Repository navigation
Conversation
timfennis
force-pushed
the
feat/new-analyser
branch
from
October 13, 2026 13:51
ae3da97 to
24c9a57
Compare
…ment targets 🧩 Key declarations and references by NodeId now that patterns and targets carry their own node identities, replacing the span-keyed lvalue list. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Function declarations join overload sets while plain bindings shadow, so each symbol now records which of the two it is. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An identifier now records every same-named symbol it can see, nearest scope first, so the type pass can choose between overloads and plain bindings. Assignment targets still resolve to the innermost symbol. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Natives now live in a root scope above the program's top level, so a top-level binding shadows a native instead of sharing its scope. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Named functions bind before their body so recursion resolves, and parameters live in a scope that ends with the function. Duplicate parameters keep the current analyser's error message. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…scope 🚧 A scope holds a name either as plain bindings or as an overload set, so a function declared after a same-named variable in one scope is an error. Across scopes, ordinary lexical shadowing applies. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Scopes now know whether they hold builtins, the program's top level, a function's parameters, or a block, so later lookups can tell when they cross a function boundary. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Each function records the outer symbols it captures, including ones it only passes through to an inner function, and the resolver collects every captured symbol so the compiler can keep them alive past their scope (#227). Every overload candidate is captured until the type pass narrows them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`x op= y` records candidates for both the in-place `op=` and the plain `op` under the assignment's node, leaving the choice to the type pass. Sequence targets keep the current analyser's error message. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ccess 🔑 `p.x` records candidates for the getter `x`, member and index targets record their setter (`x=` or `[]=`), and augmented assignment to them also records the getter it reads through. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Type names live in their own namespace beside values, so `let Int = 5` doesn't hide the type `Int`. Every name in a `let`, parameter, return or cast annotation resolves through lexical scopes to a type symbol, or reports the current analyser's unknown type message. Built-in type names are declared in the builtins scope. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A struct binds its name as a type, a constructor, and a getter and setter per field. Field types resolve before the name is bound, so self and forward references stay unknown types. Builtin shadowing, same-scope redefinitions and duplicate fields keep the current analyser's messages. The constructor follows the same-scope rule for functions, while field accessors are exempt so field names can match local variables. With structs done, the resolver has no unsupported expressions left. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
`None` is a literal since #229, so the resolver treats it like `true`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
timfennis
force-pushed
the
feat/new-analyser
branch
from
October 13, 2026 14:17
24c9a57 to
5f7036b
Compare
Stacked on #226 so these changes can be reviewed apart from the resolver work itself. The resolver on `feat/new-analyser` produced its results but nothing used them. This PR runs it in front of the existing analyser, so it handles every program and test, as phase 1 of replacing the old analyser. Phase 2 will turn the old analyser into the type pass. - **Keep going after errors:** the resolver records one error per failing statement and carries on, including inside blocks, so `a; b;` reports both names like the current analyser does. - **Resolve names before analysis:** `Analyser::analyse` runs the resolver on each top-level statement first, with the registered natives as builtins. A statement with name errors isn't analysed further, so the same error isn't reported twice. The resolver's state is part of the analyser checkpoint, so a failed REPL line rolls it back. The compiler is unchanged. **Worth a look:** - **Behaviour change:** a `fn` after a `let` of the same name in one scope is now an error, as decided for the resolver. `005_functions/034` became an error test, and the function page in the manual documents the rule. - **Test fixtures:** four analyser unit tests now declare the `[]`/`[]=` natives their index assignments use, because the resolver rejects index access without them. - **Unused fields:** the `AugmentedCandidates` and `StructSymbols` fields only the type pass will read are marked `expect(dead_code)`, so the marker becomes an error once they're used. 🤖 Generated with [Claude Code](https://claude.com/claude-code) --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Phase 2, step 1, stacked on #226. Later steps will stack on this branch. The layout pass in phase 2 will turn native symbols into the VM's global slots, so it needs each native's position in the function registry. That position happened to equal the native's `SymbolId`, because natives are declared first, but nothing guaranteed it. This makes the mapping explicit. - `NameResolver` records `native_slots`, from each native symbol to its registry index. Overloads of the same name keep separate slots. - Nothing reads it yet, so the field is marked `expect(dead_code)` outside tests, like the other fields waiting for later passes. No behaviour change. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
This branch has not been deployed
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.
No description provided.