fix(content-drive): truncate long-text field values in listing rows (#37185)

[ihoffmann-dot]

Summary

• Adds TransformOptions.LONG_TEXT_PREVIEW (declared after STORY_BLOCK_VIEW/JSON_VIEW so EnumSet iteration order runs it last) backed by a new LongTextPreviewStrategy.
• Replaces WYSIWYG/TextArea (Jsoup-extracted plain text) and Story Block (new recursive JSON-tree traversal) field values in a listing row with a ≤150-character preview, instead of the full raw value.
• Wired opt-in only at BrowserAPIImpl#dotContentMap via a new DotTransformerBuilder#longTextPreview() chain method — never added to defaultOptions, so no other transformer consumer (Content Editor, ContentResource, GraphQL, asset picker) is affected (AC-007).
• Bug found and fixed while writing the AC-008 test: if a content type's title-source field is itself WYSIWYG/TextArea (its variable is literally title), the strategy would have overwritten the already-correct, untruncated title COMMON_PROPS computes. Fixed by explicitly skipping the title key.
• Updates the @Schema description on ContentDriveResource#search (AC-005; endpoint is @Hidden, no openapi.yaml regen needed) and removes a dead item.body assertion in the Postman collection (AC-006, a listing row never carried that key).

Test plan

• ./mvnw test -pl :dotcms-core -Dtest=LongTextPreviewStrategyTest
• just test-integration-ide
• ./mvnw verify -pl :dotcms-integration -Dcoreit.test.skip=false -Dit.test=BrowserAPITest
• ./mvnw verify -pl :dotcms-postman -Dpostman.test.skip=false -Dpostman.collections=ContentDriveResource
• Confirm no System.out/System.getProperty/System.getenv introduced (checked via diff)
• Latency (p50/p95) is measured, not gated (D7) — requires the Content Drive: folder listing query scans by mod_date instead of by folder (450ms to 101ms), plus 3 related performance items #37148 benchmark dataset, not reproducible in CI

Branched off the approved spec branch per this repo's Spec-Kit flow (spec.md-only in PR1, not merged to main yet).

🤖 Generated with Claude Code

This PR fixes: #37185

Verification (2026-09-04, local)

• 12/12 unit tests (LongTextPreviewStrategyTest) pass.
• 45/45 integration tests (BrowserAPITest) pass.
• 87/87 Postman assertions (ContentDriveResource collection) pass.

Bugs found and fixed in the test code along the way (not the production fix): a package-private field access across packages (fixed via reflection), a test title exceeding the contentlet.title column's varchar(255) limit, an incorrect expected key name (__icon__ vs. the real icon) plus two File-Asset-only keys (mimeType/extension) wrongly expected on a generic-Content row.

[claude]

Claude finished @ihoffmann-dot's task in 2m 27s —— View job

---

Code Review — long-text preview for Content Drive listings

• Read LongTextPreviewStrategy, TransformOptions, DotTransformerBuilder, StrategyResolverImpl, BrowserAPIImpl wiring
• Verified strategy ordering (EnumSet ordinal order → runs last)
• Checked interaction with StoryBlockViewStrategy / JSONViewStrategy
• Reviewed null/empty and surrogate-pair edge handling

I verified the core correctness claim end-to-end: DotTransformerBuilder.build() passes EnumSet.copyOf(optionsHolder), DotContentletTransformerImpl stores that reference directly (line 71) and iterates it in resolveStrategies — so ordinal order holds and LONG_TEXT_PREVIEW (declared last) runs after STORY_BLOCK_VIEW/JSON_VIEW. I also confirmed JSONViewStrategy only touches JSONField fields, so it doesn't re-serialize the story-block Map back to a string before the preview runs. The defaultOptions set includes STORY_BLOCK_VIEW, so the story-block value is already a Map by the time extractStoryBlockPreview sees it. The design is sound and the opt-in is correctly isolated to the BrowserAPIImpl#dotContentMap call site.

New Issues

• 🟡 Medium: LongTextPreviewStrategy.java:73-77 — applyPreview unconditionally calls map.put(field.variable(), ...) for every matching field, even when the field key is absent from the map. When map.get(...) returns null, all three extractors return StringPool.BLANK, so a field that had no value (and thus no key in the transformed map) gets a new "" key inserted into the listing row. This is a subtle response-shape change: rows for content with an empty WYSIWYG/TextArea/Story-Block field would newly carry an empty-string key.
  • Assumption: DefaultTransformStrategy may or may not include unset fields in the map — if it always emits "" for these field types, this is a no-op and can be ignored.
  • What to verify: for a Content Type with an empty WYSIWYG/Story-Block field, does the listing row change from "no key" to key: ""? If BrowserAPITest already asserts on such a row, it's covered; otherwise consider guarding with if (map.containsKey(field.variable())) before overwriting. Fix this →

Everything else checks out:

• Surrogate-pair guard in truncate (lines 151-153) is correct — a high surrogate at index 149 cuts at 149, and a high surrogate landing at index 150 is naturally excluded by substring(0, 150), so no lone surrogate can be emitted.
• collectText short-circuits at MAX_PREVIEW_LENGTH before recursing/appending, avoiding full traversal of large story blocks.
• AC-008 title-key skip (line 72) correctly protects the COMMON_PROPS-computed title.
• Per-field failures are caught via Try.run(...).onFailure(...) and logged with Logger.warn — no swallowed exceptions, no System.out.
• No SQL, no new external calls, no permission-surface change; opt-in is not added to defaultOptions, so Content Editor / ContentResource / GraphQL / asset picker are unaffected (AC-007 verified).

No blocking issues. The one Medium is non-blocking and may already be a no-op depending on DefaultTransformStrategy behavior.
· branch issue-37185-content-drive-listing-longtext-projection-impl

[github-actions]

dotbot code review:

• Reviewer: meta/muse-spark-1.3 (medium)
• Overall: patch is correct
• New findings this run: 0
• Prior unresolved dotbot findings still relevant: 0
• Active findings total: 0

Opt-in LONG_TEXT_PREVIEW correctly runs last via enum ordinal order, is isolated to dotContentMap, and handles HTML/Story Block truncation with title-key protection and surrogate safety. No P0/P1 defect with concrete repo evidence was found.

Tip: comment with "/dotbot address comments" to attempt automated fixes for unresolved review threads.

_{reviewed by dotbot · meta/muse-spark-1.3 · medium}

[github-actions]

LongTextPreviewStrategy.java:74 absent field keys get inserted as empty strings

Current code:

fields.stream()
    .filter(field -> !TITTLE_KEY.equals(field.variable()))
    .forEach(field -> Try.run(() ->
        map.put(field.variable(), extractor.apply(map.get(field.variable()))))

Problem: Fields absent from the map (unset value) get a new "" key inserted, changing row shape.

Fix:

    .forEach(field -> Try.run(() -> {
        if (map.containsKey(field.variable())) {
            map.put(field.variable(), extractor.apply(map.get(field.variable())));
        }
    }))

What to verify: a Content Type with an empty WYSIWYG/Story Block field whose key is absent from Contentlet#getMap(); the listing row currently gains key: "".

[github-actions]

dotbot code review:

• Reviewer: ~z-ai/glm-latest (medium)
• Overall: patch is incorrect
• New findings this run: 1
• Prior unresolved dotbot findings still relevant: 0
• Active findings total: 1

The opt-in LONG_TEXT_PREVIEW strategy is correctly ordered last via enum ordinal, isolated to the dotContentMap call site, and handles HTML/Story Block truncation, the title-key collision, and surrogate pairs safely. The only residual issue is a low-severity possible empty-string key insertion for unset fields.

Tip: comment with "/dotbot address comments" to attempt automated fixes for unresolved review threads.

_{reviewed by dotbot · ~z-ai/glm-latest · medium}