refactor: reorganize compiler

AGENTS.md

@@ -169,17 +169,24 @@ Tree-sitter: `((a) (b))` — Plotnik: `{(a) (b)}`. The #1 syntax error.
 ```
 crates/
   plotnik-cli/         # CLI tool
-    src/commands/      # Subcommands (debug, exec, langs, types)
-  plotnik-core/        # Common code
+    src/commands/      # Subcommands (check, dump, infer, tree, langs)
+  plotnik-core/        # Common code (Interner, Symbol)
   plotnik-lib/         # Plotnik as library
     src/
-      diagnostics/     # Diagnostics (user-friendly errors)
-      parser/          # Syntactic parsing of the query
-      query/           # Analysis and representation of the parsed query
-  plotnik-langs/       # Tree-sitter language bindings (wrapped)
-  plotnik-macros/      # Proc macros of the project
+      analyze/         # Semantic analysis (symbol_table, dependencies, type_check, validation)
+      bytecode/        # Binary format definitions
+      typegen/         # Type declaration extraction (bytecode → .d.ts)
+      compile/         # Thompson NFA construction (AST → IR)
+      diagnostics/     # User-friendly error reporting
+      emit/            # Bytecode emission (IR → binary)
+      parser/          # Syntactic parsing (lexer, grammar, AST)
+      query/           # Query facade (Query, QueryBuilder, SourceMap)
+      type_system/     # Shared type primitives
+  plotnik-langs/       # Tree-sitter language bindings
+  plotnik-macros/      # Proc macros
 docs/
-  adr/                 # Architecture Decision Records (ADRs)
+  adr/                 # Architecture Decision Records
+  binary-format/       # Bytecode format specification
   lang-reference.md    # Language specification
 ```
 

crates/plotnik-cli/src/commands/infer.rs

@@ -4,7 +4,7 @@ use std::path::PathBuf;
 
 use plotnik_lib::QueryBuilder;
 use plotnik_lib::bytecode::Module;
-use plotnik_lib::codegen::typescript;
+use plotnik_lib::typegen::typescript;
 
 use super::lang_resolver::{resolve_lang, resolve_lang_required, suggest_language};
 use super::query_loader::load_query_source;

crates/plotnik-lib/src/query/dependencies.rs → crates/plotnik-lib/src/analyze/dependencies.rs

@@ -12,16 +12,15 @@ use std::collections::{HashMap, HashSet};
 
 use indexmap::{IndexMap, IndexSet};
 use plotnik_core::{Interner, Symbol};
-
-use super::source_map::SourceId;
 use rowan::TextRange;
 
+use super::symbol_table::SymbolTable;
+use super::type_check::DefId;
+use super::visitor::{Visitor, walk_expr};
 use crate::Diagnostics;
 use crate::diagnostics::DiagnosticKind;
 use crate::parser::{AnonymousNode, Def, Expr, NamedNode, Ref, Root, SeqExpr};
-use crate::query::symbol_table::SymbolTable;
-use crate::query::type_check::DefId;
-use crate::query::visitor::{Visitor, walk_expr};
+use crate::query::source_map::SourceId;
 
 /// Result of dependency analysis.
 #[derive(Clone, Debug, Default)]

crates/plotnik-lib/src/query/link.rs → crates/plotnik-lib/src/analyze/link.rs

@@ -20,16 +20,15 @@ pub struct LinkOutput {
     pub node_field_ids: IndexMap<Symbol, NodeFieldId>,
 }
 
+use super::symbol_table::SymbolTable;
+use super::utils::find_similar;
+use super::visitor::{Visitor, walk};
 use crate::diagnostics::{DiagnosticKind, Diagnostics};
 use crate::parser::ast::{self, Expr, NamedNode};
 use crate::parser::cst::{SyntaxKind, SyntaxToken};
 use crate::parser::token_src;
-
-use super::query::AstMap;
-use super::source_map::{SourceId, SourceMap};
-use super::symbol_table::SymbolTable;
-use super::utils::find_similar;
-use super::visitor::{Visitor, walk};
+use crate::query::query::AstMap;
+use crate::query::source_map::{SourceId, SourceMap};
 
 /// Link query against a language grammar.
 ///

crates/plotnik-lib/src/analyze/mod.rs

@@ -0,0 +1,32 @@
+//! Semantic analysis passes.
+//!
+//! Provides analysis passes that transform parsed AST into analyzed form:
+//! - Name resolution (symbol_table)
+//! - Dependency analysis and recursion detection (dependencies)
+//! - Type inference (type_check)
+//! - Grammar linking (link)
+//! - Semantic validation (validation)
+
+pub mod dependencies;
+mod invariants;
+pub mod link;
+pub mod refs;
+pub mod symbol_table;
+pub mod type_check;
+mod utils;
+pub mod validation;
+pub mod visitor;
+
+#[cfg(test)]
+mod dependencies_tests;
+#[cfg(all(test, feature = "plotnik-langs"))]
+mod link_tests;
+#[cfg(test)]
+mod symbol_table_tests;
+
+pub use dependencies::DependencyAnalysis;
+pub use link::LinkOutput;
+pub use symbol_table::{SymbolTable, UNNAMED_DEF};
+pub use type_check::{TypeContext, infer_types, primary_def_name};
+pub use validation::{validate_alt_kinds, validate_anchors};
+pub use visitor::{Visitor, walk_expr};

crates/plotnik-lib/src/query/symbol_table.rs → crates/plotnik-lib/src/analyze/symbol_table.rs

@@ -10,8 +10,8 @@ use crate::Diagnostics;
 use crate::diagnostics::DiagnosticKind;
 use crate::parser::{Root, ast, token_src};
 
-use super::source_map::{SourceId, SourceMap};
 use super::visitor::Visitor;
+use crate::query::source_map::{SourceId, SourceMap};
 
 /// Sentinel name for unnamed definitions (bare expressions at root level).
 /// Code generators can emit whatever name they want for this.

crates/plotnik-lib/src/query/type_check/context.rs → crates/plotnik-lib/src/analyze/type_check/context.rs

@@ -233,7 +233,6 @@ mod tests {
     use std::collections::BTreeMap;
 
     use super::*;
-    use crate::query::type_check::types::FieldInfo;
 
     #[test]
     fn builtin_types_have_correct_ids() {

crates/plotnik-lib/src/query/type_check/infer.rs → crates/plotnik-lib/src/analyze/type_check/infer.rs

@@ -16,15 +16,15 @@ use super::types::{
 };
 use super::unify::{UnifyError, unify_flows};
 
+use crate::analyze::symbol_table::SymbolTable;
+use crate::analyze::visitor::{Visitor, walk_alt_expr, walk_def, walk_named_node, walk_seq_expr};
 use crate::diagnostics::{DiagnosticKind, Diagnostics};
 use crate::parser::ast::{
     AltExpr, AltKind, AnonymousNode, CapturedExpr, Def, Expr, FieldExpr, NamedNode, QuantifiedExpr,
     Ref, Root, SeqExpr,
 };
 use crate::parser::cst::SyntaxKind;
 use crate::query::source_map::SourceId;
-use crate::query::symbol_table::SymbolTable;
-use crate::query::visitor::{Visitor, walk_alt_expr, walk_def, walk_named_node, walk_seq_expr};
 
 /// Inference context for a single pass over the AST.
 pub struct InferenceVisitor<'a, 'd> {

crates/plotnik-lib/src/query/type_check/mod.rs → crates/plotnik-lib/src/analyze/type_check/mod.rs

@@ -6,7 +6,7 @@
 mod context;
 mod infer;
 mod symbol;
-mod types;
+pub(crate) mod types;
 mod unify;
 
 #[cfg(test)]
@@ -24,11 +24,11 @@ use std::collections::BTreeMap;
 
 use indexmap::IndexMap;
 
+use crate::analyze::dependencies::DependencyAnalysis;
+use crate::analyze::symbol_table::{SymbolTable, UNNAMED_DEF};
 use crate::diagnostics::Diagnostics;
 use crate::parser::ast::Root;
-use crate::query::dependencies::DependencyAnalysis;
 use crate::query::source_map::SourceId;
-use crate::query::symbol_table::{SymbolTable, UNNAMED_DEF};
 
 use infer::infer_root;
 

crates/plotnik-lib/src/query/type_check/unify.rs → crates/plotnik-lib/src/analyze/type_check/unify.rs

@@ -143,7 +143,7 @@ fn unify_type_ids(a: TypeId, b: TypeId, field: Symbol) -> Result<TypeId, UnifyEr
 #[cfg(test)]
 mod tests {
     use super::*;
-    use crate::query::type_check::TYPE_NODE;
+    use crate::analyze::type_check::TYPE_NODE;
     use plotnik_core::Interner;
 
     #[test]

crates/plotnik-lib/src/query/alt_kinds.rs → crates/plotnik-lib/src/analyze/validation/alt_kinds.rs

@@ -3,9 +3,9 @@
 //! Checks constraints that are easier to express after parsing:
 //! - Mixed tagged/untagged alternations
 
-use super::invariants::ensure_both_branch_kinds;
-use super::visitor::{Visitor, walk, walk_alt_expr};
 use crate::SourceId;
+use crate::analyze::invariants::ensure_both_branch_kinds;
+use crate::analyze::visitor::{Visitor, walk, walk_alt_expr};
 use crate::diagnostics::{DiagnosticKind, Diagnostics};
 use crate::parser::{AltExpr, AltKind, Branch, Root};
 

crates/plotnik-lib/src/query/anchors.rs → crates/plotnik-lib/src/analyze/validation/anchors.rs

@@ -6,8 +6,8 @@
 //!
 //! This validation ensures anchors are placed where they can be meaningfully compiled.
 
-use super::visitor::{Visitor, walk_named_node, walk_seq_expr};
 use crate::SourceId;
+use crate::analyze::visitor::{Visitor, walk_named_node, walk_seq_expr};
 use crate::diagnostics::{DiagnosticKind, Diagnostics};
 use crate::parser::ast::{NamedNode, Root, SeqExpr, SeqItem};
 

crates/plotnik-lib/src/analyze/validation/mod.rs

@@ -0,0 +1,16 @@
+//! Semantic validation passes.
+//!
+//! Validates semantic constraints that aren't captured by parsing or type checking:
+//! - Alternation kind consistency (alt_kinds)
+//! - Anchor placement rules (anchors)
+
+pub mod alt_kinds;
+pub mod anchors;
+
+#[cfg(test)]
+mod alt_kinds_tests;
+#[cfg(test)]
+mod anchors_tests;
+
+pub use alt_kinds::validate_alt_kinds;
+pub use anchors::validate_anchors;

crates/plotnik-lib/src/bytecode/ir.rs

@@ -11,7 +11,7 @@ use super::effects::{EffectOp, EffectOpcode};
 use super::ids::StepId;
 use super::instructions::{Call, Match, Return, select_match_opcode};
 use super::nav::Nav;
-use crate::query::type_check::TypeId;
+use crate::analyze::type_check::TypeId;
 
 /// Symbolic reference, resolved to StepId at layout time.
 #[derive(Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)]

crates/plotnik-lib/src/bytecode/mod.rs

@@ -5,13 +5,11 @@
 mod constants;
 mod dump;
 mod effects;
-pub mod emit;
 mod entrypoint;
 mod header;
 mod ids;
 mod instructions;
 pub mod ir;
-pub mod layout;
 mod module;
 mod nav;
 mod sections;
@@ -52,6 +50,4 @@ mod dump_tests;
 #[cfg(test)]
 mod instructions_tests;
 #[cfg(test)]
-mod layout_tests;
-#[cfg(test)]
 mod module_tests;

crates/plotnik-lib/src/query/compile/capture.rs → crates/plotnik-lib/src/compile/capture.rs

@@ -8,7 +8,7 @@ use std::collections::HashSet;
 use crate::bytecode::ir::EffectIR;
 use crate::bytecode::EffectOpcode;
 use crate::parser::ast::{self, Expr};
-use crate::query::type_check::{TypeContext, TypeId, TypeShape};
+use crate::analyze::type_check::{TypeContext, TypeId, TypeShape};
 
 use super::navigation::{inner_creates_scope, is_star_or_plus_quantifier};
 use super::Compiler;

crates/plotnik-lib/src/query/compile/expressions.rs → crates/plotnik-lib/src/compile/expressions.rs

@@ -12,7 +12,7 @@ use std::num::NonZeroU16;
 use crate::bytecode::ir::{CallIR, Instruction, Label, MatchIR};
 use crate::bytecode::Nav;
 use crate::parser::ast::{self, Expr};
-use crate::query::type_check::Arity;
+use crate::analyze::type_check::Arity;
 
 use super::capture::CaptureEffects;
 use super::navigation::{check_trailing_anchor, inner_creates_scope, is_skippable_quantifier, is_star_or_plus_quantifier};

crates/plotnik-lib/src/query/compile/mod.rs → crates/plotnik-lib/src/compile/mod.rs

@@ -28,9 +28,9 @@ use crate::bytecode::ir::{Instruction, Label, ReturnIR};
 use crate::bytecode::Nav;
 use crate::parser::ast::Expr;
 
-use super::codegen::StringTableBuilder;
-use super::symbol_table::SymbolTable;
-use super::type_check::{DefId, TypeContext};
+use crate::emit::StringTableBuilder;
+use crate::analyze::symbol_table::SymbolTable;
+use crate::analyze::type_check::{DefId, TypeContext};
 
 pub use capture::CaptureEffects;
 use scope::StructScope;

crates/plotnik-lib/src/query/compile/quantifier.rs → crates/plotnik-lib/src/compile/quantifier.rs

@@ -10,7 +10,7 @@ use crate::bytecode::ir::{EffectIR, Label};
 use crate::bytecode::{EffectOpcode, Nav};
 use crate::parser::ast::{self, Expr};
 use crate::parser::cst::SyntaxKind;
-use crate::query::type_check::TypeId;
+use crate::analyze::type_check::TypeId;
 
 use super::capture::{check_needs_struct_wrapper, get_row_type_id, CaptureEffects};
 use super::navigation::{is_star_or_plus_quantifier, repeat_nav_for};

crates/plotnik-lib/src/query/compile/scope.rs → crates/plotnik-lib/src/compile/scope.rs

@@ -5,7 +5,7 @@
 use crate::bytecode::ir::{EffectIR, Instruction, Label, MatchIR, MemberRef};
 use crate::bytecode::{EffectOpcode, Nav};
 use crate::parser::ast::Expr;
-use crate::query::type_check::TypeId;
+use crate::analyze::type_check::TypeId;
 
 use super::capture::CaptureEffects;
 use super::Compiler;

crates/plotnik-lib/src/query/compile/sequences.rs → crates/plotnik-lib/src/compile/sequences.rs

@@ -7,7 +7,7 @@
 use crate::bytecode::ir::{EffectIR, Instruction, Label, MatchIR, MemberRef};
 use crate::bytecode::{EffectOpcode, Nav};
 use crate::parser::ast::{self, Expr, SeqItem};
-use crate::query::type_check::TypeShape;
+use crate::analyze::type_check::TypeShape;
 
 use super::capture::CaptureEffects;
 use super::navigation::{compute_nav_modes, is_down_nav, is_skippable_quantifier, repeat_nav_for};

crates/plotnik-lib/src/query/codegen_tests.rs → crates/plotnik-lib/src/emit/codegen_tests.rs

@@ -4,7 +4,7 @@ use indoc::indoc;
 
 use crate::Query;
 use crate::bytecode::{Header, MAGIC, Module, QTypeId, VERSION};
-use crate::query::codegen::{StringTableBuilder, TypeTableBuilder};
+use super::{StringTableBuilder, TypeTableBuilder};
 
 #[test]
 fn emit_minimal_query() {
@@ -182,12 +182,12 @@ fn string_table_builder_intern_str() {
 
 #[test]
 fn type_table_builder_builtins() {
-    use crate::query::type_check::{TYPE_NODE, TYPE_STRING, TYPE_VOID};
+    use crate::analyze::type_check::{TYPE_NODE, TYPE_STRING, TYPE_VOID};
 
     let mut builder = TypeTableBuilder::new();
 
     // Build with empty context
-    let type_ctx = crate::query::type_check::TypeContext::new();
+    let type_ctx = crate::analyze::type_check::TypeContext::new();
     let interner = plotnik_core::Interner::new();
     let mut strings = StringTableBuilder::new();
 

crates/plotnik-lib/src/bytecode/layout.rs → crates/plotnik-lib/src/emit/layout.rs

@@ -5,8 +5,8 @@
 
 use std::collections::{BTreeMap, HashSet};
 
-use super::ids::StepId;
-use super::ir::{Instruction, Label, LayoutResult};
+use crate::bytecode::StepId;
+use crate::bytecode::ir::{Instruction, Label, LayoutResult};
 
 const CACHE_LINE: usize = 64;
 const STEP_SIZE: usize = 8;

crates/plotnik-lib/src/bytecode/layout_tests.rs → crates/plotnik-lib/src/emit/layout_tests.rs

@@ -1,10 +1,10 @@
 use std::num::NonZeroU16;
 
-use super::ids::StepId;
-use super::ir::{CallIR, EffectIR, Instruction, Label, MatchIR, ReturnIR};
+use crate::bytecode::StepId;
+use crate::bytecode::ir::{CallIR, EffectIR, Instruction, Label, MatchIR, ReturnIR};
+use crate::bytecode::EffectOpcode;
+use crate::bytecode::Nav;
 use super::layout::CacheAligned;
-use crate::bytecode::effects::EffectOpcode;
-use crate::bytecode::nav::Nav;
 
 #[test]
 fn layout_empty() {