ADFA-3871 | Improve dropdown entries parsing and metadata detection

[jatezzz]

Description

This PR refines the parsing logic within the cv-image-to-xml module to improve accuracy:

• AndroidWidget: Improved dropdown/spinner entries parsing. It now gracefully falls back to text attributes or box text, safely removes trailing dropdown glyphs (e.g., ▼, ⌄), and standardizes string arrays.
• MarginAnnotationParser: Updated logic to process explicit blocks even when the canvasTags list is empty.
• MetadataDetector: Removed the broad assignmentRegex check to prevent false positives when detecting canvas metadata.

Details

Logic-related updates:

• Added toSpinnerEntries(), removeTrailingDropdownGlyph(), and isMeaningfulDropdownText() extension functions in AndroidWidget.kt to clean up text values.
• Adjusted the boolean evaluation in MarginAnnotationParser.kt to include canvasTags.isEmpty().
• Removed assignmentRegex from MetadataDetector.kt.

document_5154784628711625137.mp4

Ticket

ADFA-3871

Observation

The removal of the assignment-based metadata detection in MetadataDetector is intended to reduce false positives where standard text with colons or equals signs was being incorrectly flagged as canvas metadata.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: c2d36c9d-b830-484c-90c8-32042ce6f43c

📥 Commits

Reviewing files that changed from the base of the PR and between 0ba7a54 and de74011.

📒 Files selected for processing (1)

• cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/xml/AndroidWidget.kt

---

📝 Walkthrough

Release Notes - ADFA-3871: Improve dropdown entries parsing and metadata detection

Features & Improvements

• Enhanced Dropdown Entry Parsing: SpinnerWidget.processAttributes now intelligently sources spinner entries from multiple fallback locations (@entries attribute → @text attribute → box.text), with improved handling of dropdown glyphs and standardized string array storage

  • Parses inline entries by splitting on multiple delimiters (commas, semicolons, pipes, slashes, newlines)
  • Stores parsed entries in context.stringArrays using the pattern "{id}_array"
  • Properly handles both reference-style (@array/...) and inline entry definitions

• Reduced False Positives in Canvas Metadata Detection: Removed generic assignment pattern matching (assignmentRegex) from MetadataDetector.isCanvasMetadata() to prevent incorrectly flagging ordinary text containing colons or equals signs as canvas metadata

• Improved Annotation Processing: MarginAnnotationParser now processes explicit blocks even when canvasTags is empty, ensuring annotations are captured in scenarios with no canvas tag references

Risks & Considerations

• Fallback Parsing Risk: The fallback to box.text for spinner entries may inadvertently parse unrelated text content if both @entries and @text attributes are missing. Recommend validating that fallback text matches expected dropdown patterns in integration testing

• Metadata Detection Regression: Removing assignmentRegex may miss legitimate canvas metadata represented as simple key=value patterns. Monitor for any legitimate metadata that relies on this pattern and isn't caught by XML snippet or keyword checks

• Glyph Removal Scope: The new removeTrailingDropdownGlyph() function removes trailing Unicode dropdown symbols globally. Verify that no legitimate text values end with these glyphs (▼, ⌄, etc.) in target UI components

Walkthrough

The PR updates explicit-tag annotation assignment in MarginAnnotationParser, refactors SpinnerWidget to derive and normalize spinner entries (including storing inline lists as string arrays), and removes generic key/value inference from MetadataDetector's canvas-metadata detection.

Changes

Cohort / File(s)	Summary
Margin Annotation Processing cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/MarginAnnotationParser.kt	When parsing explicit blocks, always set annotationMap[tag] = block.annotationText if canvasTags is empty (previously assigned only when tag existed in canvasTags).
Spinner Widget Entries Management cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/xml/AndroidWidget.kt	SpinnerWidget.processAttributes now sources entries from attrs[entries], then attrs[text], then box.text (excluding literal "dropdown" markers). If source starts with @ it is copied (escaped); otherwise inline content is split/normalized into a string array stored in context.stringArrays and entries is set to an @array/{id}_array reference. XML-escaping is skipped for entries and text.
Metadata Detection Simplification cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/utils/MetadataDetector.kt	Removed the generic key/value assignmentRegex inference and related counting logic; isCanvasMetadata no longer treats generic assignments as canvas metadata and returns false when no XML/attribute/keyword matches are found.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~20 minutes

Possibly related PRs

• ADFA-3755 | Fix Spinner entries OCR parsing and validation #1234 — Related to Spinner entries handling; this PR updates entries derivation while #1234 adds entries validation/mapping.
• ADFA-3736 | Ignore canvas metadata and fix attribute parsing #1241 — Overlaps with MarginAnnotationParser and MetadataDetector changes; may conflict on metadata detection and explicit-tag assignment.
• ADFA-3107: Utilize YOLO detected tags in XML generation - Experimental #1028 — Also modifies MarginAnnotationParser's explicit-tag handling; likely directly related.

Suggested reviewers

• avestaadfa
• Daniel-ADFA

"I nibble tags and strings with glee,
margins fed when canvases flee,
spinners learn lists that once were terse,
metadata pruned from verbose verse —
hop, parse, and ship with rabbit mirth!" 🐇✨

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and specifically describes the main changes: improving dropdown entries parsing and metadata detection, which aligns with the primary objectives of the PR.
Description check	✅ Passed	The description is well-related to the changeset, providing clear details about all three main modifications and linking to the associated ticket.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/ADFA-3871-improve-parser-logic-experimental

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.
Inline comments:
In
`@cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/xml/AndroidWidget.kt`:
- Around line 97-100: The check using rawEntries.startsWith("@") should ignore
leading whitespace so values like " `@array/options`" are treated as resource
references; in the block handling AttributeKey.ENTRIES.xmlName, first compute a
trimmedLeading = rawEntries.trimStart() (or similar) and use
trimmedLeading.startsWith("@") for the branch decision, and when storing into
processed[AttributeKey.ENTRIES.xmlName] call escapeXmlAttr() on the
trimmedLeading value so the resource reference is preserved and escaped
correctly (update the condition and the value passed to escapeXmlAttr() in the
rawEntries -> processed handling).
- Around line 126-129: The current removeTrailingDropdownGlyph extension on
String uses a regex that removes ASCII 'v' and 'V' and therefore strips valid
trailing characters (function: String.removeTrailingDropdownGlyph); update the
regex to only target actual dropdown glyphs (use their Unicode code points or
named characters) and remove 'v' and 'V' from the character class—e.g., replace
Regex("\\s*[▼▽▾▿⌄˅∨vV]$") with one that omits ASCII v/V (for example
Regex("\\s*[▼▽▾▿⌄˅∨]$") or explicitly list Unicode escapes like
"\\s*[\\u25BC\\u25BD\\u25BE...]$") so legitimate trailing 'v'/'V' in values
(e.g., "SUV") are preserved.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: cd07db65-f17d-4adf-b7af-c001ab5ab9a3

📥 Commits

Reviewing files that changed from the base of the PR and between 14898ac and 0ba7a54.

📒 Files selected for processing (3)

• cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/MarginAnnotationParser.kt
• cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/xml/AndroidWidget.kt
• cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/utils/MetadataDetector.kt