to
`SetupImportsResult` and `FileSetupResult`. and a `pluginsPath` field to
`LeanPaths`, and then threads the value through these.
This PR adds preliminary support for inlay hints, as well as support for
inlay hints that denote the auto-implicits of a function. Hovering over
an auto-implicit displays its type and double-clicking the auto-implicit
inserts it into the text document.
This PR is an extension of #3910.
### Known issues
- In VS Code, when inserting an inlay hint, the inlay hint may linger
for a couple of seconds before it disappears. This is a defect of the VS
Code implementation of inlay hints and cannot adequately be resolved by
us.
- When making a change to the document, it may take a couple of seconds
until the inlay hints respond to the change. This is deliberate and
intended to reduce the amount of inlay hint flickering while typing. VS
Code has a mechanism of its own for this, but in my experience it is
still far too sensitive without additional latency.
- Inserting an auto-implicit inlay hint that depends on an auto-implicit
meta-variable causes a "failed to infer binder type" error. We can't
display these meta-variables in the inlay hint because they don't have a
user-displayable name, so it is not clear how to resolve this problem.
- Inlay hints are currently always resolved eagerly, i.e. we do not
support the `textDocument/inlayHint/resolve` request yet. Implementing
support for this request is future work.
### Other changes
- Axioms did not support auto-implicits due to an oversight in the
implementation. This PR ensures they do.
- In order to reduce the amount of inlay hint flickering when making a
change to the document, the language server serves old inlay hints for
parts of the file that have not been processed yet. This requires LSP
request handler state (that sometimes must be invalidated on
`textDocument/didChange`), so this PR introduces the notion of a
stateful LSP request handler.
- The partial response mechanism that we use for semantic tokens, where
we simulate incremental LSP responses by periodically emitting refresh
requests to the client, is generalized to accommodate both inlay hints
and semantic tokens. Additionally, it is made more robust to ensure that
we never emit refresh requests while a corresponding request is in
flight, which causes VS Code to discard the respond of the request, as
well as to ensure that we keep prompting VS Code to send another request
if it spuriously decides not to respond to one of our refresh requests.
- The synthetic identifier of an `example` had the full declaration as
its (non-canonical synthetic) range. Since we need a reasonable position
for the identifier to insert an inlay hint for the auto-implicits of an
`example`, we change the (canonical synthetic) range of the synthetic
identifier to that of the `example` keyword.
- The semantic highlighting request handling is moved to a separate
file.
### Breaking changes
- The semantic highlighting request handler is not a pure request
handler anymore, but a stateful one. Notably, this means that clients
that extend the semantic highlighting of the Lean language server with
the `chainLspRequestHandler` function must now use the
`chainStatefulLspRequestHandler` function instead.
This PR uniformizes the naming of `enum`/`enumFrom` (on `List`) and
`zipWithIndex` (on `Array` on `Vector`), replacing all with `zipIdx`. At
the same time, we generalize to add an optional `Nat` parameter for the
initial value of the index (which previously existed, only for `List`,
as the separate function `enumFrom`).
This PR fixes a significant auto-completion performance regression that
was introduced in #5666, i.e. v4.14.0.
#5666 introduced tactic docstrings, which were attempted to be collected
for every single completion item. This is slow for hundreds of thousands
of completion items. To fix this, this PR moves the docstring
computation into the completion item resolution, which is only called
when users select a specific completion item in the UI.
A downside of this approach is that we currently can't test completion
item resolution, so we lose a few tests that cover docstrings in
completions in this PR.
Avoids build time overhead until the option is proven to speed up
average projects. Adds Init.Prelude (many tiny declarations, "worst
case") and Init.List.Sublist (many nontrivial theorems, "best case")
under -DElab.async=true as new benchmarks for tracking.
This PR allows the dot ident notation to resolve to the current
definition, or to any of the other definitions in the same mutual block.
Existing code that uses dot ident notation may need to have `nonrec`
added if the ident has the same name as the definition.
Closes#6601
This PR makes it harder to create "fake" theorems about definitions that
are stubbed-out with `sorry` by ensuring that each `sorry` is not
definitionally equal to any other. For example, this now fails:
```lean
example : (sorry : Nat) = sorry := rfl -- fails
```
However, this still succeeds, since the `sorry` is a single
indeterminate `Nat`:
```lean
def f (n : Nat) : Nat := sorry
example : f 0 = f 1 := rfl -- succeeds
```
One can be more careful by putting parameters to the right of the colon:
```lean
def f : (n : Nat) → Nat := sorry
example : f 0 = f 1 := rfl -- fails
```
Most sources of synthetic sorries (recall: a sorry that originates from
the elaborator) are now unique, except for elaboration errors, since
making these unique tends to cause a confusing cascade of errors. In
general, however, such sorries are labeled. This enables "go to
definition" on `sorry` in the Infoview, which brings you to its origin.
The option `set_option pp.sorrySource true` causes the pretty printer to
show source position information on sorries.
**Details:**
* Adds `Lean.Meta.mkLabeledSorry`, which creates a sorry that is labeled
with its source position. For example, `(sorry : Nat)` might elaborate
to
```
sorryAx (Lean.Name → Nat) false
`lean.foo.12.8.12.13.8.13._sorry._@.lean.foo._hyg.153
```
It can either be made unique (like the above) or merely labeled. Labeled
sorries use an encoding that does not impact defeq:
```
sorryAx (Unit → Nat) false (Function.const Lean.Name ()
`lean.foo.14.7.13.7.13.69._sorry._@.lean.foo._hyg.174)
```
* Makes the `sorry` term, the `sorry` tactic, and every elaboration
failure create labeled sorries. Most are unique sorries, but some
elaboration errors are labeled sorries.
* Renames `OmissionInfo` to `DelabTermInfo` and adds configuration
options to control LSP interactions. One field is a source position to
use for "go to definition". This is used to implement "go to definition"
on labeled sorries.
* Makes hovering over a labeled `sorry` show something friendlier than
that full `sorryAx` expression. Instead, the first hover shows the
simplified ``sorry `«lean.foo:48:11»``. Hovering over that hover shows
the full `sorryAx`. Setting `set_option pp.sorrySource true` makes
`sorry` always start with printing with this source position
information.
* Removes `Lean.Meta.mkSyntheticSorry` in favor of `Lean.Meta.mkSorry`
and `Lean.Meta.mkLabeledSorry`.
* Changes `sorryAx` so that the `synthetic` argument is no longer
optional.
* Gives `addPPExplicitToExposeDiff` awareness of labeled sorries. It can
set `pp.sorrySource` when source positions differ.
* Modifies the delaborator framework so that delaborators can set Info
themselves without it being overwritten.
Incidentally closes#4972.
Inspired by [this Zulip
thread](https://leanprover.zulipchat.com/#narrow/channel/287929-mathlib4/topic/Is.20a.20.60definition_wanted.60.20keyword.20possible.3F/near/477260277).
This PR moves `IO.Channel` and `IO.Mutex` from `Init` to `Std.Sync` and
renames them to `Std.Channel` and `Std.Mutex`.
Note that the original files are retained and the deprecation is written
manually as we cannot import `Std` from `Init` so this is the only way
to deprecate without a hard breaking change. In particular we do not yet
move `Std.Queue` from `Init` to `Std` both because it needs to be
retained for this deprecation to work but also because it is already
within the `Std` namespace and as such we cannot maintain two copies of
the file at once. After the deprecation period is finished `Std.Queue`
will find a new home in `Std.Data.Queue`.
This PR fixes a bug in structure instance field completion that caused
it to not function correctly for bracketed structure instances written
in Mathlib style.
This PR fixes a performance issue where the Lean language server would
walk the full project file tree every time a file was saved, blocking
the processing of all other requests and notifications and significantly
increasing overall language server latency after saving.
This issue was originally reported at
https://leanprover.zulipchat.com/#narrow/channel/270676-lean4/topic/Compiling.20too.20slow/near/484386515
and uncovered in a lengthy investigation. The performance bug that
causes the Lean language server to walk the full project file tree when
the file watcher for .ilean files is triggered was introduced when the
.ileans were first introduced, whereas the specific issue of file saving
also triggering the walk was introduced by #3247 in 4.8.0 and the use of
the file watcher for .lean files, which would then also trigger the
directory walk. Combining this with VS Code's auto-save feature causes
the language server to walk the full project file tree on every change
of the document.
It somehow hasn't really been much of an issue until now, but we still
do way too much work in the watchdog main loop. I'll look into resolving
that more general issue in the future.
This now occurs for some large completions downstream of `import
Mathlib`. I'd like to get rid of this `whnf` call entirely in the
future, but this is a decent quick mitigation.
* Now `getPathToBaseStructure?` can navigate to all parent structures,
not just through subobjects.
* Adds a "resolution order" for methods. This is the order that
generalized field notation visits parent structures when trying to
resolve names. The algorithm to compute a resolution order is the
commonly used C3 (used for instance by Python). By default we use a
relaxed version of the algorithm that tolerates inconsistencies. Using
`set_option structure.strictResolutionOrder true` makes inconsistent
parent orderings into warnings.
* This makes generalized field notation be able to resolve names for all
parent structures, not just those that are embedded as subobjects.
Closes#3467. (And addresses side note in #1881.)
* Modifies `getAllParentStructures` to return *all* parents. This
improves dot completion in the editor.
It's difficult to distinguish theorems from regular definitions in the
completion menu, which is annoying when using completion for searching
one or the other. This PR makes theorem completions use the "Eureka!"
icon ()
to distinguish them more clearly from other completions.
NB: We are very limited in terms of which icons we can pick here since
[the completion kinds provided by LSP / VS
Code](https://code.visualstudio.com/docs/editor/intellisense#_types-of-completions)
are optimized for object-oriented programming languages, but I think
this choice strikes a nice balance between being easy to identify,
having some visual connection to theorem proving and not being used a
lot in other languages and thus not clashing with pre-existing
associations.
This PR simplifies the signature of `Array.mapIdx`, to take a function
`f : Nat \to \a \to \b` rather than a function `f : Fin as.size \to \a
\to \b`.
Lean doesn't actually use the extra generality anywhere (so in fact this
change *simplifies* all the call sites of `Array.mapIdx`, since we no
longer need to throw away the proof).
This change would make the function signature equivalent to
`List.mapIdx`, hence making it easier to write verification lemmas.
We keep the original behaviour as `Array.mapFinIdx`.
Should ensure we visit at most as many expr nodes as in the final expr
instead of many possibly overlapping mvar assignments. This is likely
the only way we can ensure acceptable performance in all cases.
---------
Co-authored-by: Kim Morrison <kim@tqft.net>
This PR ensures that deprecated declarations are displayed with a
strikethrough markup in the completion popup of VS Code and that the
docstring of a completion item denotes the meta-data of the deprecation.
This PR enables tactic completion in the whitespace of a tactic proof
and adds tactic docstrings to the completion menu.
Future work:
- A couple of broken tactic completions: This is due to tactic
completion now using @david-christiansen's `Tactic.Doc.allTacticDocs` to
obtain the tactic docstrings and should be fixed soon.
- Whitespace tactic completion in tactic combinators: This requires
changing the syntax of tactic combinators to produce a syntax node that
makes it clear that a tactic is expected at the given position.
Closes#1651.
This PR fixes three problems:
- When the language server is being stopped in a non-normal way without
going through the regular LSP shutdown protocol (e.g. by closing VS
Code), it could sometimes happen that both the watchdog and the file
worker were not properly terminated and lingered around forever,
resulting in zombie processes (#5296)
- When the file worker crashes and the user restarts it by making a
change to the document, the file worker would produce incorrect
diagnostics for the document until the file is restarted.
- (Minor) When the file worker would crash during initialization, the
error diagnostic would be reported on stderr instead of stdout
The deadlock-induced termination issue from #5296 should be resolved by
the following measures:
- The watchdog main task is always terminated with `IO.Process.exit` to
ensure that it terminates even if some other tasks in the process are
still running.
- The file worker communication task in the watchdog no longer waits for
the file worker process to terminate when writing to the client fails,
only when reading from the file worker fails.
- When the watchdog shuts down (either as a result of an orderly or a
non-normal shutdown), instead of waiting for the file worker
communication tasks to complete, it kills the file worker process. The
rationale behind this is that the file worker currently should have no
essential work to complete if the server is being stopped anyways, and
so waiting for the communication task is not necessary.
The file worker diagnostic desync after a crash was caused by us
tracking changes to the document of a crashed file worker twice: Once as
part of the document, and once as part of the queued messages to the
file worker. This meant that when the file worker was restarted, it
would receive the changes made to the document while the file worker was
crashed twice, leading to a desynced document state.
(Probably) fixes#5296.
