• Implement Lucene fulltext search and linked file indexing benchmarks
• Add setup initialization for PDF entries and search infrastructure
• Replace placeholder TODO implementations with functional benchmark methods
• Add teardown cleanup for index directory and resources

Implement Lucene search and indexing benchmarks with setup/teardown

• Added imports for Lucene indexing, file searching, and file utilities
• Introduced instance variables for linkedFilesIndexer, linkedFilesSearcher,
 fulltextSearchQuery, luceneIndexDir, and pdfEntries
• Enhanced init() method to create temporary Lucene index directory, mock file preferences and
 database context, and initialize PDF entries with linked files
• Implemented search() benchmark to execute fulltext search queries using LinkedFilesSearcher
• Implemented index() benchmark to measure index rebuild performance by removing and re-adding
 entries
• Added @TearDown method to close indexer and clean up temporary index directory

jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java

Description

The updated init() method now performs unrelated setup steps (temp dir creation, mocking
preferences/context, test data creation, indexing, and query creation) in one place. This reduces
readability and makes future benchmark setup changes harder to maintain.

Code

jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[R100-123]

+        luceneIndexDir = Files.createTempDirectory("jabref-benchmark-lucene");
+
+        FilePreferences filePreferences = mock(FilePreferences.class);
+        when(filePreferences.shouldFulltextIndexLinkedFiles()).thenReturn(true);
+
+        Path pdfResourceDir = Path.of("src/test/resources/pdfs");
+        BibDatabaseContext linkedFilesContext = mock(BibDatabaseContext.class);
+        when(linkedFilesContext.getDatabasePath()).thenReturn(Optional.of(pdfResourceDir.resolve("dummy.bib")));
+        when(linkedFilesContext.getFileDirectories(filePreferences)).thenReturn(List.of(pdfResourceDir));
+        when(linkedFilesContext.getFulltextIndexPath()).thenReturn(luceneIndexDir);
+
+        pdfEntries = List.of(
+                new BibEntry(StandardEntryType.PhdThesis)
+                        .withCitationKey("ExampleThesis2017")
+                        .withFiles(List.of(new LinkedFile("Example Thesis", "thesis-example.pdf", StandardFileType.PDF.getName())))
+        );
+        when(linkedFilesContext.getEntries()).thenReturn(pdfEntries);
+
+        linkedFilesIndexer = new DefaultLinkedFilesIndexer(linkedFilesContext, filePreferences);
+        linkedFilesIndexer.addToIndex(pdfEntries, mock(BackgroundTask.class));
+
+        linkedFilesSearcher = new LinkedFilesSearcher(linkedFilesContext, linkedFilesIndexer, filePreferences);
+
+        fulltextSearchQuery = new SearchQuery("title", EnumSet.of(SearchFlags.FULLTEXT));

Evidence

PR Compliance ID 10 requires new/changed methods to remain small and single-responsibility. The
added block in init() combines multiple concerns (filesystem temp dir, Mockito wiring, entry
creation, indexing, and search query setup) into a single method.

AGENTS.md
jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[100-123]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

## Issue description
`Benchmarks.init()` has grown to include multiple unrelated responsibilities (temp filesystem setup, Mockito configuration, test data creation, indexing setup, and query construction), making it harder to read and maintain.

## Issue Context
This code is benchmark setup code, but it still benefits from being structured into focused helpers to keep the benchmark class understandable.

## Fix Focus Areas
- jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[100-123]

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

Benchmarks.init() builds the PDF fixture directory using a relative path, so PDF resolution (and
thus indexing) depends on the JVM working directory. If the benchmark is launched from a different
directory, linked file resolution can fail and the Lucene index will be empty or incomplete, making
the benchmark results meaningless.

Code

jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[R105-110]

+        Path pdfResourceDir = Path.of("src/test/resources/pdfs");
+        BibDatabaseContext linkedFilesContext = mock(BibDatabaseContext.class);
+        when(linkedFilesContext.getDatabasePath()).thenReturn(Optional.of(pdfResourceDir.resolve("dummy.bib")));
+        when(linkedFilesContext.getFileDirectories(filePreferences)).thenReturn(List.of(pdfResourceDir));
+        when(linkedFilesContext.getFulltextIndexPath()).thenReturn(luceneIndexDir);
+

Evidence

The benchmark passes a relative directory ("src/test/resources/pdfs") as the base directory for
linked-file resolution. LinkedFile.findIn delegates to FileUtil.find, which checks existence via
directory.resolve(fileName) + Files.exists; if the directory is relative, the filesystem lookup is
implicitly relative to the current working directory.

jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[105-116]
jablib/src/main/java/org/jabref/model/entry/LinkedFile.java[244-270]
jablib/src/main/java/org/jabref/logic/util/io/FileUtil.java[446-495]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
The benchmark uses `Path.of("src/test/resources/pdfs")` as a relative path. Because linked-file resolution ultimately uses `Files.exists(directory.resolve(fileName))`, a relative directory makes benchmark execution depend on the JVM working directory.

### Issue Context
The benchmark is meant to be runnable via JMH/Gradle forks and also potentially from IDEs or from the generated JMH runner jar. These environments can differ in `user.dir`.

### Fix Focus Areas
- jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[100-115]

### Suggested fix approach
- Resolve the PDF fixture location in a working-directory-independent way, e.g.:
 - Add the needed PDF(s) under `jablib/src/jmh/resources/pdfs/` and load via `Benchmarks.class.getResource("/pdfs/thesis-example.pdf")`, converting the URL to a `Path`, OR
 - Compute an absolute path robustly (e.g., probe both `Path.of("src/test/resources/pdfs")` and `Path.of("jablib/src/test/resources/pdfs")` and fail fast with a clear error if neither exists).

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools

Description

Benchmarks.index() creates a new Mockito BackgroundTask mock on every benchmark invocation, adding
allocation/reflection overhead into the measured time. This distorts the indexing benchmark and can
dominate results for small indexes.

Code

jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[R161-163]

+    public void index() throws IOException {
+        linkedFilesIndexer.removeAllFromIndex();
+        linkedFilesIndexer.addToIndex(pdfEntries, mock(BackgroundTask.class));

Evidence

The index() @Benchmark method calls mock(BackgroundTask.class) each invocation. The indexer uses
the task object repeatedly per file (isCancelled/setTitle/updateProgress/updateMessage/showToUser),
so a reusable no-op BackgroundTask instance is sufficient and avoids per-iteration Mockito overhead.

jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[160-164]
jablib/src/main/java/org/jabref/logic/search/indexing/DefaultLinkedFilesIndexer.java[138-151]

Agent prompt

The issue below was found during a code review. Follow the provided context and guidance below and implement a solution

### Issue description
The JMH `index()` benchmark currently includes Mockito mock creation (`mock(BackgroundTask.class)`) in the timed section. This adds non-indexing overhead and reduces benchmark fidelity.

### Issue Context
`DefaultLinkedFilesIndexer.addToIndex` only calls methods on the `BackgroundTask` for progress/cancel checks; it doesn’t require Mockito behavior.

### Fix Focus Areas
- jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[68-76]
- jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[118-123]
- jablib/src/jmh/java/org/jabref/benchmarks/Benchmarks.java[160-164]

### Suggested fix approach
- Add a field like `private BackgroundTask<Void> noopIndexTask;`
- Initialize it once in `@Setup` using `BackgroundTask.wrap(() -> {})` (or a small custom BackgroundTask subclass).
- Use the same instance in both the setup indexing and in `index()`:
 - `linkedFilesIndexer.addToIndex(pdfEntries, noopIndexTask);`
This keeps the benchmark focused on Lucene/PDF indexing work.

ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools