Add Decode #8869
Merged
Add Decode #8869
Conversation
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
Port the Decode module from the old crates/ compiler to the new src/ compiler, using static dispatch via where clauses instead of Abilities. This includes: - Decode module with DecodeErr, Decoder opaque type, and helper functions (custom, ok, err, from_bytes_partial) - LineFmt module providing a trivial newline-separated encoding format for testing, with string_decoder implementation Also fixes several interpreter bugs discovered during implementation: - Add safeCopy helper to handle overlapping memory in @memcpy - Add bounds checking for captures_layout_idx in closure operations - Fix payload extraction in pattern matching to prefer variant_layout when computed layout is ZST but variant layout is concrete 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When constructing a tag union from a polymorphic type (like Try(val, DecodeErr) where val is a rigid/uninstantiated type variable), the expected payload layout was ZST but the actual payload was concrete (e.g., a string). This caused pattern matching to fail because extractTagValue would see ZST variant layout. Extended the existing layout mismatch fix to also detect when the expected variant layout is ZST but the actual payload is not ZST, and create a proper tag_union layout with the concrete payload size. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
LineFmt was only needed for testing the Decode module. Moved it from Builtin.roc to inline definitions in the test files. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Reverted changes to pattern matching layout resolution and tag union construction that caused segfaults when handling recursive types. Kept defensive bounds checking for builtin closure indices. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Reverted the bounds checking patterns that silently skipped operations when closure capture indices were out of bounds. If such a condition occurs, it indicates a bug that should be caught and fixed, not hidden. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove LineFmt from builtin_types in Can.zig - Remove LineFmt from modules array in BuiltinModules.zig - Replace custom safeCopy function with @memmove builtin 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Changed Decoder from an opaque type with methods to a simple type alias:
Decoder(src, val, err) :: { decode : (src -> { result : Try(val, err), rest : src }) }
Key changes:
- Removed DecodeErr opaque type - error is now parameterized
- Removed is_eq method from Decoder
- Format is captured via closure, not stored in Decoder
- Renamed custom -> decoder, from_bytes_partial -> decode
- Added decode_exact for consuming all input
- Updated tests to use new API
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove decode_type and decode_ident from BuiltinIndices - Change builtin types from "Decode" to "Decoder" in Can.zig - Update tests to use Decoder.decoder/Decoder.decode - Replace Decode.ok/err helpers with inline record syntax - Update where_clause snapshots for changed error messages 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Decoder uses :: (alias syntax) instead of := (opaque syntax), which creates s_alias_decl statements rather than s_nominal_decl. Add handling for alias declarations in: - findTypeDeclaration: Find both nominal and alias type declarations - createImportMapping: Map both nominal and alias types for display 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add decode methods to I8, I16, I32, I64 in Builtin.roc
- Uses where clause pattern: decode : src, fmt -> (Try(T, err), src)
where [fmt.decode_* : fmt, src -> (Try(T, err), src)]
- Fix qualified ident lookup for auto-imported type methods
(e.g., I32.decode now correctly resolves to Builtin.Num.I32.decode)
- Add interpreter support for where clause method dispatch
(partial - runtime type still loses nominal wrapper)
- Add decode tests for I8, I16, I32, I64 with simple format
Note: The where clause method dispatch still fails at runtime because
the StaticDispatchConstraint doesn't store which type satisfies the
constraint. This needs type system changes to fix properly.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The std.debug.print calls pull in std.Thread which isn't available on freestanding targets. These were used for debugging auto-imported type method lookup and are no longer needed. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Skip 13 decode tests that require where clause method dispatch (not yet fully implemented - StaticDispatchConstraint doesn't store which type satisfies the constraint) - Remove std.mem.eql string comparisons in interpreter that violated the forbidden pattern check - Replace runtime flex var constraint checking with TODOs since it requires cross-environment ident comparison 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The fix in commit 5b15d42 for issue #8892 changed getRuntimeLayout to use the resolved (unwrapped) type's layout instead of the nominal type's layout. This fixed non-recursive nominal types like `ValueCombinationMethod := [Modulo]` but broke recursive nominal types like `Node := [Text(Str), Element(Str, List(Node))]`. For recursive nominal types, the layout is a box wrapping the actual tag union. The fix now: 1. First gets the nominal type's layout to check if it's a box 2. If it's a box (recursive type), handle it specially by building the inner tag union first, then boxing the result 3. Otherwise use the resolved type's layout (for non-recursive nominal types) This fixes the dbg_corrupts_recursive_tag_union.roc, repeating_pattern_segfault.roc, and aoc_day2.roc tests while preserving the fix for issue #8892. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Resolved merge conflict by keeping the recursive nominal type fix while adding the new layout_rt_var field from origin/main. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add resolveTypeAnnoRef helper to follow rigid_var_lookup references - Update decode methods in Builtin.roc to use type variable dispatch pattern - Convert runtime bounds checks in StackValue.zig to debug assertions - Update test TODO to document actual root cause (nominal type propagation) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
…umption Replace defensive loop with direct single-hop resolution in production code. Debug mode validates the one-hop assumption by running the full loop and asserting it produces the same result. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The function was based on a false premise that serialization somehow breaks union-find relationships. It doesn't - the union-find is serialized intact. Tests pass without it. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This fixes where clause method dispatch for decode tests by:
1. Adding current_block_statements tracking in Check.zig to find
local nominal type declarations during deferred constraint checking
2. Adding tryUnifyWithCompatibleNominal to search for nominal types
with compatible backing (empty_record or record with 0 fields)
that have the required method
3. Fixing unify.zig to accept nominal types with .record backing
that has 0 fields when unifying with empty_record (needed for
record extension syntax like `{}.{ ... }`)
All 13 decode tests now pass.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This function was added as a hack to handle polymorphic types resolving to different sizes, but the call sites were reverted in commit 96b4c39 because they caused segfaults with recursive types. The function definition was accidentally left behind as dead code. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Resolve conflict in Check.zig: keep tryUnifyWithCompatibleNominal functions for nominal type inference with method constraints. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Resolved merge conflicts in StackValue.zig: - Updated asClosure() calls to new signature (returns optional, no roc_ops param) - Preserved debug assertions for closure layout index bounds checking Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Previously, when calling a method on a structural type (record, tuple, tag union), the compiler would scan all module statements to find a nominal type with: 1. A compatible backing type 2. The required method This was problematic because: - It was O(n) per constraint where n = number of statements - Results were non-deterministic (depended on statement order) - Users got silent, surprising type conversions Now, calling a method on a structural type (other than is_eq) produces a clear "MISSING METHOD" error. Users must explicitly annotate their variables with the nominal type they intend to use. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This field was being set and restored but never actually read anywhere. Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Summary
Test plan
🤖 Generated with Claude Code