- Match each path portion with
Matcherfrom@isaacs/nonrecursive-glob-match, in place of aRegExp. Items inMinimatch.setare now strings,Matcherobjects, orGLOBSTAR. - Remove
makeRe(),Minimatch#regexp, and theMMRegExptype. - Remove the
ASTclass. - Remove the
maxExtglobRecursionoption. - An empty path portion, such as the last one in
a/b/, only matches an empty pattern portion. Soa/b/no longer matchesa/b/?(x)ora/b/*(x), the same asa/b/*. - Add the
normalizeoption, to normalize unicode in the pattern and in each portion of the tested path. - Remove the
debugoption and thedebug()method. - Remove the
matchOne(),slashSplit(), andlevelTwoFileOptimize()methods. They are internal now. - Fix patterns with more than one
**and sections of different lengths between them, such as**/a/b/c/**/d/**, which failed to match some paths since 10.2.3. - Fix
partialmatching of patterns with**, which since 10.2.3 matched some paths under a dot portion that**can not walk, and did not match some paths that end in the pattern's last portion. partialmatching of patterns with many**no longer takes exponential time.- On Windows, a
..no longer removes the drive letter, or the host and share of a UNC path, at the start of a pattern or (at optimization level 2) a path. SoC:/a/../../xno longer matchesxor/x. - On Windows, compare the host and share of UNC paths like
//host/share/...case-insensitively, the same as drive letters. - On Windows, long UNC paths like
//?/UNC/host/share/...match//host/share/...and vice versa, the same as//?/c:/...andc:/.... - At optimization level 2, throw a
RangeErrorif splitting**/..portions would make more thanbraceExpandMaxpatterns, or add more than 1,000,000 path portions. Before, a short pattern could take minutes to compile. - At optimization level 2, dedupe patterns without comparing each
pair, so that a large set is deduped in about linear time. Fix
{**/x,x/y}, which was deduped to**/x, sox/ydid not match. - Match a numeric brace range of 100 or more values, such as
{0..999}, as one pattern portion, in place of one pattern for each value. For a glob walk, this is one directory read, in place of a stat for each value.hasMagic()is true for these patterns. minimatch()keeps at least the 256Minimatchobjects that it used most recently, so it does not compile the same pattern again.minimatch.filter()gets itsMinimatchobject from the same cache asminimatch(), the first time that it tests a path. A later change to the options object does not change that filter.- Faster to compile brace patterns that make rows that are merged,
such as
**/*.{png,jpg,gif}, because theMatchers for the rows that are merged are not made. - Fix in 11.0.6: include inherited properties in the minimatch
objects created by the
minimatch()andfilter()methods, the same as they are handled innew Minimatch. - Fix: a numeric range of 100 or more values inside a
!(...),*(...)or+(...)extglob is expanded, so it matches the same strings as its expansion. Before,file-!({1..100}).txtdid not matchfile-5.txt, and*({1..100})matched510.
- Add
braceExpandMaxoption
- Add
magicalBracesoption forescape - Fix
makeRewhenpartial: trueis set. - Fix
makeRewhen pattern ends in a final**path part.
- Require node 20 or 22 and higher
- No default export, only named exports.
- Recursive descent parser for extglob, allowing correct support for arbitrarily nested extglob expressions
- Bump required Node.js version
- Add
escape()method - Add
unescape()method - Add
Minimatch.hasMagic()method
- Add support for posix character classes in a unicode-aware way.
- Add
windowsNoMagicRootoption
- Add
optimizationLevelconfiguration option, and revert the default back to the 6.2 style minimal optimizations, making the advanced transforms introduced in 7.0 opt-in. Also, process provided file paths in the same way in optimizationLevel:2 mode, so most things that matched with optimizationLevel 1 or 0 should match with level 2 as well. However, level 1 is the default, out of an abundance of caution.
- Preprocess patterns to simplify complicated patterns and reduce
out
..pattern portions where possible. Note that this means a pattern likea/b/../*will be equivalent toa/*, and will not match the stringa/b/../c. If this causes problems, it can be addressed in a patch release by resolving..portions in the test string.
- Add nocaseMagicOnly flag
-
Handle UNC paths on Windows
This adds some slightly modified behavior when a pattern or path starts with
//on Windows.- In the case of
//?/<drive letter>:/..., the?is treated as a literal character, rather than a wildcard. That is,//?/c:will not match//x/c:. - UNC patterns starting with
//?/<drive letter>:/will match file paths starting with<drive letter>:if the drive letters match case-insensitively. - File paths starting with
//?/<drive letter>:/will match file patterns starting with<drive letter>:if the drive letters match case-insensitively.
- In the case of
-
Add
{preserveMultipleSlashes:true}option to suppress the behavior where multiple consecutive/characters would be effectively coerced into a single path portion separator.
- hybrid module supporting both
require()andimport - exported types
- use windowsPathNoEscape/allowWindowsEscape opts
- make character classes more faithful to bash glob behavior
- fix handling of escapes
- treat invalid character classes as non-matching pattern rather than escaped literals
- brace-expansion: ignore only blocks that begins with $
- Expect exclusively forward slash as path sep, same as node-glob
- makeRe: globstar should match zero+ path portions
- Fix bug with escaped '@' in patterns
- treat
nocase:trueas always having magic - expose GLOBSTAR marker
- Update to modern JS syntax
- Add
allowWindowsEscapeoption
- Added basic redos protection
- Handle unfinished
!(extglob patterns - Add
partial: trueoption