[
https://issues.apache.org/jira/browse/OPENNLP-1932?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18119051#comment-18119051
]
ASF GitHub Bot commented on OPENNLP-1932:
-----------------------------------------
krickert commented on PR #1278:
URL: https://github.com/apache/opennlp/pull/1278#issuecomment-5827823341
Thanks for the review. I pushed changes for all of it. AI provided a better
summary than myself:
> - Removed `asRegex` and `matchesPattern` outright, along with their tests.
`model-loading.xml` now says they are gone in 3.0 and that subclasses should
call `matchesWildcard(url, "*.bin")`.
> - A null `jarModelPrefix` now throws `IllegalArgumentException`, the same
as `matchesWildcard`.
> - The matcher walks both strings by code point in place, so it no longer
allocates two `int[]` per call. The Javadoc now says "code point" throughout.
> - Renamed `GlobMatcher` to `WildcardMatcher`.
> - `splitClassPath` now delegates to a new
`StringUtil.splitNonEmpty(CharSequence, char...)`. #1320 and #1321 can use the
same helper instead of their own loops.
> - Moved the `matchesWildcard` tests into a new
`AbstractClassPathModelFinderTest`. `WildcardMatcherTest` now covers only the
matcher.
> - Added `testPathologicalWildcardRejectsQuickly`, which runs `*a*a*a*b`
against 100k `a`s.
> - `DirectoryModelFinderTest` now counts the jars it copies instead of
assuming 4.
> - A class loader URL that is not a valid URI, for example one with an
unescaped space, is matched on its file part as is instead of failing the scan.
> - Shortened the model loading chapter paragraph, and put the removal note
in its own paragraph.
> - Updated the title.
> RegEx removal part 4: wildcard matching in the model resolver
> -------------------------------------------------------------
>
> Key: OPENNLP-1932
> URL: https://issues.apache.org/jira/browse/OPENNLP-1932
> Project: OpenNLP
> Issue Type: Improvement
> Reporter: Kristian Rickert
> Assignee: Kristian Rickert
> Priority: Major
> Fix For: 3.0.0
>
> Time Spent: 1h
> Remaining Estimate: 0h
>
> {{AbstractClassPathModelFinder}}, {{DirectoryModelFinder}}, and
> {{SimpleClassPathModelFinder}} converted wildcard strings to regular
> expressions and split the class path with compiled patterns. A
> package-private {{GlobMatcher}} now matches {{*}} (any run) and {{?}} (one
> code point) directly, anchored on the whole name and, as before, not crossing
> line terminators; the class path split is a character scan with
> {{String.split}} semantics.
> One behavior change, intended: characters the old conversion never quoted
> ({{( ) [ ] { } + | ^ $ \}}) had regex meaning or threw
> {{PatternSyntaxException}}; they are literal now.
> API: the protected {{asRegex}} and {{matchesPattern(URL, Pattern)}} on the
> public abstract {{AbstractClassPathModelFinder}} are replaced by protected
> {{matchesWildcard(URL, String)}}. No code in the repository used them; this
> is a 3.0 change.
> Verification: 39 globs by 46 inputs plus a 300,000-pair fuzz compared with
> the old regex path; all differences come from the unquoted characters above.
> {{GlobMatcherTest}} adds 64 cases and the module gains its first
> {{DirectoryModelFinderTest}}. The module pom adds the managed
> {{junit-jupiter-params}} dependency.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
