Conversation
Previously, arguments used in keyword function calls (generated by Julia's lowering for functions with keyword arguments) were incorrectly counted as "used", preventing proper unused argument diagnostics. For example, in `func(a; kw) = a`, the `kw` argument would not be reported as unused. This change extends the argument usage skip logic to handle: - Keyword function calls (`#funcname#N` pattern) - `kwerr` calls from `Core.TopNode` Also fixes a minor bug where `start_idx` was not used in the loop. Fixes #390
Previously, dot-prefixed completions like `xarg.x` would show irrelevant global completions when the prefix type could not be resolved. Now, completions are disabled entirely in such cases, returning no results. This provides a cleaner user experience by not showing unrelated suggestions. Property completions for non-module values may be added in the future. Fixes #389
Add completion support for Julia keywords. Keywords are sourced from `REPL.REPLCompletions.sorted_keywords` and `sorted_keyvals`. Keywords are given the lowest sort priority (`max_sort_text3`) so they should appear after local variables and globals. They are excluded from macro invocation and latex/emoji completion contexts. Closes #386.
Allow `module_name` to be omitted in `module_overrides` configuration. When `module_name` is not specified, analysis for matching files is disabled and `OutOfScope()` is returned without a module context, effectively disabling full-analysis for the files.
Rename the struct and field for clarity, as it controls analysis behavior more broadly than just module context.
Initialization options can now be configured via the `[initialization_options]` section in `.JETLSConfig.toml`, in addition to editor-specific LSP settings. This provides a client-agnostic way to configure initialization options that can be committed to version control. When both sources are present, file-based configuration takes precedence (file > client > default), as like the dynamic configuration. But I may revise this priority rule in the very near future.
Use `show_[warning|error]_message` to notify users about configuration errors in `.JETLSConfig.toml` initialization options, such as TOML parse errors or invalid option values.
Make InitOptions a ConfigSection subtype, enabling initialization options to be set via .JETLSConfig.toml and tracked by the configuration change notification system. Also improve changed_settings_message to clearly distinguish between immediately applied changes and those pending server restart.
This commit implements a prototype of Revise-based incremental full-analysis. The idea is to delegate package state management to Revise, similar to JET v0.11, enabling fast updates of analysis results when incremental changes are made while avoiding issues like #357. Although this implementation is incomplete, it works at least for JETLS itself and its dependencies, making it possible to experiment with various feature extensions that Revise-based incremental analysis enables. For example, this approach makes it practical to cache type inference results in `FileInfo` and update them on `DidChange` (though memory leak risks return). This would make dynamic diagnostics, type on hover, and inlay hints more feasible. The following incomplete aspects need to be addressed for complete migration to this new analysis approach: - This mode is only enabled when `JETLS_DEV_MODE` is active. This is because JETLS currently loads Revise only when `JETLS_DEV_MODE` is enabled, and this feature relies on that Revise instance. However, there is no guarantee that existing analysis and Revise won't interfere with each other. I will enable this mode outside of `JETLS_DEV_MODE` only after I better understand Revise internals and can achieve finer-grained control over Revise's automatic file watching, allowing us to safely load Revise (or ReviseCore?) in a secure manner. - Code loading is not yet implemented. This could be easily implemented using `Base.require`, but currently we use the approach of looking up `module_name` specified in `analysis_overrides` of initialization options from `Base.loaded_modules`. This means external packages cannot be analyzed with this new method - only packages already loaded as JETLS dependencies at server initialization. This is also related to the fact that this mode currently only works for package code. We need to reconsider how to load and analyze script code as well. - The Revise-based tracking system needs adjustment for special packages like `Base` and `Compiler`. - Generally, my understanding of Revise is still insufficient and the package monitoring approach may be immature. Also checks in the changes for .JETLSConfig.toml to enable this new analysis mode for JETLS itself.
In particular `Sys.DATAROOTDIR` is no longer defined, so we need to use `Sys.BINDIR/../..` instead.
Use hyphens as the default word separator in file names. Use underscores only when the file name corresponds directly to Julia code (e.g., module names), since Julia identifiers cannot contain hyphens. This commit renames files to follow this convention and updates all corresponding include statements. The naming rule is now documented in AGENTS.md.
Refactor `InterpretationStateCache` to use an explicit `target_modules` field instead of a lazily-computed cache, simplifying the code. Also allows the interactive `[@]analyze_call` utilities to set `target_modules` for testing purposes.
Rename `select_target_node` to `select_target_identifier` and `select_dotprefix_node` to `select_dotprefix_identifier` for better clarity about their purpose.
Previously, when a file was closed or when test structure changes removed all diagnostics for a URI, the client would not receive a notification to clear the stale diagnostics because the URI was simply absent from the diagnostics map. This adds an `ensure_cleared` parameter to `notify_diagnostics!` that guarantees an empty diagnostics notification is sent for the specified URI, ensuring the client clears any previously displayed diagnostics.
Add the ability to rename files referenced in string literals. When the
cursor is on a string that refers to an existing file path (e.g., in
`include("foo.jl")`), the rename operation now renames both the file on
disk and updates the string reference in the source code.
Add explicit `local` to the inner `message` variable to prevent it from overwriting the outer `message` used in the returned Diagnostic. Without this, the diagnostic message would be replaced by the last frame's message from the loop, causing "message must be set" errors. Probably fixes #393.
Removed `byte_ancestors` function overloads that operated on `JS.SyntaxNode` since they were no longer used anywhere.
Add static analysis for field access errors by hooking into `CC.builtin_tfunction` to intercept `getfield`, `setfield!`, `fieldtype`, and `getglobal` calls. Two new report types are introduced: - `FieldErrorReport` (`inference/field-error`): reported when accessing a non-existent field by name - `BoundsErrorReport` (`inference/bounds-error`): reported when accessing a field by an out-of-bounds integer index Note that the `inference/bounds-error` diagnostic is reported when code attempts to access a struct field using an integer index that is out of bounds, such as `getfield(x, i)` or tuple indexing `tpl[i]`, and not reported for arrays, since the compiler doesn't track array shape information. Reports from invalid `setfield!` and `fieldtype`, and general invalid argument types are left as future TODO. Also adjusts concrete evaluation logic to enable ad-hoc constant propagation after failed concrete evaluation for better accuracy.
Detected by the new field analysis.
This ensures duplicate reports are properly deduplicated according to JET's aggregation policy before converting to LSP diagnostics.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This PR releases version
2025-12-18.Checklist
release / Test JETLS.jl with release environmentrelease / Test jetls executable with release environmentPost-merge
releases/2025-12-18branch can be deleted after merging