StringMatcher.isMatch(CharSequence, int, int, int) forwards bufferEnd as bufferStart

[lenamonj]

StringMatcher.isMatch(CharSequence, int, int, int) forwards bufferEnd in the bufferStart slot:

return isMatch(CharSequenceUtils.toCharArray(buffer), start, bufferEnd, bufferEnd);

A matcher that implements only the abstract char[] overload is handed bufferStart == bufferEnd through the CharSequence path. StringSubstitutor searches a TextStringBuilder, so its matcher calls bind to that overload; with a caller-supplied prefix matcher that honours bufferStart, new StringSubstitutor(values).replace("a ${key} b") returns the input unsubstituted.

One argument changed. New StringMatcherTest fails on master with expected: <1> but was: <5> and expected: <a value b> but was: <a ${key} b>.

Verified with the default mvn goal (1894 tests, checkstyle, PMD, SpotBugs, japicmp).

Found by an automated review loop I run; the fix and test were reviewed and verified by me.

[garydgregory]

Hi @lenamonj
Nice find, thank you. There are two changes I'd like to see:

1. In testIsMatchOverloadsAgreeOnTheSameWindow(), the second case passes start = 1 with bufferStart = 2, outside the documented contract. The method tests the helper’s defensive behavior rather than required API behavior. A valid window that ends before the complete pattern would be a more useful negative case.

2. Add a direct call to the two-argument isMatch(CharSequence, int) using the custom matcher. This would explicitly cover the convenience overload that delegates through the corrected method. Explicit expected match lengths would also make the parity assertions easier to interpret.

[lenamonj]

Thanks Gary, both done in 4933c6a:

1. The negative case is now a valid window that ends before the pattern completes: isMatch("xabz", 1, 0, 2) returns 0 through both overloads. The out-of-contract case is gone.
2. New testIsMatchCharSequenceStartDelegatesThroughTheCorrectedOverload calls isMatch("xabz", 1) and isMatch("xabz", 2) directly and expects 2 and 0. All parity assertions now state the expected length explicitly.

Against master's StringMatcher all four tests fail (the two-argument call returns 0 where 2 is expected); with the fix the default mvn goal is green, 1895 tests.

[garydgregory]

Thank you @lenamonj , merged 🚀