NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
npm · #4792 most downloaded on npm
Language support infrastructure for the CodeMirror code editor
Last release 3 months ago
25 Jun 2026
Ships fairly regularly
a new release about every 2 months
Nearly every release is documented
notes for 55 of 57 stable releases
Nothing withdrawn
no release was ever pulled
6 years old
57 releases · first in 2020
One column per quarter.
Make adding a lot of folded ranges in one transaction linear rather than quadratic in range-set-building complexity.
Make adding a lot of folded ranges in one transaction linear rather than quadratic in range-set-building complexity.
Fix a crash in bracketMatching when composing at end of document.
Fix a crash in bracketMatching when composing at end of document.
Make sure brackets are highlighted in the initial editor state.
Make sure brackets are highlighted in the initial editor state.
Pause bracket matching updates during composition, to avoid disrupting Mobile Safari's fragile composition handling.
Improve finding inner language in syntax tree when the nested parse has been marked as bracketed.
Improve finding inner language in syntax tree when the nested parse has been marked as bracketed.
Nothing published for this version
Make the stream parser user 4 times smaller chunks to reduce the amount of re-parsed code on changes.
Make the stream parser user 4 times smaller chunks to reduce the amount of re-parsed code on changes.
Make sure folded ranges open when backspacing or deleting into them.
Make sure folded ranges open when backspacing or deleting into them.
Fix an issue where indentation would sometimes miss nodes in mixed-language situations.
Fix an issue where indentation would sometimes miss nodes in mixed-language situations.
Stream parsers now support a mergeTokens option that can be used to turn off automatic merging of adjacent tokens.
Stream parsers now support a mergeTokens option that can be used to turn off automatic merging of adjacent tokens.
Fix a regression introduced 6.10.7 that caused indention to sometimes crash on nested language boundaries.
Fix a regression introduced 6.10.7 that caused indention to sometimes crash on nested language boundaries.
Fix an issue where indentation for a stream language would fail to work when the parse covered only part of the document, far from the start.
Fix an issue where indentation for a stream language would fail to work when the parse covered only part of the document, far from the start.
Make sure the inner mode gets a chance to indent when indenting right at the end of a nested language section.
Fix a crash in StreamLanguage when the input range is entirely before the editor viewport.
Fix a crash in StreamLanguage when the input range is entirely before the editor viewport.
Fix an issue where a StreamLanguage could get confused when trying to reuse existing parse data when the parsed range changed.
Fix an issue where a StreamLanguage could get confused when trying to reuse existing parse data when the parsed range changed.
Join adjacent tokens of the same type into a single token in .
Join adjacent tokens of the same type into a single token in .
Call stream language indent functions even when the language is used as a nested parser.
Fix a crash in StreamParser when a parse was resumed with different input ranges.
Fix a TypeScript error when using HighlightStyle with the exactOptionalPropertyTypes typechecking option enabled.
Fix a TypeScript error when using HighlightStyle with the exactOptionalPropertyTypes typechecking option enabled.
Make delimitedIndent align to spaces after the opening token.
Fix an infinite loop that could occur when enabling bidiIsolates in documents with both bidirectional text and very long lines.
Fix an infinite loop that could occur when enabling bidiIsolates in documents with both bidirectional text and very long lines.
Fix an issue where, when a lot of code is visible in the initial editor, the bottom bit of code is shown without highlighting for one frame.
Fix an issue where, when a lot of code is visible in the initial editor, the bottom bit of code is shown without highlighting for one frame.
The new bidiIsolates extension can be used to wrap syntactic elements where this is appropriate in an element that isolates their text direction, avoi
The new bidiIsolates extension can be used to wrap syntactic elements where this is appropriate in an element that isolates their text direction, avoiding weird ordering of neutral characters on direction boundaries.
Fix an issue in StreamLanguage where it ran out of node type ids if you repeatedly redefined a language with the same token table.
Fix an issue in StreamLanguage where it ran out of node type ids if you repeatedly redefined a language with the same token table.
Allow StreamParser tokens get multiple highlighting tags.
Allow StreamParser tokens get multiple highlighting tags.
Indentation now works a lot better in mixed-language documents that interleave the languages in a complex way.
Indentation now works a lot better in mixed-language documents that interleave the languages in a complex way.
Code folding is now able to pick the right foldable syntax node when the line end falls in a mixed-parsing language that doesn't match the target node.
Make getIndentation return null, rather than 0, when there is no syntax tree available.
Make getIndentation return null, rather than 0, when there is no syntax tree available.
The new preparePlaceholder option to codeFolding makes it possible to display contextual information in a folded range placeholder widget.
The new baseIndentFor method in TreeIndentContext can be used to find the base indentation for an arbitrary node.
The new baseIndentFor method in TreeIndentContext can be used to find the base indentation for an arbitrary node.
Export DocInput class for feeding editor documents to a Lezer parser.
Export DocInput class for feeding editor documents to a Lezer parser.
Syntax-driven language data queries now support sublanguages, which make it possible to return different data for specific parts of the tree produced
Syntax-driven language data queries now support sublanguages, which make it possible to return different data for specific parts of the tree produced by a single language.
Make indentation for stream languages more reliable by having StringStream.indentation return overridden indentations from the indent context.
Make indentation for stream languages more reliable by having StringStream.indentation return overridden indentations from the indent context.
The toggleFold command folds or unfolds depending on whether there's an existing folded range on the current line.
indentUnit now accepts any (repeated) whitespace character, not just spaces and tabs.
The bracketMatchingHandle node prop can now be used to limit bracket matching behavior for larger nodes to a single subnode (for example the tag name
The bracketMatchingHandle node prop can now be used to limit bracket matching behavior for larger nodes to a single subnode (for example the tag name of an HTML tag).
Fix a bug that caused ensureSyntaxTree to return incomplete trees when using a viewport-aware parser like StreamLanguage.
Fix a bug that caused ensureSyntaxTree to return incomplete trees when using a viewport-aware parser like StreamLanguage.
Make syntax-based folding include syntax nodes that start right at the end of a line as potential fold targets.
Make syntax-based folding include syntax nodes that start right at the end of a line as potential fold targets.
Fix the indentService protocol to allow a distinction between declining to handle the indentation and returning null to indicate the line has no definite indentation.
HighlightStyle objects now have a specs property holding the tag styles that were used to define them.
HighlightStyle objects now have a specs property holding the tag styles that were used to define them.
Language objects now have a name field holding the language name.
Fix a bug where bracketMatching would incorrectly match nested brackets in syntax trees that put multiple pairs of brackets in the same parent node.
Fix a bug where bracketMatching would incorrectly match nested brackets in syntax trees that put multiple pairs of brackets in the same parent node.
Fix a bug that could cause indentRange to loop infinitely.
Fix a bug that prevented bracket matching to recognize plain brackets inside a language parsed as an overlay.
Fix a bug that prevented bracket matching to recognize plain brackets inside a language parsed as an overlay.
The indentRange function provides an easy way to programatically auto-indent a range of the document.
The foldState field is now public, and can be used to serialize and deserialize the fold state.
The foldState field is now public, and can be used to serialize and deserialize the fold state.
The foldingChanged option to foldGutter can now be used to trigger a recomputation of the fold markers.
The foldingChanged option to foldGutter can now be used to trigger a recomputation of the fold markers.
List style-mod as a dependency.
List style-mod as a dependency.
Make sure all styles in the CSS generated for a HighlightStyle have a lower precedence than the other rules defined for the style. Use a shorthand pro
Make sure all styles in the CSS generated for a HighlightStyle have a lower precedence than the other rules defined for the style. Use a shorthand property
HighlightStyle.get is now called highlightingFor.
HighlightStyle.get is now called highlightingFor.
HighlightStyles no longer function as extensions (to improve tree shaking), and must be wrapped with syntaxHighlighting to add to an editor configuration.
Language objects no longer have a topNode property.
HighlightStyle and defaultHighlightStyle from the now-removed @codemirror/highlight package now live in this package.
The new forceParsing function can be used to run the parser forward on an editor view.
The exports that used to live in @codemirror/matchbrackets are now exported from this package.
The @codemirror/fold package has been merged into this one.
The exports from the old @codemirror/stream-parser package now live in this package.
Autocompletion may now also trigger automatic indentation on input.
Autocompletion may now also trigger automatic indentation on input.
Make sure nodes that end at the end of a partial parse aren't treated as valid fold targets.
Make sure nodes that end at the end of a partial parse aren't treated as valid fold targets.
Fix an issue where the parser sometimes wouldn't reuse parsing work done in the background on transactions.
Fix an issue that could cause indentation logic to use the wrong line content when indenting multiple lines at once.
Fix an issue that could cause indentation logic to use the wrong line content when indenting multiple lines at once.
Fix an issue where the parse worker could incorrectly stop working when the parse tree has skipped gaps in it.
Fix an issue where the parse worker could incorrectly stop working when the parse tree has skipped gaps in it.
Fixes an issue where the background parse work would be scheduled too aggressively, degrading responsiveness on a newly-created editor with a large do
Fixes an issue where the background parse work would be scheduled too aggressively, degrading responsiveness on a newly-created editor with a large document.
Improve initial highlight for mixed-language editors and limit the amount of parsing done on state creation for faster startup.
The new function syntaxTreeAvailable can be used to check if a fully-parsed syntax tree is available up to a given document position.
The new function syntaxTreeAvailable can be used to check if a fully-parsed syntax tree is available up to a given document position.
The module now exports syntaxParserRunning, which tells you whether the background parser is still planning to do more work for a given editor view.
LanguageDescription.of now takes an optional already-loaded extension.
LanguageDescription.of now takes an optional already-loaded extension.
Fix an issue where a parse that skipped content with skipUntilInView would in some cases not be restarted when the range came into view.
Fix an issue where a parse that skipped content with skipUntilInView would in some cases not be restarted when the range came into view.
Fix a bug that caused indentOnInput to fire for the wrong kinds of transactions.
Fix a bug that caused indentOnInput to fire for the wrong kinds of transactions.
Fix a bug that could cause indentOnInput to apply its changes incorrectly.
Fix incorrect versions for @lezer dependencies.
Fix incorrect versions for @lezer dependencies.
CodeMirror now uses lezer 0.15, which means different package names (scoped with @lezer) and some breaking changes in the library.
CodeMirror now uses lezer 0.15, which means different package names (scoped with @lezer) and some breaking changes in the library.
EditorParseContext is now called ParseContext. It is no longer passed to parsers, but must be retrieved with ParseContext.get.
IndentContext.lineIndent now takes a position, not a Line object, as argument.
LezerLanguage was renamed to LRLanguage (because all languages must emit Lezer-style trees, the name was misleading).
Language.parseString no longer exists. You can just call .parser.parse(...) instead.
New IndentContext.lineAt method to access lines in a way that is aware of simulated line breaks.
IndentContext now provides a simulatedBreak property through which client code can query whether the context has a simulated line break.
Fix an issue where asynchronous re-parsing (with dynamically loaded languages) sometimes failed to fully happen.
Fix an issue where asynchronous re-parsing (with dynamically loaded languages) sometimes failed to fully happen.
EditorParseContext.getSkippingParser now replaces EditorParseContext.skippingParser and allows you to provide a promise that'll cause parsing to start
EditorParseContext.getSkippingParser now replaces EditorParseContext.skippingParser and allows you to provide a promise that'll cause parsing to start again. (The old property remains available until the next major release.)
Fix an issue where nested parsers could see past the end of the nested region.
The Language constructor takes an additional argument that provides the top node type.
Update dependencies to 0.18.
The Language constructor takes an additional argument that provides the top node type.
Language instances now have a topNode property giving their top node type.
TreeIndentContext now has a continue method that allows an indenter to defer to the indentation of the parent nodes.
This package now exports a foldInside helper function, a fold function that should work for most delimited node types.
This package now exports a foldInside helper function, a fold function that should work for most delimited node types.
Nothing published for this version
Parse scheduling has been improved to reduce the likelyhood of the user looking at unparsed code in big documents.
Parse scheduling has been improved to reduce the likelyhood of the user looking at unparsed code in big documents.
Prevent parser from running too far past the current viewport in huge documents.
The package now also exports a CommonJS module.
The package now also exports a CommonJS module.
Fix a bug where changing the editor configuration wouldn't update the language parser used.
Fix a bug where changing the editor configuration wouldn't update the language parser used.
### Breaking changes First numbered release.
First numbered release.
Your coding agent can read these notes before it upgrades. Set up the MCP server →