[jankapunkt]

This is our ongoing development toward the upcoming 3.1.0 release.

Merged PRs

• refactor: remove all external npm dependencies #489
• docs: improvement of architecture documentation #483
• docs: fix badges and links #487
• refactor: convert ES5 syntax to ES2015+ #488
• feat: dual-mode JQuery/native DOM backend #493
• ♻️ (blaze/attrs): optimize attribute updates by caching last values #478
• Fix #314: Use console.error for reporting template exceptions #495
• docs: migrate to vitepress #494
• Tests: jquery + native DOM backends #497
• fix: prevent stale re-renders of #each items during sequence update (#468) #501
• fix(blaze): match delegated event selectors within delegation root #513

Resolved Issues

• implement double testing (w jquery + w/o jquery) → Tests: jquery + native DOM backends #497
• #each re-rendered with stale data context #468 → fix: prevent stale re-renders of #each items during sequence update (#468) #501

Release

• 3.1.0-alpha
• 3.1.0-beta
• 3.1.0-rc
• 3.1.0

Summary by CodeRabbit

• New Features

  • Added JavaScript URL security controls via new Blaze security APIs
  • Enabled URL protocol caching for improved performance
  • Optimized attribute updates by tracking and reducing redundant handler calls

• Improvements

  • Made jQuery optional—Blaze now operates without it when unavailable
  • Removed lodash library dependencies in favor of native JavaScript equivalents
  • Expanded architecture documentation with comprehensive overview diagrams and terminology

• Tests

  • Added performance benchmarks for attribute updates and URL protocol detection

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR modernizes the Blaze templating system by removing all Lodash dependencies across multiple packages, refactoring code to use modern JavaScript constructs (const/let, arrow functions, template literals, includes()), making the DOM backend jQuery-optional with native fallbacks, adding JavaScript URL security controls with caching, and updating documentation. The changes span approximately 40 files.

Changes

Cohort / File(s)	Summary
Documentation and Architecture OVERVIEW.md, README.md, images/architecture/architecture_packages.graphml	Expanded OVERVIEW with detailed architecture narrative and diagrams; updated README badges/links to GitHub Actions; regenerated GraphML diagram with updated styling.
Lodash Dependency Removal - Package Declarations packages/blaze/package.js, packages/caching-html-compiler/package.js, packages/templating-runtime/package.js, packages/templating-tools/package.js, packages/spacebars-compiler/package.js	Removed NPM dependency declarations for lodash.has, lodash.isEmpty, and uglify-js across multiple packages.
Lodash Dependency Removal - Shrinkwrap Files packages/blaze/.npm/package/npm-shrinkwrap.json, packages/caching-html-compiler/.npm/package/npm-shrinkwrap.json, packages/templating-runtime/.npm/package/npm-shrinkwrap.json, packages/templating-tools/.npm/package/npm-shrinkwrap.json, packages/spacebars-compiler/.npm/package/npm-shrinkwrap.json, packages/.../README, packages/.../.gitignore	Removed shrinkwrap lockfiles and generated package documentation after lodash/uglify-js removal.
Blaze Core - URL Security & Attributes packages/blaze/attrs.js	Added JavaScript URL security blocking (new APIs: _allowJavascriptUrls(), _javascriptUrlsAllowed(), _clearProtocolCache()); replaced lodash.has with local hasOwn; implemented URL protocol caching with bounded Map; added tracking of lastValues per attribute to avoid redundant updates.
Blaze Core - DOM Backend packages/blaze/dombackend.js	Major refactoring to make jQuery optional: added native DOM fallback paths for parseHTML, querySelectorAll, and delegated event handling; native event delegation uses closest() for target computation; TeardownCallback converted to ES class; conditional jQuery registration for special events.
Blaze Core - Modernization packages/blaze/builtins.js, packages/blaze/events.js, packages/blaze/exceptions.js, packages/blaze/lookup.js, packages/blaze/materializer.js, packages/blaze/preamble.js, packages/blaze/template.js, packages/blaze/view.js, packages/blaze/domrange.js, packages/blaze/utils.js	Replaced lodash utilities with local hasOwn/isObject helpers; converted var→const/let, function→arrow functions, string concatenation→template literals; refactored HelperMap to ES6 class; added new utils.js with hasOwn and isObject exports.
HTML/Spacebars Tools - Core Utilities packages/blaze-tools/tojs.js, packages/blaze-tools/token_tests.js, packages/blaze-tools/tokens.js	Modernized token parsing code: replaced var→const/let, function→arrow; converted unicode/identifier regex constructions to template literals; updated escape-sequence handling with padStart(4,'0').
HTML Tools - Parsing & Tokenization packages/html-tools/charref.js, packages/html-tools/charref_tests.js, packages/html-tools/parse.js, packages/html-tools/parse_tests.js, packages/html-tools/scanner.js, packages/html-tools/templatetag.js, packages/html-tools/tokenize.js, packages/html-tools/tokenize_tests.js, packages/html-tools/utils.js	Refactored all files: replaced var→const/let, function→arrow; converted string concatenation→template literals; replaced indexOf() checks with includes(); updated error messages.
HTMLjs - DOM Building packages/htmljs/html.js, packages/htmljs/htmljs_test.js, packages/htmljs/visitors.js	Modernized DOM construction code: replaced var→const/let, function→arrow; converted error messages and HTML string building to template literals; replaced .apply() with spread syntax; optimized string checks with includes()/startsWith().
Spacebars Compiler - Main Compilation packages/spacebars-compiler/codegen.js, packages/spacebars-compiler/compile_tests.js, packages/spacebars-compiler/compiler.js	Replaced var→const/let; converted function→arrow; updated code-generation templates to template literals; removed uglify-based minification, replaced with syntax validation via Function() constructor; removed UglifyJSMinify conditional logic.
Spacebars Compiler - Utilities packages/spacebars-compiler/optimizer.js, packages/spacebars-compiler/package.js, packages/spacebars-compiler/react.js, packages/spacebars-compiler/spacebars_tests.js, packages/spacebars-compiler/templatetag.js, packages/spacebars-compiler/whitespace.js	Modernized all files: replaced var→const/let, function→arrow; converted string checks to includes(); updated template-tag position definitions; refactored whitespace handling with functional composition.
Spacebars Runtime packages/spacebars/spacebars-runtime.js, packages/spacebars/spacebars_tests.js	Replaced var→const/let, function→arrow; replaced Function.prototype.apply usage with spread operator; updated error messages to template literals.
Templating Runtime packages/templating-runtime/dynamic.js, packages/templating-runtime/dynamic_tests.js, packages/templating-runtime/templating.js	Removed lodash.has dependency, replaced with Object.prototype.hasOwnProperty.call(); modernized variable declarations and error messages; updated HMR integration with arrow functions and template literals.
Templating Tools packages/templating-tools/compile-tags-with-spacebars.js, packages/templating-tools/html-scanner.js, packages/templating-tools/html-scanner-tests.js	Removed lodash.isEmpty, replaced with Object.keys().length check; modernized variable declarations; updated closing-tag detection and error messages to template literals.
Observe Sequence packages/observe-sequence/observe_sequence.js, packages/observe-sequence/observe_sequence_tests.js	Replaced var→const/let, function→arrow; updated warning logging with spread syntax; refactored debug rendering to arrow functions and template literals; modernized cursor fetch mapping and error handling.
Spacebars Tests packages/spacebars-tests/old_templates.js, packages/spacebars-tests/old_templates_tests.js, packages/spacebars-tests/templating_tests.js	Replaced var with const/let throughout test suites and template factory functions; no logic changes.
Benchmarking tests/benchmark_ElementAttributesUpdater.html, tests/benchmark_getUrlProtocol.html	Added new benchmark pages measuring ElementAttributesUpdater (original vs. _lastValues cached) and getUrlProtocol (uncached vs. cached Map) with DOM result tables and console output.

Possibly related issues

• Remove jQuery dependency from Blaze — Native DOM rewrite of dombackend.js #490: The rewrite of packages/blaze/dombackend.js to add native DOM fallbacks and conditional jQuery usage directly addresses jQuery removal and native DOM backend implementation described in this issue.

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~50 minutes

Poem

🐰 From lodash fields we hop away,
With const and let, we modernize the day,
jQuery bows where native DOM takes flight,
Arrow functions dance in template light,
Security cached, protocols cached too—
The Blaze now shines in form anew! ✨

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 45.95% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'release 3.1.0' directly relates to the main objective of the pull request, which is development toward the upcoming 3.1.0 release. However, the title is quite generic and doesn't convey the scope of the major changes (ES5 to ES2015+ conversion, removal of external npm dependencies, jQuery/native DOM backend).
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch release-3.1.0

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 10

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (4)

packages/templating-tools/html-scanner-tests.js (1)

12-18: ⚠️ Potential issue | 🟠 Major

Fix instanceof guard precedence in checkError.

Line 16 is currently parsed as (!e) instanceof TemplatingTools.CompileError, so the condition always evaluates to false and unexpected exceptions are silently swallowed instead of being rethrown.

Proposed fix

-      if (! e instanceof TemplatingTools.CompileError) {
+      if (!(e instanceof TemplatingTools.CompileError)) {
         throw e;
       }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/templating-tools/html-scanner-tests.js` around lines 12 - 18, The
instanceof guard in the checkError function is using incorrect precedence:
change the condition from "! e instanceof TemplatingTools.CompileError" to test
the instance correctly (e.g. "!(e instanceof TemplatingTools.CompileError)" or
"if (e instanceof TemplatingTools.CompileError) { ... } else { throw e }") so
that unexpected exceptions thrown by f() are rethrown; update the checkError
function and its catch block to use the corrected instanceof check referencing
TemplatingTools.CompileError and the caught variable e.

packages/htmljs/visitors.js (1)

191-200: ⚠️ Potential issue | 🟠 Major

Filter inherited properties when iterating attributes.

The for...in loops at lines 191 and 304 iterate attribute dictionaries without filtering inherited enumerable properties. This allows properties from polluted or custom prototypes to flow through visitAttribute and serialize as HTML attributes. The inconsistency with _assign (which guards via _hasOwnProperty.call() at lines 21–27) leaves these paths unprotected.

Use Object.keys() instead of for...in:

• Line 191: for (const k of Object.keys(oldAttrs))
• Line 304: for (const k of Object.keys(attrs))

Also apply the same fix to flattenAttributes in html.js (around line 214) for consistency.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/htmljs/visitors.js` around lines 191 - 200, The for...in loops
iterate attribute maps and allow inherited enumerable properties to be
processed; change them to iterate only own keys using Object.keys to prevent
prototype-polluted attributes from reaching visitAttribute and being serialized:
in visitors.js replace the loop over oldAttrs (where oldValue is read and
visitAttribute is called and _assign is used) with a for (const k of
Object.keys(oldAttrs)) loop, and similarly replace the loop over attrs (the
second occurrence that calls visitAttribute) with for (const k of
Object.keys(attrs)); also make the analogous change in flattenAttributes
(html.js, function flattenAttributes) to iterate Object.keys(attrs) instead of
for...in so behavior matches the _assign/_hasOwnProperty protections already
used elsewhere.

packages/html-tools/charref.js (1)

2247-2259: ⚠️ Potential issue | 🟠 Major

Use prototype-safe maps for the entity matcher cache.

Lines 2250 and 2255 iterate plain-object lookup tables with for...in. If an enumerable property is added to Object.prototype before module initialization, it will be treated as an entity key and poison the generated matchers. Build these tables with null prototypes and iterate own keys only.

Proposed fix

-const getNamedEntityByFirstChar = {};
+const getNamedEntityByFirstChar = Object.create(null);
 (function () {
-  const namedEntitiesByFirstChar = {};
-  for (const ent in ENTITIES) {
+  const namedEntitiesByFirstChar = Object.create(null);
+  for (const ent of Object.keys(ENTITIES)) {
     const chr = ent.charAt(1);
     namedEntitiesByFirstChar[chr] = (namedEntitiesByFirstChar[chr] || []);
     namedEntitiesByFirstChar[chr].push(ent.slice(2));
   }
-  for (const chr in namedEntitiesByFirstChar) {
+  for (const chr of Object.keys(namedEntitiesByFirstChar)) {
     getNamedEntityByFirstChar[chr] = makeRegexMatcher(
       new RegExp('^&' + chr + '(?:' +
                  namedEntitiesByFirstChar[chr].join('|') + ')'));
   }
 })();

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/html-tools/charref.js` around lines 2247 - 2259, The cache objects
should be prototype-safe: create getNamedEntityByFirstChar and
namedEntitiesByFirstChar with null prototypes (use Object.create(null)) and
iterate only own keys (use Object.keys or for-of over Object.keys) instead of
for...in; specifically, build namedEntitiesByFirstChar with Object.create(null),
loop over Object.keys(ENTITIES) when collecting entries, then loop over
Object.keys(namedEntitiesByFirstChar) when creating matchers for
getNamedEntityByFirstChar; keep references to ENTITIES and makeRegexMatcher and
assign into getNamedEntityByFirstChar as before.

packages/spacebars-compiler/codegen.js (1)

302-328: ⚠️ Potential issue | 🔴 Critical

Potential null dereference on tagArgs.

The comment on line 306 states "tagArgs may be null", but line 307 calls tagArgs.forEach(...) directly without a null check. This will throw a TypeError if tagArgs is null or undefined.

🐛 Proposed fix to add null guard

   // tagArgs may be null
-  tagArgs.forEach((arg) => {
+  (tagArgs || []).forEach((arg) => {
     const argCode = this.codeGenArgValue(arg);

Alternatively, use optional chaining with nullish coalescing or an early return pattern.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/spacebars-compiler/codegen.js` around lines 302 - 328, In
codeGenMustacheArgs, tagArgs may be null but the function calls tagArgs.forEach
directly; add a null guard (e.g., early return when tagArgs is null/undefined or
iterate only when tagArgs exists) so the function doesn't throw a
TypeError—update codeGenMustacheArgs to check tagArgs before calling forEach and
preserve existing behavior (produce null/empty args/kwArgs and still append
Spacebars.kw when needed).

🧹 Nitpick comments (9)

README.md (1)

39-40: Consider hyphenating the compound adjective.

The phrase "Blaze related news" would be more grammatically correct as "Blaze-related news" when the compound adjective precedes the noun.

📝 Proposed grammar improvement

-definitely [join us on the Meteor forums](http://forums.meteor.com).
+definitely [join us on the Meteor forums](https://forums.meteor.com).

Also consider this optional grammar fix for line 39:

-Also, to stay up-to-date on all Blaze related news and the community you should
+Also, to stay up-to-date on all Blaze-related news and the community you should

Note: Also updated the forum link to use HTTPS for security.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@README.md` around lines 39 - 40, Change the phrase "Blaze related news" to
the compound adjective "Blaze-related news" in the README and update the forum
URL in the same sentence from "http://forums.meteor.com" to
"https://forums.meteor.com"; locate the sentence containing the link text "[join
us on the Meteor forums]" and replace the text and URL accordingly to maintain
correct grammar and a secure HTTPS link.

packages/observe-sequence/observe_sequence.js (6)

406-436: Implicit global: seqChangedToCursor lacks declaration keyword.

Same issue - this function should be declared with const.

♻️ Suggested fix

-seqChangedToCursor = function (lastSeqArray, cursor, callbacks) {
+const seqChangedToCursor = function (lastSeqArray, cursor, callbacks) {

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/observe-sequence/observe_sequence.js` around lines 406 - 436, The
function seqChangedToCursor is assigned without a declaration, creating an
implicit global; change its definition to use a proper declaration (e.g., add
"const" before seqChangedToCursor) so it becomes a local constant function,
leaving the internal logic (seqArray, observeHandle, callbacks) unchanged;
ensure no other references rely on a global and run lint/tests to confirm no
scope-related regressions.

---

270-273: Variable shadowing in callback: intentional but slightly obscure.

The destructured id in [id, pos] shadows the outer id parameter from addedBefore. While this works correctly (the iteration updates positions for all entries, not just the added one), consider renaming to entryId or key for clarity.

✨ Optional: Rename to avoid shadowing

-        Object.entries(posCur).forEach(([id, pos]) => {
+        Object.entries(posCur).forEach(([key, pos]) => {
           if (pos >= position)
-            posCur[id]++;
+            posCur[key]++;
         });

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/observe-sequence/observe_sequence.js` around lines 270 - 273, In the
addedBefore function where Object.entries(posCur).forEach(([id, pos]) => { ...
}) is used, rename the destructured variable `id` to something non-shadowing
(e.g., `entryId` or `key`) and update the body to use that new name (e.g.,
posCur[entryId]++) so it no longer shadows the outer `id` parameter; this keeps
the logic unchanged but improves clarity around the outer `id` and the callback
identifier.

---

373-404: Implicit global: seqChangedToArray lacks declaration keyword.

Same issue as seqChangedToEmpty - this function assignment creates an implicit global.

♻️ Suggested fix

-seqChangedToArray = function (lastSeqArray, array, callbacks) {
+const seqChangedToArray = function (lastSeqArray, array, callbacks) {

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/observe-sequence/observe_sequence.js` around lines 373 - 404,
Declare seqChangedToArray to avoid creating an implicit global by adding a
proper declaration (e.g., use "const seqChangedToArray = function(...)" or a
"function seqChangedToArray(...)" form) instead of the current bare assignment;
update the definition that produces seqArray (references: seqChangedToArray,
idsUsed, idStringify, Random.id, warn) so the function is locally declared and
does not leak to the global scope.

---

331-336: Same shadowing pattern.

The [id, pos] destructuring shadows the outer id parameter from removed.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/observe-sequence/observe_sequence.js` around lines 331 - 336, The
loop destructures ([id, pos]) which shadows the outer removed parameter named
id; rename the loop key to avoid shadowing (e.g., ([entryId, pos])) and update
the body to reference posCur[entryId] and compare pos against prevPosition
(computed from posCur[idStringify(id)]), so the outer id/removed stays intact;
references: removed parameter name id, posCur, idStringify, and the forEach
callback where ([id, pos]) appears.

---

369-371: Implicit global: seqChangedToEmpty is not declared with const/let.

While this line wasn't modified in this PR (no ~ marker), the function is assigned without a declaration keyword, creating an implicit global. This is inconsistent with the modernization applied elsewhere in this file.

♻️ Suggested fix

-seqChangedToEmpty = function (lastSeqArray, callbacks) {
+const seqChangedToEmpty = function (lastSeqArray, callbacks) {
   return [];
 };

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/observe-sequence/observe_sequence.js` around lines 369 - 371, The
function seqChangedToEmpty is currently assigned without a declaration, creating
an implicit global; update its definition to use a local declaration (e.g.,
prepend const or let) so it is properly scoped (replace the bare assignment to
seqChangedToEmpty with a declared binding for the function named
seqChangedToEmpty used alongside other locals in this module).

---

313-318: Same shadowing pattern as above.

The [id, elCurPosition] destructuring shadows the outer id parameter from movedBefore.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/observe-sequence/observe_sequence.js` around lines 313 - 318, The
inner callback is shadowing the outer parameter id from movedBefore; rename the
destructured key in the forEach from [id, elCurPosition] to a different
identifier (e.g., [entryId, elCurPosition]) and update all uses inside the
callback (posCur[entryId]-- / posCur[entryId]++). This preserves the outer
movedBefore id parameter and avoids shadowing in the Object.entries(...).forEach
callback.

packages/blaze-tools/tojs.js (1)

90-106: Redundant array initialization.

Line 95 re-initializes attrsArray = [] when it was already set to [] on line 93 (two lines before, inside the same if (attrs) block).

♻️ Remove redundant initialization

       attrsArray = [];
       if (HTML.isArray(attrs)) {
-        attrsArray = [];
         for (let i = 0; i < attrs.length; i++) {

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/blaze-tools/tojs.js` around lines 90 - 106, The code redundantly
sets attrsArray = [] twice inside the if (attrs) block; remove the second
initialization so attrsArray is only initialized once when attrs is truthy. Edit
the block surrounding attrsArray, attrs, HTML.isArray, hasToJS and
generateAttrsDictionary to delete the duplicate attrsArray = [] (the one
immediately after HTML.isArray(attrs) check) and leave the rest of the loop
logic unchanged.

packages/html-tools/parse.js (1)

48-65: Consider using let for consistency.

Line 51 uses var endTag while the rest of the file has been modernized to use const/let. While var works here due to hoisting across the try/catch block, using let with a declaration before the try block would be more consistent with the modernization effort.

♻️ Proposed consistency fix

     const posBefore = scanner.pos;
 
+    let endTag;
     try {
-      var endTag = getHTMLToken(scanner);
+      endTag = getHTMLToken(scanner);
     } catch (e) {
       // ignore errors from getTemplateTag
     }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/html-tools/parse.js` around lines 48 - 65, Replace the
function-scoped "var endTag" with a block-scoped declaration: declare "let
endTag" immediately before the try block and assign to it inside the try by
calling getHTMLToken(scanner); keep the existing catch behavior and subsequent
use of endTag (the if checking endTag.t/isEnd and scanner.fatal) unchanged so
behavior is identical but uses modern let scoping instead of var.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@OVERVIEW.md`:
- Around line 199-206: The unlabeled fenced code blocks flagged by MD040 should
each include a language identifier (use "text")—specifically add ```text before
the block that begins ".html file → templating-compiler …", the block that
begins "Template function called …", and the block that begins "Template
(Blueprint) …" so all three fenced blocks are labeled and markdownlint MD040 is
satisfied.
- Line 18: Replace the branch-pinned raw URL image link in the markdown (the
current
"https://raw.githubusercontent.com/meteor/blaze/architecture/images/architecture_packages.svg"
reference) with a repository-relative path to the asset (e.g. use the relative
Markdown image URL pointing at architecture/images/architecture_packages.svg) so
the docs render reliably across branches and forks; update the image markdown in
OVERVIEW.md accordingly.

In `@packages/blaze/attrs.js`:
- Around line 746-762: The current short-circuit uses shouldUpdate to skip
calling handler.update when the new value equals lastValues[k], which prevents
handlers that reconcile against the live DOM (e.g., for
value/checked/class/style) from restoring template-owned state after
user/external mutation; change the logic in the block that references
lastValues, handlers, handler and shouldUpdate so that handler.update(elem,
oldValue, value) is always called regardless of shouldUpdate, but only update
handler.value and lastValues[k] (and perform the value === null cleanup that
deletes handlers[k] and lastValues[k]) when shouldUpdate is true — keep the same
calls and symbols (handler.update, handler.value, lastValues, handlers, value,
oldValue, elem) and only move the update invocation out of the short-circuit
guard.

In `@packages/blaze/dombackend.js`:
- Around line 94-109: The undelegateEvents implementation currently removes all
wrappers for a given handler regardless of the passed type; update
undelegateEvents to only remove entries whose entry.eventType equals the
provided type: look up handlerMap via _delegateMap.get(elem), get entries =
handlerMap.get(handler), iterate and only call
elem.removeEventListener(entry.eventType, entry.wrapper) for entries where
entry.eventType === type (or matches the provided type string), remove those
matching entries from the entries collection, and only call
handlerMap.delete(handler) if no entries remain for that handler; reference
undelegateEvents, _delegateMap, handlerMap, entries, and
entry.eventType/entry.wrapper to locate the change.
- Around line 36-42: DOMBackend.parseHTML's native branch currently returns
template.content.childNodes which keeps <script> elements, whereas the jQuery
branch ($jq.parseHTML(...)) excludes scripts by default; update the native path
in DOMBackend.parseHTML to filter out SCRIPT nodes before returning so both
branches behave consistently. In practice, after creating template and
populating template.innerHTML, iterate over template.content.childNodes (or use
querySelectorAll) and return only nodes whose nodeName !== 'SCRIPT' (or nodeType
!== Node.ELEMENT_NODE || tagName !== 'SCRIPT') so script elements are excluded
from the returned array.

In `@packages/html-tools/scanner.js`:
- Around line 34-37: The truncation logic for pastInput is wrong because
substring(-CONTEXT_AMOUNT) treats negative indices as 0; in the block that sets
pastInput (using variables input, pos, CONTEXT_AMOUNT and the pastInput
variable), replace the use of substring(-CONTEXT_AMOUNT) with
slice(-CONTEXT_AMOUNT) so pastInput = `...${pastInput.slice(-CONTEXT_AMOUNT)}`
to correctly return the trailing N characters when pastInput.length >
CONTEXT_AMOUNT.

In `@packages/spacebars-compiler/compile_tests.js`:
- Around line 25-30: The postProcess function currently uses the regex
/\/\/.*$/mg which naively strips from // to end-of-line and can remove //
sequences inside string literals (e.g., "http://..."); replace that regex usage
in postProcess with a safe line-by-line comment stripper that scans each line
character-by-character, tracking quote state and escapes so it only removes //
when not inside single or double quotes (implement a helper like
stripLineCommentsPreservingStrings used from postProcess), then continue with
the existing whitespace collapse and paren-trimming steps; ensure the new code
references postProcess and removes the /\/\/.*$/mg usage.

In `@packages/spacebars-compiler/templatetag.js`:
- Around line 359-366: TemplateTag.parseCompleteTag is creating a local variable
scanner but then calls TemplateTag.parse(scannerOrString), which can create a
different Scanner and desynchronize cursor positions; change the call to use the
local scanner variable (i.e., call TemplateTag.parse(scanner)) so parsing,
parseFragment, validation and error position reporting all use the same Scanner
instance (update any other places in this function that pass scannerOrString
into internal parse functions to use scanner).

In `@packages/templating-runtime/templating.js`:
- Around line 161-164: The function signature for Template._migrateTemplate
currently declares an unused third parameter `migrate`; remove it and compute
migration intent locally instead so caller input isn't expected. Change the
signature to Template._migrateTemplate = function (templateName, newTemplate) {
and replace the reassignment line with a local const, e.g. const migrate =
Template.__pendingReplacement.includes(templateName); (or delete any unused
`migrate` variable entirely if it’s not referenced later).

In `@README.md`:
- Around line 5-6: The badge image tag lacks alternative text; update the
README.md badge <img> element (inside the <a
href="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml"> link)
to include a descriptive alt attribute (e.g., alt="GitHub Actions — blaze-tests
workflow status") so screen readers can convey the badge meaning; ensure the alt
text is concise and specific to the workflow name/status.

---

Outside diff comments:
In `@packages/html-tools/charref.js`:
- Around line 2247-2259: The cache objects should be prototype-safe: create
getNamedEntityByFirstChar and namedEntitiesByFirstChar with null prototypes (use
Object.create(null)) and iterate only own keys (use Object.keys or for-of over
Object.keys) instead of for...in; specifically, build namedEntitiesByFirstChar
with Object.create(null), loop over Object.keys(ENTITIES) when collecting
entries, then loop over Object.keys(namedEntitiesByFirstChar) when creating
matchers for getNamedEntityByFirstChar; keep references to ENTITIES and
makeRegexMatcher and assign into getNamedEntityByFirstChar as before.

In `@packages/htmljs/visitors.js`:
- Around line 191-200: The for...in loops iterate attribute maps and allow
inherited enumerable properties to be processed; change them to iterate only own
keys using Object.keys to prevent prototype-polluted attributes from reaching
visitAttribute and being serialized: in visitors.js replace the loop over
oldAttrs (where oldValue is read and visitAttribute is called and _assign is
used) with a for (const k of Object.keys(oldAttrs)) loop, and similarly replace
the loop over attrs (the second occurrence that calls visitAttribute) with for
(const k of Object.keys(attrs)); also make the analogous change in
flattenAttributes (html.js, function flattenAttributes) to iterate
Object.keys(attrs) instead of for...in so behavior matches the
_assign/_hasOwnProperty protections already used elsewhere.

In `@packages/spacebars-compiler/codegen.js`:
- Around line 302-328: In codeGenMustacheArgs, tagArgs may be null but the
function calls tagArgs.forEach directly; add a null guard (e.g., early return
when tagArgs is null/undefined or iterate only when tagArgs exists) so the
function doesn't throw a TypeError—update codeGenMustacheArgs to check tagArgs
before calling forEach and preserve existing behavior (produce null/empty
args/kwArgs and still append Spacebars.kw when needed).

In `@packages/templating-tools/html-scanner-tests.js`:
- Around line 12-18: The instanceof guard in the checkError function is using
incorrect precedence: change the condition from "! e instanceof
TemplatingTools.CompileError" to test the instance correctly (e.g. "!(e
instanceof TemplatingTools.CompileError)" or "if (e instanceof
TemplatingTools.CompileError) { ... } else { throw e }") so that unexpected
exceptions thrown by f() are rethrown; update the checkError function and its
catch block to use the corrected instanceof check referencing
TemplatingTools.CompileError and the caught variable e.

---

Nitpick comments:
In `@packages/blaze-tools/tojs.js`:
- Around line 90-106: The code redundantly sets attrsArray = [] twice inside the
if (attrs) block; remove the second initialization so attrsArray is only
initialized once when attrs is truthy. Edit the block surrounding attrsArray,
attrs, HTML.isArray, hasToJS and generateAttrsDictionary to delete the duplicate
attrsArray = [] (the one immediately after HTML.isArray(attrs) check) and leave
the rest of the loop logic unchanged.

In `@packages/html-tools/parse.js`:
- Around line 48-65: Replace the function-scoped "var endTag" with a
block-scoped declaration: declare "let endTag" immediately before the try block
and assign to it inside the try by calling getHTMLToken(scanner); keep the
existing catch behavior and subsequent use of endTag (the if checking
endTag.t/isEnd and scanner.fatal) unchanged so behavior is identical but uses
modern let scoping instead of var.

In `@packages/observe-sequence/observe_sequence.js`:
- Around line 406-436: The function seqChangedToCursor is assigned without a
declaration, creating an implicit global; change its definition to use a proper
declaration (e.g., add "const" before seqChangedToCursor) so it becomes a local
constant function, leaving the internal logic (seqArray, observeHandle,
callbacks) unchanged; ensure no other references rely on a global and run
lint/tests to confirm no scope-related regressions.
- Around line 270-273: In the addedBefore function where
Object.entries(posCur).forEach(([id, pos]) => { ... }) is used, rename the
destructured variable `id` to something non-shadowing (e.g., `entryId` or `key`)
and update the body to use that new name (e.g., posCur[entryId]++) so it no
longer shadows the outer `id` parameter; this keeps the logic unchanged but
improves clarity around the outer `id` and the callback identifier.
- Around line 373-404: Declare seqChangedToArray to avoid creating an implicit
global by adding a proper declaration (e.g., use "const seqChangedToArray =
function(...)" or a "function seqChangedToArray(...)" form) instead of the
current bare assignment; update the definition that produces seqArray
(references: seqChangedToArray, idsUsed, idStringify, Random.id, warn) so the
function is locally declared and does not leak to the global scope.
- Around line 331-336: The loop destructures ([id, pos]) which shadows the outer
removed parameter named id; rename the loop key to avoid shadowing (e.g.,
([entryId, pos])) and update the body to reference posCur[entryId] and compare
pos against prevPosition (computed from posCur[idStringify(id)]), so the outer
id/removed stays intact; references: removed parameter name id, posCur,
idStringify, and the forEach callback where ([id, pos]) appears.
- Around line 369-371: The function seqChangedToEmpty is currently assigned
without a declaration, creating an implicit global; update its definition to use
a local declaration (e.g., prepend const or let) so it is properly scoped
(replace the bare assignment to seqChangedToEmpty with a declared binding for
the function named seqChangedToEmpty used alongside other locals in this
module).
- Around line 313-318: The inner callback is shadowing the outer parameter id
from movedBefore; rename the destructured key in the forEach from [id,
elCurPosition] to a different identifier (e.g., [entryId, elCurPosition]) and
update all uses inside the callback (posCur[entryId]-- / posCur[entryId]++).
This preserves the outer movedBefore id parameter and avoids shadowing in the
Object.entries(...).forEach callback.

In `@README.md`:
- Around line 39-40: Change the phrase "Blaze related news" to the compound
adjective "Blaze-related news" in the README and update the forum URL in the
same sentence from "http://forums.meteor.com" to "https://forums.meteor.com";
locate the sentence containing the link text "[join us on the Meteor forums]"
and replace the text and URL accordingly to maintain correct grammar and a
secure HTTPS link.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: a4eaaee7-9299-4d96-ba0c-d4770b95bbaf

📥 Commits

Reviewing files that changed from the base of the PR and between da16df1 and 2cff5d2.

⛔ Files ignored due to path filters (7)

• images/architecture-static-html.png is excluded by !**/*.png
• images/architecture-templating-tools.png is excluded by !**/*.png
• images/architecture/architecture-static-html.svg is excluded by !**/*.svg
• images/architecture/architecture-templating-tools.svg is excluded by !**/*.svg
• images/architecture/architecture_packages.svg is excluded by !**/*.svg
• images/architecture_packages.png is excluded by !**/*.png
• images/architecture_packages.svg is excluded by !**/*.svg

📒 Files selected for processing (77)

• OVERVIEW.md
• README.md
• images/architecture/architecture-static-html.graphml
• images/architecture/architecture-templating-tools.graphml
• images/architecture/architecture_packages.graphml
• packages/blaze-tools/tojs.js
• packages/blaze-tools/token_tests.js
• packages/blaze-tools/tokens.js
• packages/blaze/.npm/package/.gitignore
• packages/blaze/.npm/package/README
• packages/blaze/.npm/package/npm-shrinkwrap.json
• packages/blaze/attrs.js
• packages/blaze/builtins.js
• packages/blaze/dombackend.js
• packages/blaze/domrange.js
• packages/blaze/events.js
• packages/blaze/exceptions.js
• packages/blaze/lookup.js
• packages/blaze/materializer.js
• packages/blaze/package.js
• packages/blaze/preamble.js
• packages/blaze/template.js
• packages/blaze/utils.js
• packages/blaze/view.js
• packages/caching-html-compiler/.npm/package/.gitignore
• packages/caching-html-compiler/.npm/package/README
• packages/caching-html-compiler/.npm/package/npm-shrinkwrap.json
• packages/caching-html-compiler/caching-html-compiler.js
• packages/caching-html-compiler/package.js
• packages/html-tools/charref.js
• packages/html-tools/charref_tests.js
• packages/html-tools/parse.js
• packages/html-tools/parse_tests.js
• packages/html-tools/scanner.js
• packages/html-tools/templatetag.js
• packages/html-tools/tokenize.js
• packages/html-tools/tokenize_tests.js
• packages/html-tools/utils.js
• packages/htmljs/html.js
• packages/htmljs/htmljs_test.js
• packages/htmljs/visitors.js
• packages/observe-sequence/observe_sequence.js
• packages/observe-sequence/observe_sequence_tests.js
• packages/spacebars-compiler/.npm/package/.gitignore
• packages/spacebars-compiler/.npm/package/README
• packages/spacebars-compiler/.npm/package/npm-shrinkwrap.json
• packages/spacebars-compiler/codegen.js
• packages/spacebars-compiler/compile_tests.js
• packages/spacebars-compiler/compiler.js
• packages/spacebars-compiler/optimizer.js
• packages/spacebars-compiler/package.js
• packages/spacebars-compiler/react.js
• packages/spacebars-compiler/spacebars_tests.js
• packages/spacebars-compiler/templatetag.js
• packages/spacebars-compiler/whitespace.js
• packages/spacebars-tests/old_templates.js
• packages/spacebars-tests/old_templates_tests.js
• packages/spacebars-tests/template_tests.js
• packages/spacebars-tests/templating_tests.js
• packages/spacebars/spacebars-runtime.js
• packages/spacebars/spacebars_tests.js
• packages/templating-runtime/.npm/package/.gitignore
• packages/templating-runtime/.npm/package/README
• packages/templating-runtime/.npm/package/npm-shrinkwrap.json
• packages/templating-runtime/dynamic.js
• packages/templating-runtime/dynamic_tests.js
• packages/templating-runtime/package.js
• packages/templating-runtime/templating.js
• packages/templating-tools/.npm/package/.gitignore
• packages/templating-tools/.npm/package/README
• packages/templating-tools/.npm/package/npm-shrinkwrap.json
• packages/templating-tools/compile-tags-with-spacebars.js
• packages/templating-tools/html-scanner-tests.js
• packages/templating-tools/html-scanner.js
• packages/templating-tools/package.js
• tests/benchmark_ElementAttributesUpdater.html
• tests/benchmark_getUrlProtocol.html

💤 Files with no reviewable changes (20)

• packages/caching-html-compiler/.npm/package/.gitignore
• packages/blaze/.npm/package/.gitignore
• packages/caching-html-compiler/.npm/package/npm-shrinkwrap.json
• packages/templating-tools/.npm/package/.gitignore
• packages/templating-runtime/.npm/package/.gitignore
• packages/spacebars-compiler/.npm/package/.gitignore
• packages/caching-html-compiler/package.js
• packages/spacebars-compiler/.npm/package/README
• packages/caching-html-compiler/.npm/package/README
• packages/templating-runtime/.npm/package/README
• packages/blaze/package.js
• packages/blaze/.npm/package/README
• packages/spacebars-compiler/package.js
• packages/blaze/.npm/package/npm-shrinkwrap.json
• packages/templating-runtime/.npm/package/npm-shrinkwrap.json
• packages/templating-tools/.npm/package/npm-shrinkwrap.json
• packages/templating-tools/.npm/package/README
• packages/spacebars-compiler/.npm/package/npm-shrinkwrap.json
• packages/templating-tools/package.js
• packages/templating-runtime/package.js

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

Use a repository-relative image path instead of a branch-pinned raw URL.

Line 18 is brittle and can break when branch names or paths change. Prefer a relative path so docs render consistently across branches/forks.

📌 Suggested fix

-![package overview](https://raw.githubusercontent.com/meteor/blaze/architecture/images/architecture_packages.svg)
+![package overview](./images/architecture/architecture_packages.svg)

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	
	

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@OVERVIEW.md` at line 18, Replace the branch-pinned raw URL image link in the
markdown (the current
"https://raw.githubusercontent.com/meteor/blaze/architecture/images/architecture_packages.svg"
reference) with a repository-relative path to the asset (e.g. use the relative
Markdown image URL pointing at architecture/images/architecture_packages.svg) so
the docs render reliably across branches and forks; update the image markdown in
OVERVIEW.md accordingly.

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

Add language identifiers to fenced code blocks (MD040).

Lines 199, 209, and 384 use unlabeled fenced blocks; markdownlint flags this.

🧹 Suggested fix

-```
+```text
 .html file → templating-compiler
            → caching-html-compiler
            → templating-tools.scanHtmlForTags
            → html-tools (parsing)
            → spacebars-compiler (compilation)
            → JavaScript code (output)

- +text
Template function called
→ Creates Blaze.View
→ View.render() → HTMLjs structure
→ Blaze.materialize() → DOM nodes
→ Reactive dependencies tracked (Tracker)
→ Data changes → View.invalidate()
→ Selective DOM updates (observe-sequence for {{#each}})

-```
+```text
Template (Blueprint)
    ↓ constructView()
View (Low-level rendering primitive)
    ↑ referenced by
Template Instance (Specific instantiation, created alongside View)
    ↓ renders to
HTMLjs (Intermediate representation)
    ↓ materializes to
DOM Nodes (Actual browser elements)

</details>


Also applies to: 209-217, 384-394

<details>
<summary>🧰 Tools</summary>

<details>
<summary>🪛 markdownlint-cli2 (0.22.0)</summary>

[warning] 199-199: Fenced code blocks should have a language specified

(MD040, fenced-code-language)

</details>

</details>

<details>
<summary>🤖 Prompt for AI Agents</summary>

Verify each finding against the current code and only fix it if needed.

In @OVERVIEW.md around lines 199 - 206, The unlabeled fenced code blocks flagged
by MD040 should each include a language identifier (use "text")—specifically add

block that begins "Template function called …", and the block that begins
"Template (Blueprint) …" so all three fenced blocks are labeled and markdownlint
MD040 is satisfied.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Don't short-circuit handler.update on cached values.

Line 749 skips handler.update based only on the last rendered value, but several handlers reconcile against the live element. After this change, rerendering the same template value no longer restores template-owned value/checked/class/style after user input or external DOM mutation.

Suggested fix

-    const last = lastValues[k];
-    const shouldUpdate = !(last === value ||
-                          (last == null && value == null));
-                          
-    if (shouldUpdate) {
-      // Update the handler's stored value and apply to DOM
-      handler.value = value;
-      handler.update(elem, oldValue, value);
-      lastValues[k] = value;
-      
-      // Clean up if attribute was removed (value set to null)
-      if (value === null) {
-        delete handlers[k];
-        delete lastValues[k];
-      }
-    }
+    handler.value = value;
+    handler.update(elem, oldValue, value);
+    lastValues[k] = value;
+
+    if (value === null) {
+      delete handlers[k];
+      delete lastValues[k];
+    }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/blaze/attrs.js` around lines 746 - 762, The current short-circuit
uses shouldUpdate to skip calling handler.update when the new value equals
lastValues[k], which prevents handlers that reconcile against the live DOM
(e.g., for value/checked/class/style) from restoring template-owned state after
user/external mutation; change the logic in the block that references
lastValues, handlers, handler and shouldUpdate so that handler.update(elem,
oldValue, value) is always called regardless of shouldUpdate, but only update
handler.value and lastValues[k] (and perform the value === null cleanup that
deletes handlers[k] and lastValues[k]) when shouldUpdate is true — keep the same
calls and symbols (handler.update, handler.value, lastValues, handlers, value,
oldValue, elem) and only move the update invocation out of the short-circuit
guard.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

🧩 Analysis chain

🌐 Web query:

Does jQuery.parseHTML(html, context) exclude <script> elements when keepScripts is omitted, and does setting template.innerHTML preserve <script> elements inside template.content?

💡 Result:

Yes, jQuery.parseHTML(html, context) excludes <script> elements when keepScripts is omitted (defaults to false), and yes, setting template.innerHTML preserves <script> elements inside template.content.

Citations:

• 1: https://api.jquery.com/jQuery.parseHTML/
• 2: https://github.com/jquery/jquery/blob/2.2-stable/src/core/parseHTML.js
• 3: https://api.jquery.com/jquery.parsehtml/
• 4: https://developer.mozilla.org/en-US/docs/Web/API/HTMLTemplateElement/content
• 5: https://developer.mozilla.org/en-US/docs/Web/HTML/Element/template
• 6: https://stackoverflow.com/questions/35513437/template-tag-innerhtml-breaks-script-tags

---

Filter <script> nodes in the native parseHTML path.

Lines 37-42 no longer match the jQuery backend: $.parseHTML(...) excludes scripts by default, but the <template> branch includes them unchanged. This means the same HTML produces different DOM depending on whether jQuery is present, removing the script-filtering behavior that the old backend provided.

Suggested fix

 DOMBackend.parseHTML = function (html) {
   if (_hasJQuery) {
     return $jq.parseHTML(html, DOMBackend.getContext()) || [];
   }
   const template = document.createElement('template');
   template.innerHTML = html;
+  template.content.querySelectorAll('script').forEach((script) => script.remove());
   return Array.from(template.content.childNodes);
 };

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	DOMBackend.parseHTML = function (html) {
	// Return an array of nodes.
	//
	// jQuery does fancy stuff like creating an appropriate
	// container element and setting innerHTML on it, as well
	// as working around various IE quirks.
	return $jq.parseHTML(html, DOMBackend.getContext()) || [];
	if (_hasJQuery) {
	return $jq.parseHTML(html, DOMBackend.getContext()) || [];
	}
	const template = document.createElement('template');
	template.innerHTML = html;
	return Array.from(template.content.childNodes);
	DOMBackend.parseHTML = function (html) {
	if (_hasJQuery) {
	return $jq.parseHTML(html, DOMBackend.getContext()) || [];
	}
	const template = document.createElement('template');
	template.innerHTML = html;
	template.content.querySelectorAll('script').forEach((script) => script.remove());
	return Array.from(template.content.childNodes);
	};

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/blaze/dombackend.js` around lines 36 - 42, DOMBackend.parseHTML's
native branch currently returns template.content.childNodes which keeps <script>
elements, whereas the jQuery branch ($jq.parseHTML(...)) excludes scripts by
default; update the native path in DOMBackend.parseHTML to filter out SCRIPT
nodes before returning so both branches behave consistently. In practice, after
creating template and populating template.innerHTML, iterate over
template.content.childNodes (or use querySelectorAll) and return only nodes
whose nodeName !== 'SCRIPT' (or nodeType !== Node.ELEMENT_NODE || tagName !==
'SCRIPT') so script elements are excluded from the returned array.

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

Honor type when removing native delegated handlers.

Line 106 removes every stored wrapper for handler, regardless of type. If the same callback is delegated for multiple event types on one element, the first undelegateEvents call tears all of them down in native mode.

Suggested fix

   const handlerMap = _delegateMap.get(elem);
   if (!handlerMap) return;

   const entries = handlerMap.get(handler);
   if (!entries) return;
+  let eventType = DOMBackend.Events.parseEventType(type);
+  eventType = _delegateEventAlias[eventType] || eventType;

-  for (const entry of entries) {
-    elem.removeEventListener(entry.eventType, entry.wrapper);
-  }
-  handlerMap.delete(handler);
+  const remaining = [];
+  for (const entry of entries) {
+    if (entry.eventType === eventType) {
+      elem.removeEventListener(entry.eventType, entry.wrapper);
+    } else {
+      remaining.push(entry);
+    }
+  }
+  if (remaining.length) handlerMap.set(handler, remaining);
+  else handlerMap.delete(handler);

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	undelegateEvents(elem, type, handler) {
	if (_hasJQuery) {
	$jq(elem).off(type, '**', handler);
	return;
	}
	const handlerMap = _delegateMap.get(elem);
	if (!handlerMap) return;
	const entries = handlerMap.get(handler);
	if (!entries) return;
	for (const entry of entries) {
	elem.removeEventListener(entry.eventType, entry.wrapper);
	}
	handlerMap.delete(handler);
	undelegateEvents(elem, type, handler) {
	if (_hasJQuery) {
	$jq(elem).off(type, '**', handler);
	return;
	}
	const handlerMap = _delegateMap.get(elem);
	if (!handlerMap) return;
	const entries = handlerMap.get(handler);
	if (!entries) return;
	let eventType = DOMBackend.Events.parseEventType(type);
	eventType = _delegateEventAlias[eventType] || eventType;
	const remaining = [];
	for (const entry of entries) {
	if (entry.eventType === eventType) {
	elem.removeEventListener(entry.eventType, entry.wrapper);
	} else {
	remaining.push(entry);
	}
	}
	if (remaining.length) handlerMap.set(handler, remaining);
	else handlerMap.delete(handler);

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/blaze/dombackend.js` around lines 94 - 109, The undelegateEvents
implementation currently removes all wrappers for a given handler regardless of
the passed type; update undelegateEvents to only remove entries whose
entry.eventType equals the provided type: look up handlerMap via
_delegateMap.get(elem), get entries = handlerMap.get(handler), iterate and only
call elem.removeEventListener(entry.eventType, entry.wrapper) for entries where
entry.eventType === type (or matches the provided type string), remove those
matching entries from the entries collection, and only call
handlerMap.delete(handler) if no entries remain for that handler; reference
undelegateEvents, _delegateMap, handlerMap, entries, and
entry.eventType/entry.wrapper to locate the change.

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
# Verify negative substring usage in scanner code.
rg -n -C2 --iglob '*scanner.js' 'substring\(\s*-\s*[A-Za-z0-9_$]+'

Repository: meteor/blaze

Length of output: 465

---

🏁 Script executed:

node <<'EOF'
// Verify substring vs slice behavior with negative indices
const testString = "abcdefghij"; // 10 chars
const CONTEXT_AMOUNT = 3;

console.log("Test string:", testString);
console.log("Length:", testString.length);
console.log("\nWith CONTEXT_AMOUNT = 3:");
console.log("substring(-3):", testString.substring(-3));
console.log("slice(-3):", testString.slice(-3));
console.log("\nWith CONTEXT_AMOUNT = 5:");
console.log("substring(-5):", testString.substring(-5));
console.log("slice(-5):", testString.slice(-5));

// Simulate the actual code logic
const pastInput = "This is some context text that is longer than needed";
if (pastInput.length > CONTEXT_AMOUNT) {
  const buggyVersion = `...${pastInput.substring(-CONTEXT_AMOUNT)}`;
  const fixedVersion = `...${pastInput.slice(-CONTEXT_AMOUNT)}`;
  console.log("\nActual code simulation:");
  console.log("Original pastInput length:", pastInput.length);
  console.log("Buggy version (substring):", buggyVersion);
  console.log("Fixed version (slice):", fixedVersion);
}
EOF

Repository: meteor/blaze

Length of output: 394

---

Replace substring(-CONTEXT_AMOUNT) with slice(-CONTEXT_AMOUNT) to properly truncate trailing context.

On line 36, substring(-CONTEXT_AMOUNT) doesn't truncate to the last N characters—it treats the negative index as 0 and returns the entire string. Use slice(-CONTEXT_AMOUNT) instead, which correctly returns the trailing N characters.

Proposed fix

-    pastInput = `...${pastInput.substring(-CONTEXT_AMOUNT)}`;
+    pastInput = `...${pastInput.slice(-CONTEXT_AMOUNT)}`;

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/html-tools/scanner.js` around lines 34 - 37, The truncation logic
for pastInput is wrong because substring(-CONTEXT_AMOUNT) treats negative
indices as 0; in the block that sets pastInput (using variables input, pos,
CONTEXT_AMOUNT and the pastInput variable), replace the use of
substring(-CONTEXT_AMOUNT) with slice(-CONTEXT_AMOUNT) so pastInput =
`...${pastInput.slice(-CONTEXT_AMOUNT)}` to correctly return the trailing N
characters when pastInput.length > CONTEXT_AMOUNT.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

postProcess can erase // inside string literals and weaken test validity.

At Line [29], /\/\/.*$/mg strips from // to EOL indiscriminately. That can truncate literal content (for example http://...) and let distinct outputs compare equal.

🔧 Proposed fix

-        string = string.replace(/\/\/.*$/mg, '');
+        // Strip only standalone comment lines injected by tooling.
+        string = string.replace(/^\s*\/\/.*$/mg, '');

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	const postProcess = function (string) {
	// remove initial and trailing parens
	string = string.replace(/^\(([\S\s]*)\)$/, '$1');
	if (! (Package['minifier-js'] && Package['minifier-js'].UglifyJSMinify)) {
	// these tests work a lot better with access to beautification,
	// but let's at least do some sort of test without it.
	// These regexes may have to be adjusted if new tests are added.
	// ======================== !!NOTE!! =================================
	// Since we are bringing uglify-js in from NPM, this code should no
	// longer ever be needed. Leaving it just in case.
	// ==================================+================================
	// Remove single-line comments, including line nums from build system.
	string = string.replace(/\/\/.*$/mg, '');
	string = string.replace(/\s+/g, ''); // kill whitespace
	}
	// Remove single-line comments, including line nums from build system.
	string = string.replace(/\/\/.*$/mg, '');
	string = string.replace(/\s+/g, ''); // kill whitespace
	const postProcess = function (string) {
	// remove initial and trailing parens
	string = string.replace(/^\(([\S\s]*)\)$/, '$1');
	// Remove single-line comments, including line nums from build system.
	// Strip only standalone comment lines injected by tooling.
	string = string.replace(/^\s*\/\/.*$/mg, '');
	string = string.replace(/\s+/g, ''); // kill whitespace

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/spacebars-compiler/compile_tests.js` around lines 25 - 30, The
postProcess function currently uses the regex /\/\/.*$/mg which naively strips
from // to end-of-line and can remove // sequences inside string literals (e.g.,
"http://..."); replace that regex usage in postProcess with a safe line-by-line
comment stripper that scans each line character-by-character, tracking quote
state and escapes so it only removes // when not inside single or double quotes
(implement a helper like stripLineCommentsPreservingStrings used from
postProcess), then continue with the existing whitespace collapse and
paren-trimming steps; ensure the new code references postProcess and removes the
/\/\/.*$/mg usage.

[coderabbitai]

⚠️ Potential issue | 🟠 Major

Use the local scanner for parsing to avoid cursor desynchronization.

At Line [365], TemplateTag.parse(scannerOrString) bypasses the local scanner. For string input, this parses with a different scanner instance, while subsequent logic (parseFragment, validation, error positions) continues with scanner, producing inconsistent behavior.

🔧 Proposed fix

-  const result = TemplateTag.parse(scannerOrString);
+  const result = TemplateTag.parse(scanner);

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	TemplateTag.parseCompleteTag = function (scannerOrString, position) {
	var scanner = scannerOrString;
	let scanner = scannerOrString;
	if (typeof scanner === 'string')
	scanner = new HTMLTools.Scanner(scannerOrString);
	var startPos = scanner.pos; // for error messages
	var result = TemplateTag.parse(scannerOrString);
	const startPos = scanner.pos; // for error messages
	const result = TemplateTag.parse(scannerOrString);
	if (! result)
	TemplateTag.parseCompleteTag = function (scannerOrString, position) {
	let scanner = scannerOrString;
	if (typeof scanner === 'string')
	scanner = new HTMLTools.Scanner(scannerOrString);
	const startPos = scanner.pos; // for error messages
	const result = TemplateTag.parse(scanner);
	if (! result)

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/spacebars-compiler/templatetag.js` around lines 359 - 366,
TemplateTag.parseCompleteTag is creating a local variable scanner but then calls
TemplateTag.parse(scannerOrString), which can create a different Scanner and
desynchronize cursor positions; change the call to use the local scanner
variable (i.e., call TemplateTag.parse(scanner)) so parsing, parseFragment,
validation and error position reporting all use the same Scanner instance
(update any other places in this function that pass scannerOrString into
internal parse functions to use scanner).

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

🧩 Analysis chain

🏁 Script executed:

#!/bin/bash
set -euo pipefail

# 1) Definition context
rg -nP --type=js -C3 '\bTemplate\._migrateTemplate\s*=\s*function\s*\('

# 2) All call sites
rg -nP --type=js -C2 '\bTemplate\._migrateTemplate\s*\('

# 3) Calls passing 3 arguments (potentially impacted if arg is ignored)
rg -nP --type=js -C2 '\bTemplate\._migrateTemplate\s*\(([^,]*,){2}[^)]*\)'

Repository: meteor/blaze

Length of output: 957

---

🏁 Script executed:

cat -n packages/templating-tools/code-generation.js | sed -n '1,20p'

Repository: meteor/blaze

Length of output: 833

---

🏁 Script executed:

# Look at the complete call in code-generation.js
cat -n packages/templating-tools/code-generation.js | sed -n '7,15p'

Repository: meteor/blaze

Length of output: 371

---

🏁 Script executed:

# Search for any other Template._migrateTemplate calls more broadly
rg -n '_migrateTemplate' packages/

Repository: meteor/blaze

Length of output: 362

---

Remove unused migrate parameter from _migrateTemplate.

The migrate parameter is never passed by callers. The only call site in code-generation.js:9-12 provides only 2 arguments (templateName and newTemplate), making the parameter dead code. Either remove it or, if it's intended as an optional override, change line 163 to migrate = migrate || Template.__pendingReplacement.includes(templateName); to respect caller input when provided.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/templating-runtime/templating.js` around lines 161 - 164, The
function signature for Template._migrateTemplate currently declares an unused
third parameter `migrate`; remove it and compute migration intent locally
instead so caller input isn't expected. Change the signature to
Template._migrateTemplate = function (templateName, newTemplate) { and replace
the reassignment line with a local const, e.g. const migrate =
Template.__pendingReplacement.includes(templateName); (or delete any unused
`migrate` variable entirely if it’s not referenced later).

[coderabbitai]

⚠️ Potential issue | 🟡 Minor

Add alt text to the GitHub Actions badge for accessibility.

The badge image is missing alternative text, which impacts accessibility for screen reader users.

♿ Proposed fix to add descriptive alt text

  <a href="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml">
- 	<img src="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml/badge.svg">
+ 	<img src="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml/badge.svg" alt="Blaze Tests">
  </a>

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	<a href="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml">
	<img src="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml/badge.svg">
	<a href="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml">
	<img src="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml/badge.svg" alt="Blaze Tests">
	</a>

🧰 Tools

🪛 markdownlint-cli2 (0.22.0)

[warning] 6-6: Images should have alternate text (alt text)

(MD045, no-alt-text)

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@README.md` around lines 5 - 6, The badge image tag lacks alternative text;
update the README.md badge <img> element (inside the <a
href="https://github.com/meteor/blaze/actions/workflows/blaze-tests.yml"> link)
to include a descriptive alt attribute (e.g., alt="GitHub Actions — blaze-tests
workflow status") so screen readers can convey the badge meaning; ensure the alt
text is concise and specific to the workflow name/status.

[jankapunkt]

Note, I marked #470 and #473 as optional, moving them to the next 3.2 milestone, if no PRs coming in for them, as we should move towards an alpha release

[jankapunkt]

@italojs @nachocodoner can one of you create an alpha 3.1.0-alpha release of this branches latest commit?

// cc @dupontbertrand @StorytellerCZ

[nachocodoner]

@jankapunkt: Done. blaze@3.1.0-alpha.0 available with all other related packages all published.

[dupontbertrand]

[dupontbertrand]

Status report on 3.1.0-alpha.0, with measurements, so this PR carries a clear picture of what is left.

Verified green

Full Tinytest suite from 7d151bd8 (the commit alpha.0 was published from), run on Meteor 3.5.2 rather than the 3.4 our CI test-app is pinned to:

Configuration	Passed	Failed
native DOM backend	414	0
jQuery 3.7.1 backend	418	0

Real-world: a substantial private compliance app of mine, in development and not launched yet, on Meteor 3.5.2 with published blaze@3.1.0-alpha.0 and no jquery package (so native backend) passes 1172 server + 769 browser tests, 0 failures, identical to its score on Blaze 3.0.x. Ten further projects of mine run alpha.0 on Meteor 3.5.x.

Blocking for a beta: #501 is optional, #513 is not

#513 fixes a regression the alpha actually ships. Proof that it is real and that our current suite does not cover it — the alpha with only #513's two tests added, fix withheld:

Configuration	Passed	Failed
alpha.0 + #513 tests only, native	414	2
alpha.0 + #513 complete, native	416	0
alpha.0 + #513 complete, jQuery 3.7.1	420	0

The two failures are delegated event fires in inclusion wrapped by an element inside {{#if}} (#512) and delegated event fires on direct child of a wrapper element (#512). Impact on my side: two silently dead controls on an admin screen while its eleven render tests stayed green.

Proposed checklist for 3.1.0

• fix(blaze): match delegated event selectors within delegation root #513 reviewed and merged — only known regression in the alpha, fix scoped to dombackend.js
• fix: prevent stale re-renders of #each items during sequence update (#468) #501 reviewed (#each re-rendered with stale data context #468 stale #each data context) — mergeable since June 6, long-standing bug rather than a 3.1 regression, but the fix is ready
• feat: enable server-side Blaze template rendering (foundation for SSG/SSR) #507 — milestone call, 3.1.0 or 3.2. It gates feat: static-render core package for Blaze SSG/SSR meteor#14355; see below
• feat(blaze): Add visual error indicator for development #481 → 3.2 (changes requested), feat: add Blaze.setExceptionLogLevel() to configure exception log level #509 → 3.2 (conflicting), feat: inline expressions in Spacebars #500 already 3.2
• 3.1.0-beta.0 version bump commit (I can prepare it)
• beta published, forum thread opened to collect app reports
• 3.1.0 published, packages/non-core/blaze submodule bumped in meteor/meteor

#507 is a cross-repo decision, not a Blaze-internal one

#507 (server-side Blaze.toHTML(), the SSG/SSR foundation) is MERGEABLE against release-3.1.0, +278/-163 across 5 packages, zero reviews. It carries milestone 3.2 only because the 3.1 vs 3.2 call was left open in April and never resolved.

Meanwhile meteor/meteor#14355 (the static-render core package) is MERGEABLE and milestoned Release 3.6. @nachocodoner asked there, on July 29, for E2E scenarios on an existing E2E app with an SSR route. That E2E can only assert graceful degradation while packages/non-core/blaze is pinned to 3.0.3, because the server-side Template registry does not exist in Blaze 3.0.x.

So the two PRs are deadlocked on each other: #14355 waits on a Blaze release, #507 waits on a milestone decision.

Taking #507 into 3.1.0 closes the chain in the order the core team is already planning: Blaze 3.1.0 final → submodule bump in meteor/meteor → #14355 lands with real E2E coverage in 3.6. Leaving it for 3.2 parks #14355 behind a release with no date.

The cost, stated plainly: #507 is not a no-op. Lifting the 'client' scoping puts blaze and spacebars in the server bundle and compiles .html for every architecture, for any app using templating, at roughly 500 bytes per template server-side. Backward compatible, documented in the PR's guide changes, but a real change of shape for a release currently framed around jQuery becoming optional. It also means the beta must be validated with #507 included, which is one more full run of the matrix above. I will do that run if the call goes that way.

Two CI gaps worth fixing here

1. Our test-app is pinned to METEOR@3.4. The alpha is being used on 3.5.x in the wild and nothing covered that combination until now. Bumping it is the single highest-value CI change.
2. When you bump it, the test-app job stops being native. Since Meteor 3.5, test-in-console bundles jquery (meteor/meteor 70570e76f1), and Blaze's weak jquery dependency loads it, so the job that is supposed to exercise the native backend silently runs jQuery back-compat instead. I had to use a local jquery-free copy of test-in-console to get a genuine native run. Worth asserting Blaze._DOMBackend._hasJQuery === false in that job so the configuration can never drift silently again.

Related, for meteor/meteor rather than here: all three Blaze E2E apps on release-3.6 (apps/blaze, apps/blaze-router, apps/full-blaze) and all four Blaze-bearing skeletons pin jquery, so a core-side validation of 3.1 would also only exercise the jQuery path. One jquery-free variant would cover the native backend.

[nachocodoner]

Thanks for the detailed report. I followed up with meteor/meteor#14754, pointing the Blaze submodule to 7d151bd8 and adding your two delegated-event cases to an existing E2E app. They run with both native and jQuery backends, in development and production.

Locally, I can confirm the regressions: 8 tests pass and 4 fail, with both cases failing only on the native backend in both modes. Applying only the dombackend.js change from #513 makes all 12 pass, and the full Blaze E2E group passes 36/36. These are local results; CI is still pending. The PR keeps the original alpha pin so the failures remain visible. But we should definitely merge #513 for the next round and ensure meteor/meteor#14754 gets passed.

Blaze submodule changes now trigger E2E, and all three Blaze apps verify and log the checked-out commit, resolved version, compiled source, and active backend. I think having this regression coverage in Meteor core is important for the stability of Blaze 3.1.0: it gives us repeatable checks through real apps and Meteor’s build pipeline, and keeps those cases covered as both projects evolve.

Also, I would add resolving this PR's conflicts between release-3.1.0 and master to the proposed checklist.

Once the fixes are merged, the conflicts are resolved, and the updated submodule passes CI, I think that should be our signal to publish the first 3.1.0-beta.0 and communicate it to the community for broader testing.

[nachocodoner]

Also, on your point about #507 being a decision across both repos, I agree we should plan the two changes together. I'm more in favor of keeping #507 for Blaze 3.2 and targeting Meteor 3.7 or later for meteor/meteor#14355. That would also mean moving the Meteor PR out of the 3.6 milestone.

The changes in #507 remove the compiler's archMatching: 'web' restriction and make the Template runtime, blaze, and spacebars available on the server. As the guide in that PR explains, explicit imports control which templates reach the server, while eager-loaded templates add registration work at startup. Those are changes to existing apps using templating, even when they don't enable static-render. The PR estimates around 500 bytes per compiled template; we haven't measured the startup or memory impact ourselves.

The E2E results from meteor/meteor#14754 give us a useful baseline for the current Blaze integration and the event regressions. They don't validate #507 or the new SSG/SSR path. I think that integration deserves its own E2E scenarios with real server-rendered routes, checks for existing apps, and measurements of the additional build and startup work.

For now, I think we already have enough to deliver with a stable Blaze 3.1.0 and the native backend. I'd focus the first beta on the regression fixes, resolving the branch conflicts, and getting broader app feedback. We can continue the SSG/SSR work toward the following release with time to validate both PRs together and keep the risk manageable.

[jankapunkt]

@nachocodoner how do we move forward with test-in-console having jQuery added again?

[nachocodoner]

@nachocodoner how do we move forward with test-in-console having jQuery added again?

We will need to be a PR on the core again, and based on meteor/meteor#14754 where we actually see the new Blaze 3.1.0 adopted as the default coming from the version. Hopefully we can deliver the whole thing for Meteor 3.6.

[jankapunkt]

Sorry for being out of the loop lately, I had some private scrambles... What is the ETA for 3.6? Do we need a meeting to align ourselves? I am already reviewing a few things, I don't want to be the blocking factor here 😅

[dupontbertrand]

Thanks @nachocodoner, agreed on all points.

• feat: enable server-side Blaze template rendering (foundation for SSG/SSR) #507 / feat: static-render core package for Blaze SSG/SSR meteor#14355: fine with Blaze 3.2 and Meteor 3.7+. I can't edit milestones on meteor/meteor, could you move #14355 out of Release 3.6?
• fix: prevent stale re-renders of #each items during sequence update (#468) #501: updated with release-3.1.0 now that fix(blaze): match delegated event selectors within delegation root #513 is in (merge commit). The only conflict was both PRs appending tests to the end of spacebars-tests/template_tests.{html,js}; both blocks kept. Green in native (420/0) and jQuery (424/0) on the pinned test-app release. @jankapunkt, ready for review whenever you have a moment.
• Evaluate Blaze 3.1.0 with Meteor 3.6 meteor#14754 CI: none of its 3 red jobs was related to Native (no-jQuery) event backend silently drops delegated events: scope heuristic compares view.name strings, not view identity #512. I opened ci: fix Blaze E2E git check and html scanner test for Blaze 3.1 meteor#14757 against that branch with the two fixes.
• release-3.1.0 ↔ master conflicts: 34 files, no code. They come from the 3.0.3 release: version bumps in package.js / .versions (all lower than the 3.1.0-alpha versions), HISTORY.md, and site/package-lock.json. Happy to open a PR merging master into release-3.1.0 (keeping the 3.1.0 versions and carrying the 3.0.3 history entry) if that helps.

[jankapunkt]

release-3.1.0 ↔ master conflicts: 34 files, no code. They come from the 3.0.3 release: version bumps in package.js / .versions (all lower than the 3.1.0-alpha versions), HISTORY.md, and site/package-lock.json. Happy to open a PR merging master into release-3.1.0 (keeping the 3.1.0 versions and carrying the 3.0.3 history entry) if that helps.

Yes that would help a lot 👍

[jankapunkt]

#515 is merged, so is #511 however the merge conflict seems to remain 🤔

[dupontbertrand]

My bad, I should have flagged this on #515: it needed a merge commit. The squash kept the resolved content but dropped master as a second parent, so git still sees the 3.0.3 commits as unmerged and #492 reports the same conflicts.

The content on release-3.1.0 is already right, so the fix only records the history, with no file change. -s ours is safe here because #515 already brought every master change over. Two options:

• Directly, since you have push rights:

  git checkout release-3.1.0 && git pull
  git merge -s ours origin/master -m "Merge master into release-3.1.0 (record history only)"
  git push origin release-3.1.0
  

• Or via Merge master into release-3.1.0 (record history only) #516, which contains exactly that commit (0 files changed). Please merge it with Create a merge commit, not squash.

I checked locally: the resulting tree is identical to release-3.1.0, master becomes an ancestor, and #492 no longer conflicts.

[jankapunkt]

@dupontbertrand @nachocodoner I moved everything but #501 to 3.2 so this is the remaining last item to review and merge toward 3.1.0

[nachocodoner]

Could we do a documentation-gap review before merging this release?

A few examples I noticed:

• The migration guide says jQuery was removed, but the implementation makes it optional and continues using it as the DOM backend when installed.
• The Templates API says TemplateInstance.$ always returns a jQuery object, while without jQuery it returns an array of DOM elements.
• The JSDoc and TypeScript declarations do not accurately represent both return types.
• Some migrated pages still contain Hexo-specific directives, such as {% apibox %}, {% dtdd %}, and the changelog tag, which VitePress does not process correctly.
• The documentation build and deployment should be verified in CI to catch these issues.

These should be reviewed so the documentation accurately explains the new optional-jQuery behavior and the migrated site renders correctly.

Also, lets keep coderabbitai comments resolved or closed, but not remaining open.

[jankapunkt]

@nachocodoner I would address these remaining points, do you think it makes sense to publish a beta in parallel? I think most of these do not involve significant code changes so I think a beta + announcement in the forums for early adoption tests can make sense, what do you think?

[nachocodoner]

Yes, that is indeed the next step for it. We will do it next.

[jankapunkt]

Hey @nachocodoner FYI since we have a reported issue in #519 and #520 we may have to release another beta / rc before finally publishing 3.1.0.

[nachocodoner]

Yes, please. Let me know if you can work them out so we can include them on a next beta release. Let's try to get the stable version of Blaze 3.1.0 into the next minor release, Meteor 3.7, at least. If I have time to dive into them, I will. Otherwise, I will need help with the fixes and reviews.

cc @dupontbertrand