A single slot always occupies the same flattened index, so filling,
clearing, or replacing it moves no other child -- only the new child's own
index needs setting. The generated single-slot setters now pass that index
to SetChildNode (a compile-time constant when no collection precedes the
slot; SetChild forwards its argument), which assigns childIndex directly and
never invalidates -- mirroring the IL AST's SetChildInstruction(ref, value,
index). Previously a set-from-null could not know the index here and fell
back to invalidating, forcing a later O(children) EnsureChildIndices rebuild.
With the indices now kept current by construction, NextSibling/PrevSibling
inline the validity check and skip the (non-inlinable) EnsureChildIndices
call in the overwhelmingly common already-valid case.
Together these idle the renumber machinery on a System.Private.CoreLib
decompile: EnsureChildIndices calls 53.4M -> 1.6M, actual rebuilds
1.97M -> 183K, elements renumbered 3.40M -> 757K. Output is byte-identical
and the Pretty suite stays green with CheckInvariant validating the
directly-assigned indices after every transform.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Removing a collection element invalidated the parent's whole flattened
index set, so the next sibling navigation rebuilt it in O(children) -- and
that renumber, not the array shift, was the dominant cost: a reverse
(tail-first) removal, which shifts nothing, was still quadratic, isolating
EnsureChildIndices as the culprit.
When a node's only collection is also its last slot it owns the contiguous
range [base, base + Count) with nothing after it, so an element's flattened
index is just base + its local position (base is the slot's declaration
position, since the preceding slots are all single children). On that
fast-path -- which the generator now flags, passing the base index -- Add
indexes only the appended element, Insert/Remove renumber only the shifted
suffix, and IndexOf is base-relative O(1); none of them invalidate. Other
shapes (several collections, or a slot after the collection) keep the
invalidate-and-rebuild fallback. Tail and scattered removal, and removal
during traversal, no longer pay the per-operation renumber.
Microbenchmark, removing every element of an N-element block: tail-first at
N=32000 went 1493 ms -> 0.5 ms (now O(N)); front-first is ~1.6x faster and
no longer renumbers (its residual cost is the inherent array shift). Output
is byte-identical and the Pretty suite stays green with CheckInvariant
validating the maintained indices after every transform.
Assisted-by: Claude:claude-opus-4-8:Claude Code
A node's children carry a cached flattened childIndex, rebuilt by
EnsureChildIndices after any structural mutation invalidates it. The
single-slot setter and the collection indexer invalidated the whole set on
every in-place replace, so the next sibling navigation (e.g. the visitor's
NextSibling walk) rebuilt all indices in O(children). A transform that
replaces each element of a block while traversing it was therefore O(N^2).
But replacing a child in place does not move anything: a single slot always
occupies the same flattened index, and a replaced collection element keeps
its position. So carry the old child's index to the new child and skip the
invalidation. Setting or clearing a slot still invalidates, since the new
child's index is not known locally; a stale carried value is corrected by
the next renumber anyway.
Microbenchmark (replace every statement in an N-statement block): at
N=32000, 2158 ms -> 1 ms. Output is byte-identical and the Pretty suite
stays green with CheckInvariant active.
Assisted-by: Claude:claude-opus-4-8:Claude Code
The generated GetChildNodes materialized a List<AstNode> (plus a boxed
List.Enumerator at every foreach) for each node, so AstNode.Children and
the visitor's per-node child walk allocated three objects per traversal.
Decompiling System.Private.CoreLib that came to ~1.7 GB of extra garbage,
roughly +7% over the linked-list model the slot tree replaced. A yield
iterator removes the List but trades it for an equally costly per-node
state machine, so it is not enough on its own.
Enumerate children through a by-value struct enumerator over the existing
FirstChild/NextSibling primitives, capturing each child's successor before
it is yielded so a transform may still remove or replace the current child
mid-traversal. AstNodeCollection<T> gets the same struct treatment for a
direct foreach. Child enumeration now allocates nothing, bringing total
allocations back to the linked-list baseline at byte-identical output
(full Pretty suite green with CheckInvariant active).
Assisted-by: Claude:claude-opus-4-8:Claude Code
The slot kind is now the canonical typed CSharpSlotInfo<T> in Slots, matched by
object identity (the IL SlotInfo model), instead of a parallel SlotKind enum.
CSharpSlotInfo.Kind points at the canonical shared slot; a Slots constant is its
own kind (null Kind, never read -- only per-node slots are asked for their kind),
so there is no self-reference. Child access (GetChild/GetChildren/SetChild,
GetCollectionByKind, AddChild/InsertChild) and the polymorphic
node.Slot.Kind == Slots.X comparisons key on the canonical reference; the
generated SlotKind enum is removed. No behavior change.
Assisted-by: Claude:claude-opus-4-8:Claude Code
GetChild/GetChildren/SetChild take a typed CSharpSlotInfo<T> and infer the child
type from the slot, delegating to the existing kind-based lookups. Call sites can
pass a node's static slot (e.g. node.GetChild(PropertyDeclaration.GetterSlot))
instead of an explicit type argument plus a SlotKind.
Assisted-by: Claude:claude-opus-4-8:Claude Code
A DEBUG-only structural check, the analog of ILInstruction.CheckInvariant, run
after every AST transform in RunTransforms. It recursively asserts that each
required (non-optional) single slot holds a child, every child's Parent points
back, the stored flattened index matches the slot position, and the runtime type
fits the slot, so a transform that corrupts the tree fails at that transform
instead of as a downstream output diff. CSharpSlotInfo gains an IsOptional flag
(emitted by the generator) so the check can tell required from optional slots.
Assisted-by: Claude:claude-opus-4-8:Claude Code
The flags word existed to back per-node bit state, but its only user was Identifier.IsVerbatim,
and it was retained as the seed for a deferred flag-packing optimization. Measuring that optimization
showed it saves essentially nothing -- the CLR already packs a node's bools and narrow enums into the
object's existing alignment padding (ParameterDeclaration and BinaryOperatorExpression: 0 bytes saved
by hand-packing). So IsVerbatim becomes a plain bool (free, it lands in padding) and the flags field is
removed. Removing it shrinks every AstNode by 8 bytes -- the field occupied its own aligned slot rather
than pairing with childIndex as assumed (a minimal node measured 64 -> 56 bytes), a universal win across
the whole tree that the per-node packing never delivered.
Assisted-by: Claude:claude-opus-4-8:Claude Code
GetChildByRole<T> returned default! while it can legitimately yield null (empty slot
or a kind the node does not declare); it now returns T?. Callers that already null-checked
are unchanged; the few that rely on a structural guarantee (name tokens, label identifier,
property accessors) now state it with an explicit !, and MethodDeclaration drops a redundant
cast that the lie required.
Also removes two members left dead by the Role removal: the AddAnnotation override that
only forwarded to the base, and GetCSharpNodeBefore, which ignored 'this' and had no callers.
On the output path, GetChildNodes() allocates a snapshot List per call; the read-only
.Children traversals that never mutate the tree now walk the live FirstChild/NextSibling chain
(O(n), allocation-free, since child indices are cached), an EmptyStatement child check uses
HasChildren, and the label-statement sibling scan breaks on the first match instead of walking
to the end of the block. The generated GetChildNodes snapshot is kept where traversal may mutate.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Nodes use nullable reference types now, so the generated null-object
node and its hasNullNode / NeedsNullNode / NullNodeBaseCtorParamCount
plumbing are dead. Drop them and reword the affected comments.
AstNode.Children and the visitors' VisitChildren walked children via
FirstChild/NextSibling, an O(slots) index rescan per step. The generator now
emits GetChildNodes, which materializes the children in slot order in one pass
(O(children)); Children and VisitChildren iterate that. The snapshot makes the
walk tolerant of the visitor removing or replacing the current child (the
mutation pattern the old capture-next loop supported) without re-feeding the
loop.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Slot identity already lived in node.Slot/CSharpSlotInfo/SlotKind after
the storage flip, so the parallel Role/Role<T> child model was redundant.
The Role-keyed mutation/query API is reexpressed over SlotKind, and the
role-index packing on AstNode flags is gone.
Because a node's Slot is now derived from its index in its parent rather
than stored on the child, the located-AST reattach in
InsertMissingTokensDecorator must capture the child's slot kind before
Remove() detaches it.
Assisted-by: Claude:claude-opus-4-8:Claude Code
With every optional slot nullable, the null-object pattern is dead. Generated
non-nullable getters return the backing field directly, which surfaced a last
tier of slots the decompiler legitimately leaves empty (omitted range operands,
an implicitly-typed array creation, unnamed parameters, an unbound generic
argument, and others) and flips them to nullable too. The machinery is then
removed entirely: the per-node null classes, the .Null statics and
VisitNullNode, AstNode.IsNull, the role null object, and Identifier.Null.
AcceptVisitor becomes unconditionally generated, and consumers move from
.IsNull to is null and from unconditional visits to ?.AcceptVisitor.
Assisted-by: Claude:claude-opus-4-8:Claude Code
NodeType was NRefactory's coarse node category, but only three reads remained
here: two checks now expressed as "is not Trivia" and one debug assert for the
pattern category. Remove the enum, the abstract property, every per-node
override, and the generator's emission, preserving the pattern-placeholder case
through an IPatternPlaceholder marker interface the output-visitor assert
checks. Also remove the unused PrimitiveExpression.AdvanceLocation helper.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Preparation for making optional single-value slots nullable: a role may omit
its null object and a slot may be cleared via SetChildByRole(role, null), and
the generated DoMatch emits MatchOptional for a nullable single child so an
absent child matches correctly. Both are inert until a slot is actually made
nullable.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Introduce the successor to node.Role for child-slot identity: the generator
emits a CSharpSlotInfo per [Slot], exposed as node.Slot, plus a shared SlotKind
enum for the polymorphic "is this node in an embedded-statement / condition /
base-type slot?" comparisons a per-node identity cannot express. Migrate the
printer and transform position checks from node.Role to node.Slot and
node.Slot.Kind, and read identifier children and role-keyed writes through the
typed properties. Role is still present and is removed later; output is
unchanged.
Assisted-by: Claude:claude-opus-4-8:Claude Code
The pattern matcher walked collections through INode.Role/FirstChild/
NextSibling, skipping siblings of a different role. Now that each
AstNodeCollection<T> is already the per-role child list, the engine matches two
collections by list index, and INode sheds Role/FirstChild/NextSibling
entirely. A collection exposes its IReadOnlyList<INode> view through a cached
adapter rather than implementing the interface directly, so a typed collection
does not become ambiguous for LINQ. Characterization tests pin the matcher's
behavior first.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Children were kept in a per-node doubly-linked list with the slot accessors
layered over it as a view. Storage now is the slot model: each node stores its
children in generated backing fields, AstNodeCollection<T> is backed by a
List<T>, and the flattened child-index space is owned by generated
GetChildCount/GetChild/SetChild/GetChildSlot members, with sibling navigation,
the role API and Clone re-expressed over them and indices renumbered lazily. A
DEBUG CheckInvariant runs after each transform, the analog of the IL
pipeline's per-transform check, so a transform that corrupts the tree fails at
that transform. Output is unchanged.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Comments and preprocessor directives were positional children interleaved
into the child list, and punctuation, keywords and operators were token-node
children. Add a leading/trailing trivia side-channel for comments and
directives, emit it from the output visitor, and re-home every comment
receiver onto it (including inside-block comments as comment-only empty
statements and undecodable attribute arguments as an ErrorExpression). With
locations and sequence points no longer sourced from token nodes, stop
reconstructing them on the locations path and delete CSharpTokenNode,
CSharpModifierToken and InsertSpecialsDecorator. The AST no longer carries
token children or positional comments; output is byte-identical.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Source locations were virtual, computed by recursing to the first and last
child, whose leftmost and rightmost leaves are token nodes; sequence-point
coordinates likewise came from reconstructed token nodes. Store locations as
fields assigned while printing, and derive sequence-point coordinates from the
surrounding real nodes plus the decompiler's fixed formatting, so neither
depends on token children. The using/foreach await modifier becomes a plain
bool field. Characterization gates lock the emitted locations and PDB
coordinates, which are unchanged.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Each child of a C# AST node is declared as a [Slot] partial property, and the
source generator emits the accessor bodies and an ordered slot schema
(SlotCount/GetSlotRole/IsCollectionSlot) from them. Generating the schema
keeps slot order from being mis-stated by hand and lets a DEBUG invariant
check declared slot order against document order on every decompile. The node
hierarchy is converted family by family; the EntityDeclaration leaves flatten
their inherited Attributes/ReturnType/NameToken into each leaf's ordered slot
set. Storage stays the NRefactory linked list at this stage, so only the
declaration model changes and output is unchanged.
Assisted-by: Claude:claude-opus-4-8:Claude Code
The generator emits the IAstVisitor interface, the AcceptVisitor overloads,
and the null-node and pattern-placeholder nodes from [DecompilerAstNode]
declarations, so drop the hand-written equivalents across the C# AST: per-node
AcceptVisitor/DoMatch, the #region Null / #region PatternPlaceholder blocks,
IAstVisitor.cs, and now-dead usings. Also adds AccessorKind and moves
IdentifierExpressionBackreference into the PatternMatching folder.
Assisted-by: Claude:claude-opus-4-8:Claude Code
Introduce a Roslyn source generator that emits the visitor boilerplate for
the C# AST from [DecompilerAstNode]-tagged node declarations: the
IAstVisitor interface, the AcceptVisitor overloads, the pattern-placeholder
nodes, and the initial DoMatch support. AccessorKind lets an accessor's
keyword be chosen independently of its role, an early step toward shedding
the NRefactory role model.
The C# AST inherited NRefactory's freezable model (IFreezable, Freeze,
IsFrozen, a frozen flag bit, and ThrowIfFrozen guards on every mutator),
but the decompiler never uses it: nothing calls Freeze(), not even the
generated null-node singletons, so every IsFrozen guard only ever
evaluated false. The decompiler is single-threaded and never shares or
freezes nodes. Remove the whole apparatus as preparation for the
slot-based AST rewrite, which has no place for it. Roles are untouched
here, so the flags word keeps its role index; only the freed frozen bit
goes away.
Siegfried Pammer
Peter Crabtree
Daniel Grunwald
dotnet format
Christoph Wille