Skip to content

Put every position index on one generic tree (unify step 1) - #516

Merged
DongyuZhao merged 292 commits into
mainfrom
claude/project-thread-raxrtq
Nov 15, 2026
Merged

DongyuZhao merged 292 commits into
mainfrom
claude/project-thread-raxrtq

Conversation

@DongyuZhao

Copy link
Copy Markdown
Contributor

Step 1 of the unified stream / edit / LSP plan: one generic tree carries every position-ordered index, so stream, edit and locate share the same structure.

What changes

  • core/tree.{h,c}: a B+ tree of fixed-size entries. Leaves hold 64 entries and inner nodes fan out to 16. A focus (front) splits the tree: the front is in new-text terms and the back stays in old-text terms until the splice's tree_shift. The tree also has:
    • pending shifts;
    • optional end measures (tree_past);
    • key ordering for label trees (tree_seek);
    • handles that a moved callback keeps current.
  • Moved onto the tree:
    • input lines (a line's facts are dropped as it moves to the back);
    • reach and uses;
    • committed tables, where listed entries go from id to handle, and their label trees;
    • roster places (record id to handle);
    • headings and heading-name stems;
    • session child widths;
    • the label key tree (the old one in map.c is deleted).
  • core/zipper.h deleted.
  • Read-only document access goes through the pure tree_peek / tree_seek. Concurrent reads of one document never mutate it, which keeps the thread-safety contract in markdown_core.h.
  • Stage gate kept:
    • leaves are not zeroed;
    • the next line's start is read from its terminator bytes;
    • an edit's line insert is a cold path;
    • properties walk lines in runs (tree_run);
    • a new child takes its offset from the line at hand (a debug assert checks this).
  • Build lists: tree.c is added to the Swift, Kotlin and ES build lists.

The second commit is a -s ours merge of the branch's old, already-merged history (#515). It brings in no content and only makes the push fast-forward.

Testing

  • Debug build: full ctest passes, 151/151, including all incremental shards.
  • Stage gate under gcc 15 (Ubuntu 26.04 chroot), the two documents that were over before the squeeze:
    • metadata-types-host-dialect +1.58%
    • mixed-definitions-paired-dialect +1.68%
  • The stream gate is left to CI.
  • clang -Wshorten-64-to-32 is clean, clang-format has been applied, and Prettier is clean.

Generated by Claude Code

claude added 30 commits October 29, 2026 03:59
The splice's guard named List, Callout and Table in the engine. The
elements that write the block before them now declare its kinds
(`trail_kinds`: the paragraph's block identifier, the table's caption),
and the dialect marks them TRAILED.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The edit's limit is computed once per reading and rechecked only on the
cold path, so a parse without an edit keeps the line loop it had. A
structure now says it is trailed (`trailed`), folded into the kind record
with its other facts, in place of a per-element list walked at seal.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
An edit's rest point no longer reads a child's absolute place: the
session keeps the widths of the root's children in chunks of 64, made at
the first edit and mended after each one for the chunks of the children
it read and the first it took back. A debug build checks every mend
against widths made anew and against the children's places.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The lines of a flushed parse past where an edit puts it back stand in a
back list, last first, each measured from the text's end, so an edit
before them moves none: a reopen moves only the lines between the old
and the new rest point, and a splice drops the ones the reading read
again and copies nothing.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The reach of the readings past where an edit puts the parse back stands
in the edit's back, measured from the text's end, as its lines do: a
splice drops what the reading read again and notes nothing anew.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
What an edit puts aside of the registry and of the headings is measured
from the text's end, as its lines and reach are, so the splice takes it
back in the new text's terms and no delta crosses the element hooks.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Nothing reads a published node's place again: the old children an edit
keeps or takes back are measured from where the child before the old
tail ends, by their extents, for the reading's candidates, its splice,
its records and its publish, so the splice no longer walks the tail to
move the places.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The FINAL prefix's edit check made GCC 15 stop inlining
publish_relation_begin into publish_matched, a per-step cost on the
stream session path. It is always inlined now, as its neighbours are.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The widths and the reread range are made at a session's first edit, so a
session that is only streamed keeps the size it had: the 32 bytes they
added moved the session into another malloc size class, and the heap's
layout after it, which cost some stream windows 3%.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…, stage 1)

## What

Before this change, an edit inside a list item, callout or footnote
reread the whole root child that held it: one edit near the end of a
long list read the entire list again. Now the reading starts at a rest
point inside that container, at any depth. It reopens the chain of
containers from the root's child down to the rest point. Then it reads
on from there until the top-level resync (E1a) takes the old blocks
back.

Plan: `e1b-nested.md`, stage E1b-1. Resync and splice at depth is the
next stage.

## How

- **Nested rest point** (`session_rest_in`, session.c). The search
  descends while the held child is a container whose element sets
  `reopens` and the edited line is past its first line. At each level it
  picks the last child that:
  - begins after a blank (or prefix-only) line;
  - has a gap before its previous sibling, so no lazy line or split
    reference paragraph runs into it;
  - has no reading that started before the chain spanning its start.
  Child windows come from `markdown_core_node_window`.
- **Reopen the chain** (`S_reopen_chain`, blocks.c). Each chain node
  goes OPEN again, and its kept children are folded again. The new
  `reopen_block` hook lets an element undo what its close decided. Here
  that is the list's looseness. The engine still names no kinds: the
  hook and the `reopens` flag come from the elements. Elements opt in
  with `reopens`: list, callout, footnote and specimen. Tables are left
  out.
- **The blank state before the rest point** (`markdown_core_parser_rest_blank`).
  It tells each chain level whether the line before the rest point was
  blank to it, so the deepest node's last-line-blank flag matches a
  fresh parse.
- **The reach log.** It keeps each entry's raw end and builds a separate
  prefix maximum (`furthest`) for the span queries. This lets the root
  chain node's own close entry be found and dropped at reopen (it closes
  again where the reading closes it). `markdown_core_parser_spanned_inside`
  finds readings that start inside the chain.
- **Publish** (ast.c). Each chain node is recorded as open under its
  owner, with its kept children FINAL. Definitions in a chain node keep
  their place in the table (`publish_chain_rank`), so nested footnotes
  list in source order.

Hot paths are unchanged. Everything new runs at edit time or behind
`edit->depth`.

## Results

- The differential fuzz compares against a fresh `Document.parse`:
  20,000 container-heavy documents × 20 edits, plus 10,000 mixed ones.
  All match. The rest points it hit were at depths 1 to 4.
- Bytes reread per edit over `specs/incremental`, before → after:
  - list-ascii-16k: 11,694 → 5,833.
  - list-utf8-16k: 12,125 → 6,196.
  - Every other family is the same or lower. The whole set averages
    10,719 → 10,466.
- The incremental oracle suites and the api_incremental tests pass.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
- A container with one child (a one-item list, a quote holding one list)
  is now descended too: the search needs a child to go into, not two.
- A reopened root child closed by its own end condition (a directive's
  closer) notes its close at its last line, as one the input's end
  closes does, so that entry is the chain's own close too.
- Definition lists and directives reopen (`reopens`), as the plan's
  E1b-1 lists them. A definition's body is not published and has no
  window, so no rest point is looked for below one.

Bytes reread per edit over specs/incremental: quote-16k 15.9k -> 10.8k,
opener-directive-16k 16.3k -> 10.3k, nameless-directives 213 -> 164; the
whole set 10,466 -> 9,891 (main: 10,719).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…pth (step E1b, stage 2)

A reading reopened inside a container now resyncs at the child the chain
let go, at any level, and splices the rest of the old children back: each
block of the chain closes as it closed before, with the flags its children
fold refolded. Lists whose items abut (`abuts`) rest and resync between
adjacent items and at their first item. `trailed` becomes per-kind
(`trailed_kinds`), so list items no longer block a resync. The `hard`
incremental family now runs on the session subject.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The engine names no element kinds: the lookahead that tells the line
leaves the block open inside the chain block asks for a child of the kind
that block already holds, not for a paragraph.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The reopen kept the content marks whose source fell before the rest
point, but marks read inside a nested block hold sources relative to that
block, so they never compared past it: a nested list read again on every
edit grew line_marks without bound (22 MB after 2000 edits of the 16k
list fixture, 11x a fresh parse). Every map the flush made is complete, so
the reopen now truncates to the identity run.

The fuzz runner reads each input twice: once as appends, as before, and
once switching to edits at the first head byte with the high bit set, so
the edit oracle sees random inputs too. Edits whose ends fall after a
continuation byte at a line start are skipped, since such positions are
ambiguous in UTF-16.

An api test makes a thousand edits to a small nested list and checks the
session holds no more than after the first hundred; it fails without the
fix (4x growth).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
… part 2)

The compaction of plan 12.4 (E-D4). The pool keeps the slots each reading
lets go for the next, so it holds what the most text the session held
needed. The session keeps that peak, and an edit that leaves less than
half of it reads the whole text: it drops the pool's released slots and
slab first, so the new nodes take new slabs, and again after, so the old
slabs go with the old nodes. The input's lines, the content marks, the
line facts and the edit's scratch then shrink to the text
(markdown_core_parser_compact, markdown_core_array_shrink). Ids stay, as
for any whole reading.

On the 16k list fixture an edit that deletes 90% of the text now leaves
1.35x the storage of a fresh parse, down from 7.6x. An api test checks
2x after such an edit, and fails without the compaction.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Three cases where a snapshot settled what a longer input reads otherwise,
found by the edit-aware fuzz runner (4.6, then 4.1 at the flush):

- A `$` pair whose body a code span begins makes no formula, but the
  pairing still took the delimiters between its two away, so a reading
  from a settle point between them paired a `*` the whole reading drops.
  The rejected pair now holds the run undecided from its opener
  (markdown_core_inline_state_undecide).
- An attribute block whose quote the input's end leaves open, or closes
  at the very end, reads that quote as an unquoted value's byte; the
  input going on undoes it. The recogniser now says it ran out
  (`ran_out`), and the run reports the end read as it does for a block
  that does not close.
- A topic's name reads the character after it whole, up to four bytes,
  but reported an end read only within three of the end. It now reports
  within four, and five after a `/`.

Each case is a source of session_flush_matches_a_parse, which fails
without its fix.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…t openers

A formula scanner asks how many `$` openers are unmatched to read a `$` as
a closer. An opener that an emphasis pair around it takes away still counted
at those scans, so a reading from past the emphasis, without the opener,
could read a later `$` as an opener instead. The core now notes the rules
whose scanners ask that count, and the start of a removed opener of such a
rule is an undecided start.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The wire's markdown_core_wire_session_edit answers an edit's document as a
step's message. Swift's MarkdownSession.edit(from:to:with:), and Kotlin's
and ECMAScript's edit(start, end, text), replace the text between two
cursors in the session's unit and return the parse of the new text. An edit
before the flush throws the new code sessionStreaming (SESSION_STREAMING);
one out of bounds or backwards throws outOfBounds and leaves the context.

Each binding sweeps an insertion and its undo over every cursor of a text
and checks each document against a fresh parse, and checks that a node the
edit did not reach is the old node.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Document.content holds existential markup, which == cannot compare.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The allocator seam audit reads a probe armed after the parse as one that
measures nothing; the test refuses the read's next allocation instead.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
An authored anchor can hold the mark a long name's key starts with, and
its bytes could then be a computed long name's key: a short anchor
reserved an unrelated long heading's name. A name that starts with the
mark is now known by its key whatever its length, and no computed name
holds the mark, as the projection drops controls.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
A long heading's anchor, and a literal held in two parts, are made at
their first read, which can be refused storage after the parse
succeeded. The Swift builder read them as calls that cannot fail. It now
makes them as it queues each node and throws the refusal from the parse
or session step; the records built after read what the engine kept.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
When an edit shortens a heading that parsed from its seams, undoing the
seams keeps its old children while the heading reads as a paragraph; the
records named its old id but compared the paragraph's kind, so the
heading's content lost its ids. The publish now matches the kind it named
the node as, as it does the id.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
A heading named from its projection clears the name once, the key's
choice of path inlines into the naming loops, and the two-lane Horner
takes four bytes each turn of its loop.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…given hash

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
A heading's node keeps its base (the projection of its text, split after
an edit that leaps inside it) and a numeric suffix; its anchor is the base
and "-k". One naming model serves every length: the committed pass and
an edit's renames read names as pieces and compare them byte for byte,
keyed by their measure, so there is no 1024-byte cut and no digest key.

- Stems keep no bytes, only a measure and a witness (first member's base,
  else a child's, else the heir or relic), and are removed by id.
- An edit's new bases wait for the edit's end and are written in place
  (the old base's kept front and back around the new middle).
- The public anchor borrows the base when it is whole and unsuffixed, and
  is otherwise joined on first read.
- A heading whose text projects to nothing borrows a static "section".
- Roster ANCHOR keys are hashed by measure.

Also fixes a locate crash present before this change: a heading's run that
reached the text's end, prepared before a later heading's run began, lost
the sink root and read NULL when it finished. Only a run that reaches the
end now replaces the root.

The OOM dump sweep parses a fresh document per allocation, since a read
keeps the anchors it joins.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…ader refuse

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…names as bytes

The measure is now a polynomial in a generator G of the residues modulo
2^61 - 1, read through byte tables, so no repeating pattern of byte changes
cancels; the Codex family of colliding headings no longer shares a measure.
The committed naming pass, which reads every base whole, files names under a
byte hash read in parts (markdown_core_key_stream), so a base's hash is read
once for all its suffixes; edits keep the measure. Holders whose bases are of
the name's base size are compared by suffix and one comparison of bases.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…enerator

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
…er for every heading

Swapping the projection's storage into the node left the pass a buffer to
grow again from nothing for the next heading, several reallocations per
heading; a copy is one allocation, as the joined anchor was before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
core/tree.{h,c} is one B+ tree of fixed-size entries (leaves of 64,
fanout 16) with a focus that splits it into a front in new text terms
and a back in old text terms, pending shifts, optional end measures
(tree_past), key ordering for label trees, and handles kept current by
a `moved` callback. It replaces the zipper and every hand-rolled index:

- input lines (facts dropped as lines move to the back), reach, uses
- committed tables (listed entries by id -> handle) and their labels
- roster places (record id -> handle)
- headings and heading-name stems
- session child widths
- the label key tree (map.c)

core/zipper.h is gone. Document reads use the pure tree_peek and
tree_seek, so concurrent read-only access stays safe.

The stream driver keeps its per-line cost within the stage gate: leaves
are not zeroed, the next line's start is read from the terminator, an
edit's line insert is a cold path, properties walk lines in runs
(tree_run), and a new child takes its offset from the line at hand.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
@DongyuZhao
DongyuZhao requested a review from a team November 14, 2026 20:29
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 14, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-15T02:30:58.195542Z 8dcc814 New commits
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

Copy link
Copy Markdown
Contributor Author

Stage gate on 22bdc15: 1573/1584 stages pass. All of source_to_buffer passes. The 11 failures are in buffer_to_ast, all on tiny documents (base 3.8k to 11k Ir), each about +300 Ir over base.

I traced one of them, leaf-formula-paired-dialect, under callgrind (gcc 15). S_finish_parse makes the same 6 markdown_core_free calls in base and head. In head those calls cost 384 Ir instead of 77, and there are no other differences in that stage.

I believe the cause is where the first free() of the process now falls:

  • On main, the line and reach arrays grow with realloc during source_to_buffer. Those reallocs do the first free() of the process, including glibc's tcache_free_init.
  • With the trees, the parse never frees anything, so the first free() lands in buffer_to_ast.
  • Seen in the profile: in head, tcache_free_init (141 Ir) is charged to the frees in S_finish_parse. I haven't confirmed where it falls on main.
  • Seen in the profile: on the same document, source_to_buffer is about 1k Ir under base (45,650 vs 46,671).

The fix I can see is to make the parse free something early, which would only move glibc's one-time cost from one stage to the other. I'm asking the owner whether to accept these 11 rows.


Generated by Claude Code

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 232d3d0ab6

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread packages/markdown-core/elements/ast.c Resolved
An edit's splice cut, shifted and unlabeled one table at a time, so a
reference definition whose label fell to a later heading looked the
heading up in old-text terms and lost the label. Take and shift every
table first, then move the labels on. A sweep of whole-line edits over
texts whose labels cross tables checks the trees and tables against a
fresh parse.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
@github-actions

github-actions Bot commented Nov 14, 2026 •

Copy link
Copy Markdown

Benchmark

Commit: 8dcc8141a0e8dbd6d06b42d7d20ed9e774f409b5 · CI run · CI status: failure

Run and full reports.

Benchmark is required when CI inputs require execution. This ordinary PR comment does not create a review thread to resolve.

Parse stages

Baseline: 30dc7361dd076011eec98cc70daeada04d7af3c1. 792 document workloads, measured in the same job.

Core instructions (Ir) Base PR PR / base
Source → buffer 61,981,423 61,731,375 0.9960×
Buffer → AST 224,962,044 224,915,450 0.9998×
Both stages 286,943,467 286,646,825 0.9990×
Both stages, ASCII documents 143,039,098 142,870,507 0.9988×
Both stages, UTF-8 documents 143,904,369 143,776,318 0.9991×

Totals sum this finite workload; they are not elapsed time or a general speedup claim. Parser creation and release are not parsing and are in no figure here.

Stage budget (+2% per document and stage): 1,573/1,584 passed, 11 exceeded. Required when CI inputs require execution.

Largest stage ratios (up to 10 document stages)
Case Stage Base Ir PR Ir PR / base
formula-bracket-block-paired-dialect Buffer → AST 3,778 4,051 1.0723×
leaf-formula-paired-dialect Buffer → AST 3,778 4,051 1.0723×
formula-bracket-block-utf8-paired-dialect Buffer → AST 3,778 4,051 1.0723×
leaf-formula-utf8-paired-dialect Buffer → AST 3,778 4,051 1.0723×
common-code-fence-backtick-utf8-paired-dialect Buffer → AST 4,018 4,291 1.0679×
common-code-fence-backtick-utf8-paired-common Buffer → AST 4,018 4,291 1.0679×
common-code-fence-tilde-utf8-paired-dialect Buffer → AST 4,018 4,291 1.0679×
common-code-fence-tilde-utf8-paired-common Buffer → AST 4,018 4,291 1.0679×
directive-leaf-bare-paired-dialect Buffer → AST 4,246 4,519 1.0643×
directive-leaf-bare-utf8-paired-dialect Buffer → AST 4,246 4,519 1.0643×

Corpus: 0ac517a3db11cc260db2fe63097f40cd67fb8881a2eab551570082c36b52c034 · Grammar: 003131631321aed6ce88b38f9b46bdfcb504b23929ae39571ed04e245b65c97f.
Full reference comparisons, all workloads, toolchain identities and raw profiles are in the run artifacts.

Grammar-equivalent reference comparisons

A = Core on dialect input; B = Core on common input; R = cmark/cmark-gfm on common input. Ir covers the two parse stages. Whole means the declared language; local excludes the unmatched host. Each input is measured once. No AST equivalence or equal native output cost is asserted.

Certificate Alphabet Scope Units Bytes A/B A/B B/R A/R
alpha-list-grammar-v2 ASCII whole 26 5,579/5,596 1.0015× 1.4202× 1.4223×
anchor-grammar-v2 ASCII whole 12 2,405/2,513 0.9182× 1.1006× 1.0106×
anchor-last-empty-grammar-v2 ASCII whole 12 458/374 1.0362× 1.5018× 1.5562×
anonymous-container-grammar-v2 ASCII whole 12 2,429/2,357 1.0510× 1.5134× 1.5905×
attribute-autolink-grammar-v2 ASCII whole 12 479/551 0.7781× 1.4729× 1.1461×
attribute-bare-grammar-v2 ASCII whole 12 240/444 0.9555× 1.4803× 1.4144×
attribute-code-grammar-v2 ASCII whole 12 575/635 0.8261× 1.4951× 1.2352×
attribute-double-grammar-v2 ASCII whole 12 1,649/1,661 1.0258× 1.3879× 1.4237×
attribute-empty-grammar-v2 ASCII whole 12 252/264 0.9970× 1.5360× 1.5314×
attribute-escaped-value-grammar-v2 ASCII whole 12 446/458 0.9855× 1.4299× 1.4092×
attribute-fence-grammar-v2 ASCII whole 12 587/659 0.5133× 1.3128× 0.6739×
attribute-heading-grammar-v2 ASCII whole 12 1,589/1,577 0.9690× 1.7035× 1.6507×
attribute-image-grammar-v2 ASCII whole 12 1,718/1,790 0.9479× 1.4417× 1.3665×
attribute-link-grammar-v2 ASCII whole 12 1,706/1,778 0.9452× 1.3826× 1.3069×
attribute-newline-grammar-v2 ASCII whole 12 446/410 1.0637× 1.5334× 1.6310×
attribute-order-grammar-v2 ASCII whole 12 632/575 0.9893× 1.4767× 1.4609×
attribute-reference-grammar-v2 ASCII whole 12 3,822/3,774 1.0091× 0.8904× 0.8985×
attribute-single-grammar-v2 ASCII whole 12 1,649/1,661 1.0258× 1.3879× 1.4238×
attribute-unquoted-grammar-v2 ASCII whole 12 1,625/1,661 1.0263× 1.3879× 1.4244×
block-id-container-list-grammar-v2 ASCII whole 12 2,405/2,501 0.9218× 1.1078× 1.0212×
block-id-container-table-grammar-v2 ASCII whole 12 1,453/1,549 0.8643× 0.6747× 0.5832×
block-id-escaped-grammar-v2 ASCII whole 12 204/204 1.0000× 1.4414× 1.4414×
block-id-list-grammar-v2 ASCII whole 12 3,539/3,707 0.8373× 0.9277× 0.7768×
block-id-paragraph-grammar-v2 ASCII whole 12 2,369/2,489 0.8424× 1.0871× 0.9157×
callout-grammar-v2 ASCII whole 12 1,415/1,515 0.9649× 1.2791× 1.2342×
citation-affixes-grammar-v2 ASCII whole 12 587/611 1.3089× 1.5406× 2.0165×
citation-group-grammar-v2 ASCII whole 12 297/537 0.9747× 1.4876× 1.4501×
cite-author-grammar-v2 ASCII whole 12 180/288 1.0668× 1.5356× 1.6381×
cite-normal-grammar-v2 ASCII whole 12 204/288 1.2349× 1.5356× 1.8963×
cite-suppress-grammar-v2 ASCII whole 12 192/288 1.0509× 1.5356× 1.6138×
class-span-grammar-v2 ASCII whole 12 314/314 1.0187× 1.5185× 1.5468×
comment-empty-grammar-v2 ASCII whole 12 216/228 1.0233× 1.5461× 1.5821×
common-anchor-generation-grammar-v2 ASCII whole 12 348/348 1.0000× 2.0887× 2.0887×
common-angle-autolink-grammar-v2 ASCII whole 12 686/686 1.0000× 1.4462× 1.4462×
common-atx-1-grammar-v2 ASCII whole 12 1,371/1,371 1.0000× 2.0014× 2.0014×
common-atx-2-grammar-v2 ASCII whole 12 1,383/1,383 1.0000× 1.9996× 1.9996×
common-atx-3-grammar-v2 ASCII whole 12 1,395/1,395 1.0000× 1.9985× 1.9985×
common-atx-4-grammar-v2 ASCII whole 12 1,407/1,407 1.0000× 1.9963× 1.9963×
common-atx-5-grammar-v2 ASCII whole 12 1,419/1,419 1.0000× 1.9951× 1.9951×
common-atx-6-grammar-v2 ASCII whole 12 1,431/1,431 1.0000× 1.9934× 1.9934×
common-bullet-dash-grammar-v2 ASCII whole 12 2,357/2,357 1.0000× 1.5052× 1.5052×
common-bullet-plus-grammar-v2 ASCII whole 12 2,357/2,357 1.0000× 1.4995× 1.4995×
common-bullet-star-grammar-v2 ASCII whole 12 2,357/2,357 1.0000× 1.5012× 1.5012×
common-code-fence-backtick-grammar-v2 ASCII whole 12 405/405 1.0000× 1.2708× 1.2708×
common-code-fence-tilde-grammar-v2 ASCII whole 12 405/405 1.0000× 1.2709× 1.2709×
common-code-newline-grammar-v2 ASCII whole 12 314/314 1.0000× 1.5533× 1.5533×
common-code-span-grammar-v2 ASCII whole 12 434/434 1.0000× 1.5447× 1.5447×
common-emphasis-asterisk-grammar-v2 ASCII whole 12 491/491 1.0000× 1.4295× 1.4295×
common-emphasis-underscore-grammar-v2 ASCII whole 12 491/491 1.0000× 1.4298× 1.4298×
common-entities-grammar-v2 ASCII whole 12 770/770 1.0000× 1.0110× 1.0110×
common-eol-cr-grammar-v2 ASCII whole 12 1,493/1,493 1.0000× 1.4665× 1.4665×
common-eol-crlf-grammar-v2 ASCII whole 12 1,529/1,529 1.0000× 1.4955× 1.4955×
common-eol-lf-grammar-v2 ASCII whole 12 1,493/1,493 1.0000× 1.4669× 1.4669×
common-escapes-grammar-v2 ASCII whole 12 422/422 1.0000× 1.0292× 1.0292×
common-hardbreak-backslash-grammar-v2 ASCII whole 12 2,297/2,297 1.0000× 1.4516× 1.4516×
common-hardbreak-spaces-grammar-v2 ASCII whole 12 2,309/2,309 1.0000× 1.4467× 1.4467×
common-html-block-tag-grammar-v2 ASCII whole 12 204/204 1.0000× 1.3454× 1.3454×
common-html-cdata-grammar-v2 ASCII whole 12 192/192 1.0000× 1.3607× 1.3607×
common-html-comment-grammar-v2 ASCII whole 12 369/369 1.0000× 1.5259× 1.5259×
common-html-complete-tag-grammar-v2 ASCII whole 12 288/288 1.0000× 1.3225× 1.3225×
common-html-declaration-grammar-v2 ASCII whole 12 180/180 1.0000× 1.3507× 1.3507×
common-html-instruction-grammar-v2 ASCII whole 12 228/228 1.0000× 1.3290× 1.3290×
common-html-raw-grammar-v2 ASCII whole 12 276/276 1.0000× 1.2773× 1.2773×
common-image-grammar-v2 ASCII whole 12 1,694/1,694 1.0000× 1.4509× 1.4509×
common-indented-code-grammar-v2 ASCII whole 12 278/278 1.0000× 1.4849× 1.4849×
common-inline-comment-grammar-v2 ASCII whole 12 489/489 1.0000× 1.4843× 1.4843×
common-inline-html-grammar-v2 ASCII whole 12 1,755/1,755 1.0000× 1.4340× 1.4340×
common-link-direct-grammar-v2 ASCII whole 12 1,682/1,682 1.0000× 1.3859× 1.3859×
common-list-loose-grammar-v2 ASCII whole 12 2,417/2,417 1.0000× 1.4999× 1.4999×
common-list-parenthesis-grammar-v2 ASCII whole 12 2,357/2,357 1.0000× 1.5055× 1.5055×
common-paragraph-grammar-v2 ASCII whole 12 2,465/2,465 1.0000× 1.5658× 1.5658×
common-quote-grammar-v2 ASCII whole 12 2,405/2,405 1.0000× 1.4855× 1.4855×
common-reference-collapsed-grammar-v2 ASCII whole 12 2,511/2,511 1.0000× 0.7253× 0.7253×
common-reference-duplicate-grammar-v2 ASCII whole 12 3,425/3,425 1.0000× 0.6046× 0.6046×
common-reference-full-grammar-v2 ASCII whole 12 3,798/3,798 1.0000× 0.9156× 0.9156×
common-reference-shortcut-grammar-v2 ASCII whole 12 2,487/2,487 1.0000× 0.7281× 0.7281×
common-setext-one-grammar-v2 ASCII whole 12 2,333/2,333 1.0000× 2.0237× 2.0237×
common-setext-two-grammar-v2 ASCII whole 12 2,333/2,333 1.0000× 2.0236× 2.0236×
common-softbreak-grammar-v2 ASCII whole 12 2,285/2,285 1.0000× 1.4469× 1.4469×
common-tabs-grammar-v2 ASCII whole 12 1,395/1,395 1.0000× 1.4874× 1.4874×
common-thematic-dash-grammar-v2 ASCII whole 12 2,381/2,381 1.0000× 1.4949× 1.4949×
common-thematic-star-grammar-v2 ASCII whole 12 2,381/2,381 1.0000× 1.4930× 1.4930×
common-thematic-underscore-grammar-v2 ASCII whole 12 2,357/2,357 1.0000× 1.4935× 1.4935×
common-unresolved-reference-grammar-v2 ASCII whole 12 1,563/1,563 1.0000× 1.2484× 1.2484×
composition-directive-callout-grammar-v2 ASCII whole 12 1,503/1,491 0.9173× 1.5049× 1.3804×
composition-literal-islands-grammar-v2 ASCII whole 12 470/470 0.9988× 1.4910× 1.4893×
cross-anchor-grammar-v2 ASCII whole 12 273/285 0.8659× 1.5323× 1.3269×
cross-block-anchor-grammar-v2 ASCII whole 12 285/297 0.8667× 1.5255× 1.3221×
cross-embed-grammar-v2 ASCII whole 12 498/534 0.8407× 1.4594× 1.2270×
cross-embed-absent-grammar-v2 ASCII whole 12 249/261 0.8359× 1.5538× 1.2987×
cross-embed-empty-grammar-v2 ASCII whole 12 261/297 0.8335× 1.5215× 1.2682×
cross-embed-table-grammar-v2 ASCII whole 12 1,881/1,869 0.9678× 0.9638× 0.9328×
cross-link-grammar-v2 ASCII whole 12 486/522 0.8342× 1.4630× 1.2205×
cross-link-absent-grammar-v2 ASCII whole 12 237/249 0.8324× 1.5551× 1.2945×
cross-link-empty-grammar-v2 ASCII whole 12 249/285 0.8259× 1.5230× 1.2578×
cross-link-table-grammar-v2 ASCII whole 12 1,869/1,857 0.9712× 0.9658× 0.9379×
cross-local-grammar-v2 ASCII whole 12 228/240 0.8609× 1.5610× 1.3439×
cross-raw-label-grammar-v2 ASCII whole 12 312/336 0.8520× 1.5094× 1.2860×
decimal-list-grammar-v2 ASCII whole 12 2,364/2,364 1.0000× 1.5058× 1.5058×
default-list-grammar-v2 ASCII whole 12 2,357/2,357 1.0015× 1.5064× 1.5086×
definition-blocks-grammar-v2 ASCII whole 12 2,417/2,417 1.0363× 1.5289× 1.5845×
directive-attribute-only-grammar-v2 ASCII whole 12 326/374 0.9304× 1.5018× 1.3973×
directive-both-parts-grammar-v2 ASCII whole 12 1,637/1,661 1.0964× 1.3879× 1.5216×
directive-leaf-bare-grammar-v2 ASCII whole 12 72/108 0.4770× 1.4266× 0.6806×
directive-nested-grammar-v2 ASCII whole 12 1,673/1,625 0.8571× 1.5201× 1.3029×
directive-unbraced-grammar-v2 ASCII whole 12 1,491/1,419 0.9172× 1.4932× 1.3696×
directive-unicode-grammar-v2 ASCII whole 12 1,637/1,661 1.0532× 1.3691× 1.4420×
embed-dimensions-grammar-v2 ASCII local 12 378/378 1.0000× 1.5719× 1.5719×
empty-directive-grammar-v2 ASCII whole 12 204/228 1.0730× 1.5681× 1.6827×
enclosed-default-list-grammar-v2 ASCII whole 12 2,381/2,357 1.0020× 1.5055× 1.5085×
fallback-footnote-grammar-v2 ASCII whole 12 204/204 1.0000× 1.2774× 1.2774×
fallback-list-limit-grammar-v2 ASCII whole 12 1,455/1,455 1.0000× 1.4265× 1.4265×
fallback-strike-grammar-v2 ASCII whole 12 1,455/1,455 1.0000× 0.9787× 0.9787×
fallback-task-separator-grammar-v2 ASCII whole 12 503/503 1.0000× 1.1558× 1.1558×
footnote-nested-inline-grammar-v2 ASCII whole 12 419/419 1.1197× 1.4504× 1.6240×
footnote-retention-grammar-v2 ASCII whole 12 8,088/8,088 1.0000× 0.7594× 0.7594×
formula-backtick-grammar-v2 ASCII whole 12 429/405 1.0267× 1.5572× 1.5988×
formula-bracket-grammar-v2 ASCII whole 12 453/405 1.1439× 1.5572× 1.7812×
formula-bracket-block-grammar-v2 ASCII whole 12 357/357 1.0186× 1.3530× 1.3781×
formula-parenthesis-grammar-v2 ASCII whole 12 453/405 1.1479× 1.5572× 1.7874×
gfm-bare-autolink-grammar-v2 ASCII whole 12 695/695 1.0000× 1.0217× 1.0217×
gfm-footnote-grammar-v2 ASCII whole 12 3,593/3,593 1.0000× 0.8088× 0.8088×
gfm-footnote-blocks-grammar-v2 ASCII whole 12 4,627/4,627 1.0000× 0.8633× 0.8633×
gfm-footnote-cycle-grammar-v2 ASCII whole 12 4,614/4,614 1.0000× 0.7642× 0.7642×
gfm-pipe-ragged-grammar-v2 ASCII whole 12 815/815 1.0000× 0.8040× 0.8040×
gfm-pipe-table-grammar-v2 ASCII whole 12 2,714/2,714 1.0000× 0.9284× 0.9284×
gfm-strike-grammar-v2 ASCII whole 12 1,479/1,479 1.0000× 0.9897× 0.9897×
gfm-task-lower-grammar-v2 ASCII whole 12 1,383/1,383 1.0000× 1.0225× 1.0225×
gfm-task-open-grammar-v2 ASCII whole 12 1,383/1,383 1.0000× 1.0160× 1.0160×
gfm-task-upper-grammar-v2 ASCII whole 12 1,383/1,383 1.0000× 1.0182× 1.0182×
heading-explicit-id-grammar-v2 ASCII whole 12 2,405/2,513 0.8328× 1.4092× 1.1735×
heading-reference-grammar-v2 ASCII whole 12 2,140/3,198 1.1589× 0.6527× 0.7564×
image-dimensions-grammar-v2 ASCII local 12 378/378 1.0000× 1.5719× 1.5719×
inline-directive-grammar-v2 ASCII whole 12 1,491/1,515 1.0642× 1.3849× 1.4739×
inline-footnote-grammar-v2 ASCII whole 12 1,467/1,539 1.0324× 1.3822× 1.4270×
insertion-strong-grammar-v2 ASCII whole 12 1,479/1,479 0.9976× 1.4355× 1.4321×
leaf-comment-grammar-v2 ASCII whole 12 333/357 1.2172× 1.3530× 1.6468×
leaf-directive-grammar-v2 ASCII whole 12 1,383/1,395 1.0535× 1.2705× 1.3385×
leaf-fence-grammar-v2 ASCII whole 12 441/357 1.2759× 1.3530× 1.7262×
leaf-formula-grammar-v2 ASCII whole 12 333/357 1.0049× 1.3530× 1.3596×
loose-definition-grammar-v2 ASCII whole 12 1,596/1,620 1.0403× 1.5174× 1.5786×
mark-formatting-grammar-v2 ASCII whole 12 1,479/1,479 0.9996× 1.4355× 1.4349×
metadata-grammar-v2 ASCII local 12 1,469/1,469 1.0000× 1.4696× 1.4696×
metadata-literal-grammar-v2 ASCII local 12 1,754/1,754 1.0000× 1.4895× 1.4895×
metadata-types-grammar-v2 ASCII local 12 1,541/1,541 1.0000× 1.4929× 1.4929×
metadataempty-grammar-v2 ASCII local 12 1,541/1,541 1.0000× 1.4929× 1.4929×
mixed-definitions-grammar-v2 ASCII whole 12 2,795/3,047 0.9565× 1.5640× 1.4961×
named-container-grammar-v2 ASCII whole 12 2,417/2,441 0.9424× 1.5328× 1.4446×
opaque-comment-grammar-v2 ASCII whole 12 429/429 0.9373× 1.5494× 1.4523×
opaque-display-grammar-v2 ASCII whole 12 429/429 1.0082× 1.5494× 1.5622×
opaque-formula-grammar-v2 ASCII whole 12 405/405 1.0178× 1.5572× 1.5849×
record-span-grammar-v2 ASCII whole 12 1,649/1,661 1.0258× 1.3879× 1.4237×
roman-list-grammar-v2 ASCII whole 26 5,628/5,596 1.0165× 1.4202× 1.4437×
run-insertion-grammar-v2 ASCII whole 12 429/429 0.9986× 1.5432× 1.5410×
run-mark-grammar-v2 ASCII whole 12 429/429 0.9986× 1.5432× 1.5410×
run-strike-grammar-v2 ASCII whole 12 429/429 0.9971× 1.5432× 1.5386×
run-sub-grammar-v2 ASCII whole 12 278/278 0.9951× 1.4805× 1.4732×
run-super-grammar-v2 ASCII whole 12 278/278 1.0005× 1.4805× 1.4812×
script-empty-grammar-v2 ASCII whole 12 192/288 0.9384× 1.5452× 1.4500×
script-escaped-space-grammar-v2 ASCII whole 12 326/374 0.9546× 1.3567× 1.2951×
script-formatting-grammar-v2 ASCII whole 12 216/240 1.1337× 1.5015× 1.7023×
span-empty-grammar-v2 ASCII whole 12 314/314 1.0187× 1.5185× 1.5468×
span-nested-grammar-v2 ASCII whole 12 1,601/1,601 1.0347× 1.3756× 1.4233×
specimen-graph-grammar-v2 ASCII whole 12 3,581/3,593 0.9270× 0.8124× 0.7532×
specimen-groups-grammar-v2 ASCII whole 12 7,022/7,034 0.9654× 1.2062× 1.1645×
specimen-reset-grammar-v2 ASCII whole 12 3,629/4,699 0.8589× 1.0291× 0.8839×
task-heading-grammar-v2 ASCII whole 12 1,407/1,359 1.0103× 1.9810× 2.0014×
task-nested-grammar-v2 ASCII whole 12 2,477/2,477 1.0000× 1.0208× 1.0208×
task-unicode-grammar-v2 ASCII whole 12 1,407/1,383 1.0019× 1.0225× 1.0244×
task-value-grammar-v2 ASCII whole 12 2,429/2,429 1.0000× 1.0304× 1.0304×
topic-grammar-v2 ASCII whole 12 180/192 0.9302× 1.6706× 1.5540×
upper-list-grammar-v2 ASCII whole 26 5,631/5,596 1.0047× 1.4202× 1.4270×
upper-roman-list-grammar-v2 ASCII whole 26 5,628/5,596 1.0162× 1.4202× 1.4432×
alpha-list-grammar-v2 UTF-8 whole 26 12,810/12,827 1.0015× 1.2282× 1.2300×
anchor-grammar-v2 UTF-8 whole 12 4,121/4,229 0.9153× 0.9894× 0.9056×
anchor-last-empty-grammar-v2 UTF-8 whole 12 734/650 1.0058× 1.4364× 1.4447×
anonymous-container-grammar-v2 UTF-8 whole 12 5,534/5,462 1.0518× 1.2588× 1.3240×
attribute-autolink-grammar-v2 UTF-8 whole 12 838/910 0.7687× 1.4118× 1.0853×
attribute-bare-grammar-v2 UTF-8 whole 12 276/480 0.9515× 1.4633× 1.3923×
attribute-code-grammar-v2 UTF-8 whole 12 1,275/1,335 0.8363× 1.3334× 1.1152×
attribute-double-grammar-v2 UTF-8 whole 12 3,641/3,653 1.0558× 1.1915× 1.2579×
attribute-empty-grammar-v2 UTF-8 whole 12 288/300 0.9942× 1.5150× 1.5062×
attribute-escaped-value-grammar-v2 UTF-8 whole 12 722/734 0.9596× 1.3862× 1.3302×
attribute-fence-grammar-v2 UTF-8 whole 12 1,287/1,359 0.5406× 1.2544× 0.6781×
attribute-heading-grammar-v2 UTF-8 whole 12 3,581/3,569 1.0347× 1.5404× 1.5939×
attribute-image-grammar-v2 UTF-8 whole 12 3,793/3,865 0.9732× 1.2947× 1.2601×
attribute-link-grammar-v2 UTF-8 whole 12 3,781/3,853 0.9713× 1.2131× 1.1783×
attribute-newline-grammar-v2 UTF-8 whole 12 722/686 1.0802× 1.5124× 1.6336×
attribute-order-grammar-v2 UTF-8 whole 12 1,074/934 1.0002× 1.4284× 1.4286×
attribute-reference-grammar-v2 UTF-8 whole 12 5,978/5,930 1.0346× 0.6785× 0.7020×
attribute-single-grammar-v2 UTF-8 whole 12 3,641/3,653 1.0558× 1.1915× 1.2580×
attribute-unquoted-grammar-v2 UTF-8 whole 12 3,617/3,653 1.0548× 1.1915× 1.2568×
block-id-container-list-grammar-v2 UTF-8 whole 12 4,121/4,217 0.9191× 0.9995× 0.9187×
block-id-container-table-grammar-v2 UTF-8 whole 12 1,776/1,872 0.8718× 0.6865× 0.5985×
block-id-escaped-grammar-v2 UTF-8 whole 12 204/204 1.0000× 1.4414× 1.4414×
block-id-list-grammar-v2 UTF-8 whole 12 5,255/5,423 0.8333× 0.8610× 0.7175×
block-id-paragraph-grammar-v2 UTF-8 whole 12 4,085/4,205 0.8402× 0.9756× 0.8198×
callout-grammar-v2 UTF-8 whole 12 3,131/3,231 1.0484× 1.0460× 1.0966×
citation-affixes-grammar-v2 UTF-8 whole 12 1,287/1,311 1.2603× 1.3533× 1.7055×
citation-group-grammar-v2 UTF-8 whole 12 416/656 0.9737× 1.4860× 1.4469×
cite-author-grammar-v2 UTF-8 whole 12 216/324 1.0677× 1.5229× 1.6260×
cite-normal-grammar-v2 UTF-8 whole 12 240/324 1.2307× 1.5229× 1.8742×
cite-suppress-grammar-v2 UTF-8 whole 12 228/324 1.0476× 1.5229× 1.5954×
class-span-grammar-v2 UTF-8 whole 12 554/554 0.9878× 1.4114× 1.3942×
comment-empty-grammar-v2 UTF-8 whole 12 252/264 1.0233× 1.5282× 1.5638×
common-anchor-generation-grammar-v2 UTF-8 whole 12 420/420 1.0000× 2.0948× 2.0948×
common-angle-autolink-grammar-v2 UTF-8 whole 12 722/722 1.0000× 1.4443× 1.4443×
common-atx-1-grammar-v2 UTF-8 whole 12 3,087/3,087 1.0000× 1.9089× 1.9089×
common-atx-2-grammar-v2 UTF-8 whole 12 3,099/3,099 1.0000× 1.9079× 1.9079×
common-atx-3-grammar-v2 UTF-8 whole 12 3,111/3,111 1.0000× 1.9067× 1.9067×
common-atx-4-grammar-v2 UTF-8 whole 12 3,123/3,123 1.0000× 1.9054× 1.9054×
common-atx-5-grammar-v2 UTF-8 whole 12 3,135/3,135 1.0000× 1.9044× 1.9044×
common-atx-6-grammar-v2 UTF-8 whole 12 3,147/3,147 1.0000× 1.9031× 1.9031×
common-bullet-dash-grammar-v2 UTF-8 whole 12 5,462/5,462 1.0000× 1.2639× 1.2639×
common-bullet-plus-grammar-v2 UTF-8 whole 12 5,462/5,462 1.0000× 1.2584× 1.2584×
common-bullet-star-grammar-v2 UTF-8 whole 12 5,462/5,462 1.0000× 1.2602× 1.2602×
common-code-fence-backtick-grammar-v2 UTF-8 whole 12 829/829 1.0000× 1.2065× 1.2065×
common-code-fence-tilde-grammar-v2 UTF-8 whole 12 829/829 1.0000× 1.2067× 1.2067×
common-code-newline-grammar-v2 UTF-8 whole 12 590/590 1.0000× 1.4560× 1.4560×
common-code-span-grammar-v2 UTF-8 whole 12 710/710 1.0000× 1.4821× 1.4821×
common-emphasis-asterisk-grammar-v2 UTF-8 whole 12 850/850 1.0000× 1.3856× 1.3856×
common-emphasis-underscore-grammar-v2 UTF-8 whole 12 850/850 1.0000× 1.3858× 1.3858×
common-entities-grammar-v2 UTF-8 whole 12 1,046/1,046 1.0000× 0.9917× 0.9917×
common-eol-cr-grammar-v2 UTF-8 whole 12 3,449/3,449 1.0000× 1.2482× 1.2482×
common-eol-crlf-grammar-v2 UTF-8 whole 12 3,485/3,485 1.0000× 1.2734× 1.2734×
common-eol-lf-grammar-v2 UTF-8 whole 12 3,449/3,449 1.0000× 1.2483× 1.2483×
common-escapes-grammar-v2 UTF-8 whole 12 698/698 1.0000× 1.0342× 1.0342×
common-hardbreak-backslash-grammar-v2 UTF-8 whole 12 5,402/5,402 1.0000× 1.2014× 1.2014×
common-hardbreak-spaces-grammar-v2 UTF-8 whole 12 5,414/5,414 1.0000× 1.1956× 1.1956×
common-html-block-tag-grammar-v2 UTF-8 whole 12 240/240 1.0000× 1.3387× 1.3387×
common-html-cdata-grammar-v2 UTF-8 whole 12 228/228 1.0000× 1.3422× 1.3422×
common-html-comment-grammar-v2 UTF-8 whole 12 793/793 1.0000× 1.4703× 1.4703×
common-html-complete-tag-grammar-v2 UTF-8 whole 12 324/324 1.0000× 1.3030× 1.3030×
common-html-declaration-grammar-v2 UTF-8 whole 12 216/216 1.0000× 1.3344× 1.3344×
common-html-instruction-grammar-v2 UTF-8 whole 12 264/264 1.0000× 1.3195× 1.3195×
common-html-raw-grammar-v2 UTF-8 whole 12 312/312 1.0000× 1.2725× 1.2725×
common-image-grammar-v2 UTF-8 whole 12 3,733/3,733 1.0000× 1.2815× 1.2815×
common-indented-code-grammar-v2 UTF-8 whole 12 554/554 1.0000× 1.3815× 1.3815×
common-inline-comment-grammar-v2 UTF-8 whole 12 913/913 1.0000× 1.4172× 1.4172×
common-inline-html-grammar-v2 UTF-8 whole 12 3,507/3,507 1.0000× 1.2259× 1.2259×
common-link-direct-grammar-v2 UTF-8 whole 12 3,721/3,721 1.0000× 1.1939× 1.1939×
common-list-loose-grammar-v2 UTF-8 whole 12 5,522/5,522 1.0000× 1.2710× 1.2710×
common-list-parenthesis-grammar-v2 UTF-8 whole 12 5,462/5,462 1.0000× 1.2587× 1.2587×
common-paragraph-grammar-v2 UTF-8 whole 12 5,570/5,570 1.0000× 1.2960× 1.2960×
common-quote-grammar-v2 UTF-8 whole 12 5,546/5,546 1.0000× 1.2521× 1.2521×
common-reference-collapsed-grammar-v2 UTF-8 whole 12 2,900/2,900 1.0000× 0.5369× 0.5369×
common-reference-duplicate-grammar-v2 UTF-8 whole 12 3,847/3,847 1.0000× 0.4418× 0.4418×
common-reference-full-grammar-v2 UTF-8 whole 12 5,903/5,903 1.0000× 0.7015× 0.7015×
common-reference-shortcut-grammar-v2 UTF-8 whole 12 2,876/2,876 1.0000× 0.5382× 0.5382×
common-setext-one-grammar-v2 UTF-8 whole 12 5,438/5,438 1.0000× 1.8547× 1.8547×
common-setext-two-grammar-v2 UTF-8 whole 12 5,438/5,438 1.0000× 1.8547× 1.8547×
common-softbreak-grammar-v2 UTF-8 whole 12 5,390/5,390 1.0000× 1.1978× 1.1978×
common-tabs-grammar-v2 UTF-8 whole 12 3,147/3,147 1.0000× 1.2843× 1.2843×
common-thematic-dash-grammar-v2 UTF-8 whole 12 5,486/5,486 1.0000× 1.2490× 1.2490×
common-thematic-star-grammar-v2 UTF-8 whole 12 5,486/5,486 1.0000× 1.2474× 1.2474×
common-thematic-underscore-grammar-v2 UTF-8 whole 12 5,462/5,462 1.0000× 1.2474× 1.2474×
common-unresolved-reference-grammar-v2 UTF-8 whole 12 3,279/3,279 1.0000× 1.0963× 1.0963×
composition-directive-callout-grammar-v2 UTF-8 whole 12 3,255/3,243 0.9218× 1.2989× 1.1973×
composition-literal-islands-grammar-v2 UTF-8 whole 12 746/746 0.9989× 1.4383× 1.4368×
cross-anchor-grammar-v2 UTF-8 whole 12 407/419 0.8810× 1.4606× 1.2868×
cross-block-anchor-grammar-v2 UTF-8 whole 12 419/431 0.8837× 1.4572× 1.2878×
cross-embed-grammar-v2 UTF-8 whole 12 1,005/1,041 0.8787× 1.3566× 1.1921×
cross-embed-absent-grammar-v2 UTF-8 whole 12 332/344 0.8409× 1.5152× 1.2741×
cross-embed-empty-grammar-v2 UTF-8 whole 12 344/380 0.8384× 1.4836× 1.2438×
cross-embed-table-grammar-v2 UTF-8 whole 12 4,104/4,092 0.9888× 0.9545× 0.9438×
cross-link-grammar-v2 UTF-8 whole 12 993/1,029 0.8725× 1.3579× 1.1848×
cross-link-absent-grammar-v2 UTF-8 whole 12 320/332 0.8368× 1.5170× 1.2694×
cross-link-empty-grammar-v2 UTF-8 whole 12 332/368 0.8311× 1.4878× 1.2366×
cross-link-table-grammar-v2 UTF-8 whole 12 4,092/4,080 0.9957× 0.9522× 0.9482×
cross-local-grammar-v2 UTF-8 whole 12 279/291 0.8729× 1.5295× 1.3352×
cross-raw-label-grammar-v2 UTF-8 whole 12 399/423 0.8669× 1.4725× 1.2765×
decimal-list-grammar-v2 UTF-8 whole 12 5,469/5,469 1.0000× 1.2591× 1.2591×
default-list-grammar-v2 UTF-8 whole 12 5,462/5,462 1.0015× 1.2594× 1.2613×
definition-blocks-grammar-v2 UTF-8 whole 12 5,558/5,558 1.0390× 1.2905× 1.3409×
directive-attribute-only-grammar-v2 UTF-8 whole 12 602/650 0.9093× 1.4364× 1.3060×
directive-both-parts-grammar-v2 UTF-8 whole 12 3,629/3,653 1.1353× 1.1915× 1.3527×
directive-leaf-bare-grammar-v2 UTF-8 whole 12 108/144 0.4856× 1.4014× 0.6806×
directive-nested-grammar-v2 UTF-8 whole 12 3,665/3,617 0.8700× 1.3224× 1.1504×
directive-unbraced-grammar-v2 UTF-8 whole 12 3,243/3,171 0.9222× 1.2833× 1.1835×
directive-unicode-grammar-v2 UTF-8 whole 12 3,593/3,617 1.0485× 1.1860× 1.2435×
embed-dimensions-grammar-v2 UTF-8 local 12 885/885 1.0000× 1.4247× 1.4247×
empty-directive-grammar-v2 UTF-8 whole 12 240/264 1.0563× 1.5459× 1.6330×
enclosed-default-list-grammar-v2 UTF-8 whole 12 5,486/5,462 1.0020× 1.2587× 1.2612×
fallback-footnote-grammar-v2 UTF-8 whole 12 240/240 1.0000× 1.1781× 1.1781×
fallback-list-limit-grammar-v2 UTF-8 whole 12 3,171/3,171 1.0000× 1.2257× 1.2257×
fallback-strike-grammar-v2 UTF-8 whole 12 3,171/3,171 1.0000× 0.9585× 0.9585×
fallback-task-separator-grammar-v2 UTF-8 whole 12 862/862 1.0000× 1.1363× 1.1363×
footnote-nested-inline-grammar-v2 UTF-8 whole 12 778/778 1.1057× 1.4025× 1.5508×
footnote-retention-grammar-v2 UTF-8 whole 12 11,825/11,825 1.0000× 0.8103× 0.8103×
formula-backtick-grammar-v2 UTF-8 whole 12 853/829 1.0150× 1.3996× 1.4206×
formula-bracket-grammar-v2 UTF-8 whole 12 877/829 1.2900× 1.3996× 1.8055×
formula-bracket-block-grammar-v2 UTF-8 whole 12 781/781 1.0181× 1.2312× 1.2534×
formula-parenthesis-grammar-v2 UTF-8 whole 12 877/829 1.2935× 1.3996× 1.8103×
gfm-bare-autolink-grammar-v2 UTF-8 whole 12 971/971 1.0000× 1.0346× 1.0346×
gfm-footnote-grammar-v2 UTF-8 whole 12 5,377/5,377 1.0000× 0.8614× 0.8614×
gfm-footnote-blocks-grammar-v2 UTF-8 whole 12 7,800/7,800 1.0000× 0.8953× 0.8953×
gfm-footnote-cycle-grammar-v2 UTF-8 whole 12 6,432/6,432 1.0000× 0.8304× 0.8304×
gfm-pipe-ragged-grammar-v2 UTF-8 whole 12 1,174/1,174 1.0000× 0.8108× 0.8108×
gfm-pipe-table-grammar-v2 UTF-8 whole 12 5,938/5,938 1.0000× 0.9240× 0.9240×
gfm-strike-grammar-v2 UTF-8 whole 12 3,195/3,195 1.0000× 0.9755× 0.9755×
gfm-task-lower-grammar-v2 UTF-8 whole 12 3,099/3,099 1.0000× 1.0060× 1.0060×
gfm-task-open-grammar-v2 UTF-8 whole 12 3,099/3,099 1.0000× 0.9944× 0.9944×
gfm-task-upper-grammar-v2 UTF-8 whole 12 3,099/3,099 1.0000× 0.9992× 0.9992×
heading-explicit-id-grammar-v2 UTF-8 whole 12 4,154/4,262 0.8638× 1.2271× 1.0599×
heading-reference-grammar-v2 UTF-8 whole 12 2,206/3,297 1.1087× 0.4904× 0.5437×
image-dimensions-grammar-v2 UTF-8 local 12 885/885 1.0000× 1.4247× 1.4247×
inline-directive-grammar-v2 UTF-8 whole 12 3,243/3,267 1.0729× 1.2073× 1.2954×
inline-footnote-grammar-v2 UTF-8 whole 12 3,183/3,255 1.0266× 1.2084× 1.2405×
insertion-strong-grammar-v2 UTF-8 whole 12 3,195/3,195 0.9976× 1.2317× 1.2287×
leaf-comment-grammar-v2 UTF-8 whole 12 757/781 1.2122× 1.2312× 1.4924×
leaf-directive-grammar-v2 UTF-8 whole 12 3,135/3,147 1.1517× 1.0306× 1.1869×
leaf-fence-grammar-v2 UTF-8 whole 12 865/781 1.2979× 1.2312× 1.5979×
leaf-formula-grammar-v2 UTF-8 whole 12 757/781 1.0052× 1.2312× 1.2375×
loose-definition-grammar-v2 UTF-8 whole 12 3,736/3,760 1.0650× 1.2661× 1.3484×
mark-formatting-grammar-v2 UTF-8 whole 12 3,195/3,195 0.9996× 1.2317× 1.2312×
metadata-grammar-v2 UTF-8 local 12 3,425/3,425 1.0000× 1.2544× 1.2544×
metadata-literal-grammar-v2 UTF-8 local 12 4,134/4,134 1.0000× 1.2628× 1.2628×
metadata-types-grammar-v2 UTF-8 local 12 3,533/3,533 1.0000× 1.2988× 1.2988×
metadataempty-grammar-v2 UTF-8 local 12 3,533/3,533 1.0000× 1.2988× 1.2988×
mixed-definitions-grammar-v2 UTF-8 whole 12 6,494/6,746 0.9607× 1.3638× 1.3102×
named-container-grammar-v2 UTF-8 whole 12 5,558/5,582 0.9453× 1.2886× 1.2181×
opaque-comment-grammar-v2 UTF-8 whole 12 853/853 0.9141× 1.3984× 1.2783×
opaque-display-grammar-v2 UTF-8 whole 12 853/853 0.9979× 1.3984× 1.3955×
opaque-formula-grammar-v2 UTF-8 whole 12 829/829 1.0074× 1.3996× 1.4100×
record-span-grammar-v2 UTF-8 whole 12 3,641/3,653 1.0558× 1.1915× 1.2579×
roman-list-grammar-v2 UTF-8 whole 26 12,859/12,827 1.0168× 1.2282× 1.2488×
run-insertion-grammar-v2 UTF-8 whole 12 853/853 0.9986× 1.3870× 1.3851×
run-mark-grammar-v2 UTF-8 whole 12 853/853 0.9986× 1.3870× 1.3851×
run-strike-grammar-v2 UTF-8 whole 12 853/853 0.9971× 1.3870× 1.3829×
run-sub-grammar-v2 UTF-8 whole 12 518/518 0.9886× 1.4420× 1.4255×
run-super-grammar-v2 UTF-8 whole 12 518/518 0.9938× 1.4420× 1.4331×
script-empty-grammar-v2 UTF-8 whole 12 228/324 0.9378× 1.5287× 1.4336×
script-escaped-space-grammar-v2 UTF-8 whole 12 602/650 0.9514× 1.3080× 1.2445×
script-formatting-grammar-v2 UTF-8 whole 12 252/276 1.1325× 1.4771× 1.6728×
span-empty-grammar-v2 UTF-8 whole 12 554/554 0.9878× 1.4114× 1.3942×
span-nested-grammar-v2 UTF-8 whole 12 3,557/3,557 1.0662× 1.1905× 1.2692×
specimen-graph-grammar-v2 UTF-8 whole 12 5,365/5,377 0.9260× 0.8626× 0.7988×
specimen-groups-grammar-v2 UTF-8 whole 12 11,942/11,954 0.9610× 0.9696× 0.9318×
specimen-reset-grammar-v2 UTF-8 whole 12 5,411/6,514 0.8570× 0.8080× 0.6925×
task-heading-grammar-v2 UTF-8 whole 12 3,123/3,075 1.0109× 1.8971× 1.9177×
task-nested-grammar-v2 UTF-8 whole 12 5,582/5,582 1.0000× 0.9921× 0.9921×
task-unicode-grammar-v2 UTF-8 whole 12 3,123/3,099 1.0019× 1.0060× 1.0079×
task-value-grammar-v2 UTF-8 whole 12 5,534/5,534 1.0000× 1.0019× 1.0019×
topic-grammar-v2 UTF-8 whole 12 216/228 0.9318× 1.6080× 1.4982×
upper-list-grammar-v2 UTF-8 whole 26 12,862/12,827 1.0048× 1.2282× 1.2341×
upper-roman-list-grammar-v2 UTF-8 whole 26 12,859/12,827 1.0164× 1.2282× 1.2483×
Rejection of constructs only Core implements

No reference implements these constructs, so none is measured. B = Core on the common input; C = Core on its control, the same input with the rejected construct's trigger bytes replaced by letters of the same width.

Certificate Alphabet Rejects Units B/C (B − C) Ir per unit
fallback-attributes-grammar-v2 ASCII attribute-block 12 1.0277× 956
fallback-citation-grammar-v2 ASCII citation 12 1.0440× 243
fallback-comment-grammar-v2 ASCII percent-comment 12 1.1436× 4,421
fallback-cross-link-grammar-v2 ASCII cross-link 12 1.1440× 1,045
fallback-directive-grammar-v2 ASCII inline-directive 12 1.2426× 1,320
fallback-formula-grammar-v2 ASCII inline-formula 12 1.1693× 5,210
fallback-image-dimensions-grammar-v2 ASCII image-dimensions 12 1.0006× 6
fallback-insertion-grammar-v2 ASCII insertion 12 1.0612× 1,884
fallback-mark-grammar-v2 ASCII mark 12 1.0612× 1,884
fallback-noninitial-metadata-grammar-v2 ASCII metadata-envelope 12 no control —
fallback-script-grammar-v2 ASCII superscript 12 1.6452× 3,554
fallback-span-grammar-v2 ASCII bracketed-span 12 1.0292× 1,001
fallback-topic-grammar-v2 ASCII topic 12 1.0525× 287
fallback-attributes-grammar-v2 UTF-8 attribute-block 12 1.0603× 2,047
fallback-citation-grammar-v2 UTF-8 citation 12 1.0443× 240
fallback-comment-grammar-v2 UTF-8 percent-comment 12 1.2582× 7,708
fallback-cross-link-grammar-v2 UTF-8 cross-link 12 1.1666× 1,197
fallback-directive-grammar-v2 UTF-8 inline-directive 12 1.2460× 1,317
fallback-formula-grammar-v2 UTF-8 inline-formula 12 1.3133× 9,351
fallback-image-dimensions-grammar-v2 UTF-8 image-dimensions 12 1.0006× 6
fallback-insertion-grammar-v2 UTF-8 insertion 12 1.0639× 1,908
fallback-mark-grammar-v2 UTF-8 mark 12 1.0639× 1,908
fallback-noninitial-metadata-grammar-v2 UTF-8 metadata-envelope 12 no control —
fallback-script-grammar-v2 UTF-8 superscript 12 1.6859× 3,835
fallback-span-grammar-v2 UTF-8 bracketed-span 12 1.0617× 2,082
fallback-topic-grammar-v2 UTF-8 topic 12 1.0818× 437

Attribute grammar

Names and values Engine Lists Values Instructions (Ir) Data reads Data writes
ASCII markdown-core 4,000 12,724 6,716,785 1,659,686 897,971
ASCII lexbor 4,000 12,724 9,508,591 2,337,430 1,396,917
UTF-8 markdown-core 4,000 12,724 8,163,440 2,184,831 907,764
UTF-8 lexbor 4,000 12,724 10,782,778 2,576,993 1,416,849

Core / lexbor instructions on the same recovered attributes: ASCII 0.7064×, UTF-8 0.7571×.

Streams

Subjects: reparse (R), session (S) and binding (B), Ir per window of chunks and for the flush, R counting the two parse stages of each Document.parse; S runs every stream R runs at every step (those of at most 1,024 chunks, and the shapes 6.2 leaves to 6.3 at their smallest size) and every longer stream of a shape the language keeps local, and B the streams of at most 1,024 chunks. 768 workloads · baa4015502b54854.

Stream regressions (+2% per workload and subject in p95, max and total, against 30dc7361dd07): 754/768 passed, 14 exceeded.

Documents Family Size Workloads Windows R p50 R p95 R max R total S p50 S p95 S max S total S / R total Highest step S / R One-shot total
Grammar corpus tokens 384 161,409 252,363 770,175 1,523,229 48,721,226,662 33,467 121,364 505,296 5,296,378,000 0.206× 2.782× 139,991,554
Grammar corpus scalars 384 279,626 199,528 605,783 1,522,160 69,850,253,067 21,265 36,197 218,632 1,316,231,894 0.297× 2.238× 139,991,554

Bindings' C side (B): the MCB3 encode of each document an S window publishes and the release of its message, without building the platform values; no gate.

Documents Family Size B p50 B p95 B max B total B / S total
Grammar corpus tokens 63,539 167,947 296,344 8,456,467,236 1.597×
Grammar corpus scalars 18,245 59,066 122,298 1,450,156,098 1.102×

Edits: Ir per window of edits, R counting the two parse stages of the Document.parse of a window's last edit and S the session's edits of the window, every one; one window per edit except at a shape's larger sizes; no gate. 384 edit workloads.

Documents Family Size Scripts Edits Windows R p50 R p95 R max R total S p50 S p95 S max S total
Grammar corpus typing 384 24,576 24,576 382,478 834,075 1,525,371 8,822,243,311 113,118 297,373 732,545 3,080,483,997
Grammar corpus lines 1,152 6,066 6,066 389,611 864,326 1,525,894 2,237,578,269 42,398 573,078 1,778,963 765,005,904
Grammar corpus markers 384 6,912 6,912 368,328 813,033 1,531,247 2,408,240,108 60,624 566,444 1,693,585 925,254,409
Grammar corpus ranges 384 3,840 3,840 826,988 1,395,056 2,219,609 3,438,651,287 112,914 944,942 1,854,885 700,720,753
Grammar corpus far 384 24,576 24,576 385,242 861,296 1,525,666 9,085,102,649 37,146 704,779 1,774,907 3,549,781,226
Grammar corpus random 384 23,926 23,926 57,541 518,857 1,524,228 3,114,311,627 35,213 220,402 1,455,908 1,509,261,564

Completion latency

Callgrind Ir per completion request: the didChange of one edit and the completion at the end of its text. At most 8 requests a script, at evenly spaced edits. The budget is the plan's proposed p95 wall time on a device, not compared with these counts; no gate.

Documents Budget p95 Workloads Scripts Skipped Requests p50 p95 Max Edit p95 Completion p95
Grammar corpus 1 ms 384 3,072 36 21,150 108,652 682,140 3,631,053 526,857 233,973
Shapes, 16 KiB 1 ms 34 402 1 2,906 151,183 6,336,855 27,606,532 4,006,797 2,864,775
Shapes, 64 KiB 10 ms 40 184 0 1,376 186,274 22,748,859 48,160,263 12,150,844 11,144,213
Shapes, 256 KiB 10 ms 34 136 0 1,088 200,981 61,836,225 178,767,092 4,111,042 42,695,216

Headings prepare runs its loop only when a heading waits, a tree of one
leaf erases in that leaf, and an erase of nothing returns inline.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
An insertion that joins the front at the end appends in place, a cut of
the last leaf's tail leaves the leaf and lowers only the counts above it,
a one-leaf tree inserts in its leaf, a label table looks a label up once,
and headings in order are read from their stretch of storage.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
The first leaf of a tree starts at eight entries and doubles to a full
leaf while it is the only one, taken anew and copied so small blocks come
from the allocator's caches. Definition tables search the leaf of the
front's last entry directly, the uses flush walks runs, and withdraw skips
tables whose front ends before the root.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
A first leaf that grows from eight entries costs every small document
its growth in the parse stage; the leaf is taken at its full size, as
before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
Line places, the dropped facts and a leap's places step through the line
records leaf by leaf, forward with markdown_core_tree_run and back with
the new markdown_core_tree_back, instead of finding each line by rank.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AvzArumvqU8RFP8mWYPXpj
@DongyuZhao
DongyuZhao merged commit dab88b8 into main Nov 15, 2026
99 of 102 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants