Upgrade Mago to 1.44.0

Author: AJenbo

Cargo.lock

@@ -18,16 +18,16 @@ sqlparser = "0.62"
 tracing = "0.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 tokio = { version = "1.39", features = ["full"] }
-mago-syntax = "=1.43.0"
-mago-allocator = "=1.43.0"
-mago-database = "=1.43.0"
+mago-syntax = "=1.44.0"
+mago-allocator = "=1.44.0"
+mago-database = "=1.44.0"
 mago-docblock = "=1.42.0"
-mago-names = "=1.43.0"
-mago-span = "=1.43.0"
+mago-names = "=1.44.0"
+mago-span = "=1.44.0"
 mago-type-syntax = "=1.42.0"
-mago-formatter = { version = "=1.43.0", features = ["serde"] }
-mago-php-version = "=1.43.0"
-mago-composer = "=1.43.0"
+mago-formatter = { version = "=1.44.0", features = ["serde"] }
+mago-php-version = "=1.44.0"
+mago-composer = "=1.44.0"
 ignore = "0.4"
 globset = "0.4"
 regex = "1"

Cargo.toml

@@ -58,6 +58,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- **Formatting a short method chain starting with `new X(...)` no longer breaks it across lines unnecessarily.** When the constructor call's own arguments were long enough to wrap, the formatter also forced a short trailing chain like `(new Foo(...))->bar()` onto separate lines even though it would have fit on one. Upstream fix from mago 1.44.0.
 - **`@method` and `@property` tags on an implemented interface are now always applied.** A class that declared no docblock of its own missed the magic methods and properties its interfaces declared, so they did not complete, hover, or resolve, and calls to them were reported as unknown members. Tags on an interface (and on the interfaces it extends) are now picked up regardless of what the implementing class documents.
 - **Find References, Rename, and Go to Implementation no longer look stalled during startup indexing.** A search started while the background index is still parsing the workspace waits for that index to finish, since acting on a partial index would silently miss results. That wait now shows in the request's own progress bar as "Waiting for workspace index" alongside the index's live file counts, instead of sitting at "Resolving…" with no indication of what it is waiting for.
 - **Type narrowing against `@phpstan-assert`/`@psalm-assert` no longer leaks memory.** Evaluating a narrowing call such as `Assert::isInstanceOf($x, Foo::class)` or a custom function/method with the same annotations allocated a small amount of memory that was never freed. This ran on every conditional touched during completion, hover, diagnostics, and go-to-definition, so memory held by a long-running editor session grew slowly but permanently the more the project was edited. Fixed by no longer leaking the allocation.

docs/CHANGELOG.md

@@ -195,7 +195,7 @@ unlikely to move the needle for most users.
 | E6  | Stub install prompt for non-Composer projects                                                                                                                               | Low         | Low         |
 | E7  | [Stub-based framework patches](todo/external-stubs.md#e7-stub-based-framework-patches)                                                                                      | Medium      | Medium      |
 |     | **[Performance](todo/performance.md)**                                                                                       |             |             |
-| P29 | [Migrate to `mago-phpdoc-syntax`](todo/performance.md#p29-migrate-to-mago-phpdoc-syntax) (drop deprecated `mago-docblock` / `mago-type-syntax`)                             | Medium      | Medium      |
+| P29 | [Migrate to `mago-phpdoc-syntax`](todo/performance.md#p29-migrate-to-mago-phpdoc-syntax) (drop deprecated `mago-docblock` / `mago-type-syntax`; sequenced behind P30)         | Medium      | High        |
 | P30 | [Evaluate migrating parse/resolve/docblock pipeline to `mago-hir`](todo/performance.md#p30-evaluate-migrating-parseresolvedocblock-pipeline-to-mago-hir) (blocked on upstream API stabilizing — see triggers) | Medium-High | High        |
 | P43 | [`init_single_project` is the longest single-threaded stretch of a run](todo/performance.md#p43-init_single_project-is-the-longest-single-threaded-stretch-of-a-run)        | Medium-High | Medium      |
 | P16 | [Pre-parsed stub format (eliminate raw PHP embedding)](todo/performance.md#p16-pre-parsed-stub-format-eliminate-raw-php-embedding)                                          | High        | Medium-High |

docs/todo.md

@@ -680,33 +680,63 @@ body-return inference, since fixed by memoization).
 
 ## P29. Migrate to `mago-phpdoc-syntax`
 
-**Impact: Medium · Effort: Medium**
+**Impact: Medium · Effort: High**
 
 As of the 1.42/1.43 mago releases, `mago-docblock` and
 `mago-type-syntax` are both deprecated and frozen at 1.42.0. Upstream
 folded them into a single unified crate, `mago-phpdoc-syntax`, which
 parses docblock comments and their embedded type and constant
-expressions in one pass and is meant to replace both. The deprecation
-notices point at `mago_phpdoc_syntax::PHPDocParser` (for
-`parse_phpdoc_with_span`) and `mago_phpdoc_syntax::parse_type` (for
-`parse_str`).
-
-We currently call the deprecated functions at four sites, each wrapped
-in `#[allow(deprecated)]`:
-
-- `src/docblock/parser.rs` — `mago_docblock::parse_phpdoc_with_span`
-- `src/php_type.rs` — `mago_type_syntax::parse_str` (type-string parse
-  and the round-trip test helper)
-- `src/symbol_map/docblock.rs` — `mago_type_syntax::parse_str`
-
-`mago-phpdoc-syntax` first shipped at 1.44.0, so this migration is
-coupled to bumping the rest of the mago toolchain from 1.43.0 to
-1.44.0 (mixing 1.43 and 1.44 would pull two incompatible copies of the
-shared `mago-span` types). Do the version bump and the API migration
-together, drop `mago-docblock` and `mago-type-syntax` from
-`Cargo.toml`, and remove the `#[allow(deprecated)]` wrappers. The
-unified single-pass parser should also be a small win on the docblock
-parse hot path.
+expressions in one pass and is meant to replace both.
+
+This is **not** a same-shaped API rename. The deprecation notices
+point at `mago_phpdoc_syntax::PHPDocParser` (replacing
+`parse_phpdoc_with_span`) and `mago_phpdoc_syntax::parse_type`
+(replacing `parse_str`), but the new crate produces a fundamentally
+different CST: the old flat `mago_docblock::document::{Document, Tag,
+TagKind}` (one enum with `tags_by_kind(TagKind::X)` lookups) is
+replaced by a `TagValue` enum with ~60 per-tag-kind structs
+(`ThrowsTagValue`, `TemplateTagValue`, `ParamTagValue`, ...), and
+`mago_type_syntax::cst::Type` (69 variants) is replaced by an equally
+restructured type CST. Confirmed by reading both crate sources
+directly rather than relying on docs.rs.
+
+Grepping for `mago_docblock`/`mago_type_syntax` usage turns up far
+more than the three call sites in the old writeup — at last count 14
+files, including the core of the shared type engine
+(`src/php_type/mod.rs`, `src/php_type/parse.rs` build `PhpType` off
+`mago_type_syntax::cst::Type` directly) and every consumer that
+filters tags by `TagKind` (`src/docblock/tags.rs`, `templates.rs`,
+`conditional.rs`, `virtual_members.rs`, `completion/phpdoc/mod.rs`,
+`code_actions/update_docblock.rs`, `code_actions/phpstan/add_throws.rs`,
+`completion/source/throws_analysis/scanning.rs`, `hover/formatting.rs`).
+This is a full rewrite of the docblock/type parsing foundation, sized
+for several sprint items, not one.
+
+**Sequencing with P30:** [P30](#p30-evaluate-migrating-parseresolvedocblock-pipeline-to-mago-hir)
+evaluates `mago-hir`, which — if it pans out — replaces this entire
+hand-rolled layer (docblock tag parsing and type-string parsing both)
+with a single upstream pass. Hand-migrating to `mago-phpdoc-syntax`
+now risks being thrown away if `mago-hir` is adopted later. Do not
+start the CST rewrite until P30's triggers have been evaluated one way
+or the other — either `mago-hir` is adopted (making this migration
+moot) or its triggers are judged unmet for the foreseeable future
+(making this migration the fallback worth doing).
+
+The rest of the mago toolchain (`mago-syntax`, `mago-allocator`,
+`mago-database`, `mago-names`, `mago-span`, `mago-formatter`,
+`mago-php-version`, `mago-composer`) is already on 1.44.0 — that bump
+was independent of this CST rewrite and safe to do ahead of it (see
+the `[Unreleased]` changelog entry for the one user-visible effect, a
+`mago-formatter` chain-breaking fix). `mago-docblock` and
+`mago-type-syntax` are still pinned at 1.42.0 pending this migration.
+Do not go past 1.44.0 for the rest of the toolchain until this
+migration (or P30) is underway: `mago-allocator` 1.45.0 rewrites
+`LocalArena`/`SharedArena` into a unified `Arena` trait (new
+`boxed.rs`, `iter.rs`, `collections.rs`, `vec.rs` modules) — a real
+breaking change to the arena API this codebase uses throughout the
+parse hot path (`src/parser/ast_update.rs`'s `with_reusable_arena`,
+the parse-worker threads, etc.) — and taking it piecemeal ahead of a
+real reason to touch that code would be its own unreviewed migration.
 
 ---