Request Timeouts started plaguing IDE due to numerous `executionContext/***Visualization` requests. While caused by a bug they revealed a bigger problem in the Language Server when serving large amounts of requests:
1) Long and short lived jobs are fighting for various locks. Lock contention leads to some jobs waiting for a longer than desired leading to unexpected request timeouts. Increasing timeout value is just delaying the problem.
2) Requests coming from IDE are served almost instantly and handled by various commands. Commands can issue further jobs that serve request. We apparently have and always had a single-thread thread pool for serving such jobs, leading to immediate thread starvation.
Both reasons increase the chances of Request Timeouts when dealing with a large number of requests. For 2) I noticed that while we used to set the `enso-runtime-server.jobParallelism` option descriptor key to some machine-dependent value (most likely > 1), the value set would **only** be available for instrumentation. `JobExecutionEngine` where it is actually used would always get the default, i.e. a single-threaded ThreadPool. This means that this option descriptor was simply misused since its introduction. Moved that option to runtime options so that it can be set and retrieved during normal operation.
Adding parallelism intensified problem 1), because now we could execute multiple jobs and they would compete for resources. It also revealed a scenario for a yet another deadlock scenario, due to invalid order of lock acquisition. See `ExecuteJob` vs `UpsertVisualisationJob` order for details.
Still, a number of requests would continue to randomly timeout due to lock contention. It became apparent that
`Attach/Modify/Detach-VisualisationCmd` should not wait until a triggered `UpsertVisualisationJob` sends a response to the client; long and short lived jobs will always compete for resources and we cannot guarantee that they will not timeout that way. That is why the response is sent immediately from the command handler and not from the job executed after it.
This brings another problematic scenario:
1. `AttachVisualisationCmd` is executed, response sent to the client, `UpsertVisualisationJob` scheduled.
2. In the meantime `ModifyVisualisationCmd` comes and fails; command cannot find the visualization that will only be added by `UpsertVisualisationJob`, which might have not yet been scheduled to run.
Remedied that by checking visualisation-related jobs that are still in progress. It also allowed for cancelling jobs which results wouldn't be used anyway (`ModifyVisualisationCmd` sends its own `UpsertVisualisationJob`). This is not a theoretical scenario, it happened frequently on IDE startup.
This change does not fully solve the rather problematic setup of numerous locks, which are requested by short and long lived jobs. A better design should still be investigated. But it significantly reduces the chances of Request Timeouts which IDE had to deal with.
With this change I haven't been able to experience Request Timeouts for relatively modest projects anymore.
I added the possibility of logging wait times for locks to better investigate further problems.
Closes#7005
- Add type detection for `Mixed` columns when calling column functions.
- Excel uses column name for missing headers.
- Add aliases for parse functions on text.
- Adjust `Date`, `Time_Of_Day` and `Date_Time` parse functions to not take `Nothing` anymore and provide dropdowns.
- Removed built-in parses.
- All support Locale.
- Add support for missing day or year for parsing a Date.
- All will trim values automatically.
- Added ability to list AWS profiles.
- Added ability to list S3 buckets.
- Workaround for Table.aggregate so default item added works.
Fixes performance problems observed when creating/resolving errors (#6674):
|before|after|
|---|---|
|||
This also helps with #6637, although I haven't been able to reproduce the degree of slowness shown there so I can't confirm that this resolves that issue.
# Important Notes
- Disable visualizations until shown. [Faster startup, and all graph changes.]
- 6x faster message deserialization. [Saves 400ms when making a change with many visualizations open.]
- Fast edge recoloring. [Saves 100-150ms when disconnecting an edge in Orders.]
- Add a checked implementation of a `profiler` data structure, used instead of the fast `unsafe` version when `debug-assertions` are enabled.
Several small changes:
- Dropdowns: Populate `GridView` lazily (fixes#6865).
- Clear disconnected edges when editing node (fixes case 1 in #7018).
- Fix regression in node selection rendering (2nd bug in #6975).
- Update profiler docs. The hotkey to *prOfile without exiting* is now `Ctrl+Alt+O` (`Ctrl+Alt+P` has been requisitioned by the CB).
Node selection:
| Pre-`Rectangle` | This PR |
| --- | --- |
|  |  |
# Important Notes
- `Rectangle`: When `inset > border`, the extra space is now between the body and the border, not outside the border.
- More robust node layering logic. Now an inconsistent layer order cannot occur, even if something strange happens (like editing an expression and an edge at the same time).
- The dynamic drop down in the `drop_down` example scene doesn't show any entries before (or after) this, so I can't test the dynamic case.
This adds the spec for all update actions, but implements the common input validation framework and `Insert`. Tests for remaining actions are marked as pending - these will be implemented in a subsequent PR.
close#6936
Changelog:
- add: new suggestion type Getter that is not exposed to the api
- update: do not return suggestion of type getter when doing a global search (without specifying self types)
Private suggestions and modules mentioned in the issue will be filtered out after we finish the work on the new (refined) exports algorithm.
# Important Notes
This PR adds facilities for controllers to be aware of what shortcut command is currently being processed. This allows grouping consequences of single user action into a single transaction without hard-coding it separately for all the separate paths case-by-case, which turned out to be challenging and error-prone.
Additionally, a number of minor fixes were carried over from #6877:
* workaround for #6718;
* avoiding creating spurious transactions when dealing with node positions;
* dropping any non-user user-triggered transactions that occur during the IDE project initialization.
Before, we prevented default on every keyboard event, making usage of HTML inputs impossible (both on the dashboard and inside visualizations).
# Important Notes
The list of shortcuts was based on comments in #6364 (where the `preventDefault` was introduced).
As demonstrated in https://github.com/enso-org/enso/actions/runs/5175688022/jobs/9323585204?pr=6940
```
org.enso.interpreter.test.instrument.RuntimeAsyncCommandsTest *** ABORTED ***
org.graalvm.polyglot.PolyglotException: java.util.NoSuchElementException
at java.base/java.util.WeakHashMap$HashIterator.nextEntry(WeakHashMap.java:811)
at java.base/java.util.WeakHashMap$EntryIterator.next(WeakHashMap.java:848)
at java.base/java.util.WeakHashMap$EntryIterator.next(WeakHashMap.java:846)
at org.enso.interpreter.runtime.ThreadExecutors.shutdown(ThreadExecutors.java:46)
at org.enso.interpreter.runtime.EnsoContext.shutdown(EnsoContext.java:198)
at org.enso.interpreter.EnsoLanguage.finalizeContext(EnsoLanguage.java:179)
at org.enso.interpreter.EnsoLanguage.finalizeContext(EnsoLanguage.java:65)
at org.graalvm.truffle/com.oracle.truffle.api.LanguageAccessor$LanguageImpl.finalizeContext(LanguageAccessor.java:326)
at org.graalvm.truffle/com.oracle.truffle.polyglot.PolyglotLanguageContext.finalizeContext(PolyglotLanguageContext.java:404)
at org.graalvm.truffle/com.oracle.truffle.polyglot.PolyglotContextImpl.finalizeContext(PolyglotContextImpl.java:2925)
```
There was an inherent race condition between edit, close & open commands which could not be prevented solely using locks. `EditFileCmd` triggered `EnsureCompiledJob` which was applying edits collected over time. At the same `CloseFileCmd` and `OpenFileCmd` were executed asynchronously and required locks on compilation unit and file lock.
Additionally, open file was resetting the module's runtime source irrespective of any edits that could already have been applied with the asynchronous execution in `EnsureCompiledJob`. This was visible especially during early manipulation of the project when open/close was performed due to a bug in IDE (#6843).
Now commands can be run either synchronously or asynchronously. Only that way can we ensure that `close` & `open` commands finish by the time any editions are being applied to module's sources.
Closes#6841.
# Important Notes
In the given video, `"foo"` would be greyed out because it would never be part of the module's (runtime) sources. Therefore no IR would be generated for it or instrumentation, meaning it would be present in `expressionUpdates` information necessary for IDE.
[Kazam_screencast_00014.webm](https://github.com/enso-org/enso/assets/292128/226a17b8-729a-415a-803f-003a9695b2f1)
close#6900
This is a follow-up to the discussion of the imports/exports meeting.
Right now we have no control over the visibility of atom constructor arguments. One way to hide them is a convention of filtering getters by an `internal` prefix or suffix.
Related to #6912
It essentially solves it by removing any builtins that would take an EnsoDate/EnsoTimeOfDay/EnsoTimeZone and replacing them with Java utils that do the same operation.
This is not a proper solution - the builtin conversion is still invalid for the date/time types - but at this moment we may just no longer use the invalid conversion so it is much less of an issue. We still need to be aware of this if we want to introduce builtins taking date/time in the future.
Implements #6544 (eliminates 10/42 of the constantly-displayed draw calls).
Fixes#6717. Improves startup CPU time by 5% (250ms, loading Orders on my dev box).
# Important Notes
- Edges: New implementation uses only Rectangle under most conditions.
- Node and action area: Replace some shapes with Rectangle.
- List view: Replace some shapes with Rectangle.
- Display object hierarchy: The lowest-level shape instance types no longer have their own display objects.
- Includes initial support for using `Rectangle` to display triangles.
`executionContext/create` method has an optional `context_id` parameter. Supplying this argument makes the user's session more reproducible. I.e. this way the language server can recreate the user's session by recording the requests.
* Save backend type; fix `-startup.project`
* Attempt to fix for cloud
* Fix for cloud and implement automatic opening
* Fixes
* Add missing functionality
* Switch default backend to local backend
* Fix type error
* Fix saving backend
* Make loading message appear instantly
* Fix context menu positioning
* Fixes and QoL improvements
* Style scrollbar on `document.body`
* Fix `-startup.project`; other minor fixes
* Open project immediately when creating from template; minor fix
* Finally fix spinner bugs
* Fix some minor bugs
* Fix bugs when closing project
* Disallow deleting local projects while they are still running
* Close modals when buttons are clicked
Fixes#6763
The bug was caused by pushing stack frames in the engine first, then failing to get graph controller - in that case we didn't update graph, but kept the stack, so graph was not synchronized with stack.
This PR changes the approach: we try to open graph first, only then push frames to stack. If _any_ frame will fail, we try to pop those we pushed so far, to restore the previous state. The same fix was applied for leaving nodes.
Also, I realized, that running several "enter node" / "leave node" actions could mix push/pop operations, making a mess with our execution context state. I've added a mutex to ensure we won't do this: the contesting operation will simply fail.
# Important Notes
In case when _restoring_ state fails, I did not have any better idea than just trying until it succeed.
First part for #6498 - refactoring of the upload infrastructure, in preparation for `update_database_table`.
Implemented a `Set` data structure which was long needed.
The APIs are added and an initial implementation is created, but it is not complete - but it has grown significantly already so the remaining implementation will be done as a separate PR.
Adds some basic ability for a function to ensure that it is only executed from within a transaction.
At the beginning of the execution `EnsureCompiledJob` acquired write compilation lock. When compiling individual modules it would then
- acquire file lock
- acquire read compilation lock
The second one was spurious since it already kept the write lock. This sequence meant however that `CloseFileCmd` or `OpenFileCmd` can lead to a deadlock when requests come in close succession. This is because commands:
- acquire file lock
- acquire read compilation lock
So `EnsureCompiledJob` might have the (write) compilation lock but the commands could have file lock. And the second required lock for either the job or the command could never be acquired.
Flipping the order did the trick.
Partially solves #6841.
# Important Notes
For some reason we don't get updates for the newly added node, as illustrated in the screenshot, but that could be related to the close/open action. Will need to dig more.
- Add the missing dropdowns for `Locale` and `Encoding`.
- Correct a few mismatched type signatures.
- Adjust `order_by` calls with a single `Sort_Column` to call in a Vector.
- Adjust parameter names for `transpose`.
- Fix for the table viz: escape HTML and `suppressFieldDotNotation`.
- Use `Filter_Condition.Equal True` for the default filter.
- Adjust `Data.fetch` to return the response on success when parse fails. Rename `parse` to `try_auto_parse`.
- Add various aliases for methods.
- Add tests for `Table.set` when using a `Vector`, `Range` or `Date_Range`.
- Add check for mismatched length on `Table.set`.
Previously, a `RuntimeException` would be thrown when an attempt would be made to curry a conversion function. That is problematic for IDE where `executionFailed` means we can't enter functions due to lack of method pointers info.
Closes#6897.
# Important Notes
A more generic solution that allows to recover from execution failures will need a follow up.
Closes#5227
# Important Notes
- This lays first steps towards #6292 - we get pure Enso variants of MultiValueKey.
- Another part refactors `LongStorage` into `AbstractLongStorage` allowing it to provide alternative implementations of the underlying storage, in our case `LongRangeStorage` generating the values ad-hoc and `LongConstantStorage` - currently unused but in the future it can be adapted to support constant columns (once we implement similar facilities for other types).
