ADFA-3709 | Refactor OCR fuzzy parsing and modularize YOLO to XML conversion

[jatezzz]

Description

This PR improves the accuracy and maintainability of the computer vision to XML pipeline.

• Updated FuzzyAttributeParser to include comprehensive regex-based sanitization for common OCR typos (e.g., misreading '0' as 'o', '5' as 's') and common UI label typos (e.g., "usemame" to "Username").
• Introduced strict fuzzy matching for categorical XML values such as inputType, gravity, and textStyle to ensure valid Android XML output.
• Refactored YoloToXmlConverter to reduce cyclomatic complexity. The large mapping and appending logic was broken down into smaller, specialized functions (assignTextToParents, matchAnnotationsToElements, appendTextViewAttributes, etc.).

Details

Logic-related changes. Verified parsing improvements and corrected UI typo mappings via application logs. XML generation now cleanly delegates to component-specific builder methods.

Ticket

ADFA-3709

[coderabbitai]

Warning

Rate limit exceeded

@jatezzz has exceeded the limit for the number of commits that can be reviewed per hour. Please wait 44 minutes and 49 seconds before requesting another review.

Your organization is not enrolled in usage-based pricing. Contact your admin to enable usage-based pricing to continue reviews beyond the rate limit, or try again in 44 minutes and 49 seconds.

⌛ How to resolve this issue?

After the wait time has elapsed, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout.

Please see our FAQ for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: af25b656-6d01-4041-9cdf-5650646eb7ad

📥 Commits

Reviewing files that changed from the base of the PR and between 5727d0d and eb6b697.

📒 Files selected for processing (3)

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

📝 Walkthrough

Walkthrough

FuzzyAttributeParser adds regex-based OCR normalization, fuzzy categorical matching, and refined id/number cleaning. YoloToXmlConverter refactors layout generation: scales widgets, assigns OCR text boxes to parents, derives canvas tags, matches annotations to elements, and extracts attribute-emission into new helpers.

Changes

Cohort / File(s)	Summary
Fuzzy attribute parsing cv-image-to-xml/.../FuzzyAttributeParser.kt	Added precompiled regexes for id/number cleanup and OCR character normalization, implemented matchCategoricalValue(...), updated cleanValue(...) and cleanId(...) (including btm→btn fix), and simplified delimiter handling and unsafe !! usages.
YOLO → XML conversion pipeline cv-image-to-xml/.../YoloToXmlConverter.kt	Refactored generateXmlLayout flow: scale widgets (exclude widget_tag), assign OCR text boxes to parent boxes, derive canvas tags, match annotations to elements via new matchAnnotationsToElements, and moved view attribute logic into appendTextViewAttributes, appendEditTextAttributes, appendImageViewAttributes. Defaults for missing width/height now use scaled box dims.

Sequence Diagram

sequenceDiagram
    participant Converter as YoloToXmlConverter
    participant Scaler as Scaler
    participant TextAssigner as TextAssigner
    participant TagDeriver as TagDeriver
    participant Matcher as AnnotationMatcher
    participant Builder as XmlBuilder
    participant Parser as FuzzyAttributeParser

    Converter->>Scaler: Scale YOLO widget boxes (exclude widget_tag)
    Scaler-->>Converter: scaledBoxes

    Converter->>TextAssigner: assignTextToParents(scaledBoxes, ocrBoxes)
    TextAssigner-->>Converter: parentTextMappings

    Converter->>TagDeriver: derive canvas tags (widget_tag + OCR tag text)
    TagDeriver-->>Converter: canvasTags

    Converter->>Matcher: matchAnnotationsToElements(canvasTags, scaledBoxes, parentTextMappings)
    Matcher-->>Converter: finalAnnotations

    Converter->>Builder: build XML from sortedBoxes + finalAnnotations
    Builder->>Parser: sanitize/normalize attribute values
    Parser-->>Builder: cleanedAttributes
    Builder-->>Converter: xmlLayout

Loading

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~60 minutes

Possibly related PRs

• ADFA-3108: Improve CV-to-XML accuracy with fuzzy search and OCR refinement  #1047: Modifies the same FuzzyAttributeParser and integrates sanitization into YoloToXmlConverter’s parsing flow.
• ADFA-3107: Utilize YOLO detected tags in XML generation - Experimental #1028: Changes YoloToXmlConverter’s annotation/text-assignment and XML-generation pipeline with overlapping helpers.

Suggested reviewers

• Daniel-ADFA
• avestaadfa
• hal-eisen-adfa

Poem

🐇 In pixels' hush I hop and parse,

O to 0, and l to 1 I sparse;
Text finds parents, tags align,
Boxes scale and attributes shine—
A tiny rabbit cheers the XML march.

🚥 Pre-merge checks | ✅ 2 | ❌ 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 (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main changes: refactoring OCR fuzzy parsing in FuzzyAttributeParser and modularizing YOLO to XML conversion in YoloToXmlConverter.
Description check	✅ Passed	The description is directly related to the changeset, covering the three main improvements: regex-based sanitization, fuzzy matching for categorical values, and refactoring to reduce complexity.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch refactor/ADFA-3709-ocr-fuzzy-parsing-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]

🧹 Nitpick comments (1)

cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/FuzzyAttributeParser.kt (1)

490-492: Consider using kotlin.random.Random for more idiomatic Kotlin.

Math.random() works but kotlin.random.Random.nextInt(1000) is more idiomatic and explicit about the integer range.

♻️ Suggested improvement

+import kotlin.random.Random
+
 // In cleanId function:
-            return "view_${(Math.random() * 1000).toInt()}"
+            return "view_${Random.nextInt(1000)}"

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In
`@cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/FuzzyAttributeParser.kt`
around lines 490 - 492, Replace the use of Math.random() in the
FuzzyAttributeParser return expression with the idiomatic Kotlin random integer
API: use kotlin.random.Random.nextInt(1000) to generate the 0–999 int and
construct the ID string (i.e., change the expression inside the if in
FuzzyAttributeParser where it returns "view_${(Math.random() * 1000).toInt()}"
to use Random.nextInt(1000)); add an import for kotlin.random.Random if
required.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In
`@cv-image-to-xml/src/main/java/org/appdevforall/codeonthego/computervision/domain/FuzzyAttributeParser.kt`:
- Around line 490-492: Replace the use of Math.random() in the
FuzzyAttributeParser return expression with the idiomatic Kotlin random integer
API: use kotlin.random.Random.nextInt(1000) to generate the 0–999 int and
construct the ID string (i.e., change the expression inside the if in
FuzzyAttributeParser where it returns "view_${(Math.random() * 1000).toInt()}"
to use Random.nextInt(1000)); add an import for kotlin.random.Random if
required.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro

Run ID: d745e406-afeb-4484-9b73-4ee1ef882fd6

📥 Commits

Reviewing files that changed from the base of the PR and between 017ffd5 and bc85cf0.

📒 Files selected for processing (2)

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

[coderabbitai]

Actionable comments posted: 3

🤖 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/FuzzyAttributeParser.kt`:
- Line 152: The regex ocrLetterBToSixRegex incorrectly maps OCR 'B' to the digit
6; update the mapping and its name to reflect that 'B' is visually confused with
'8' (e.g., rename to ocrLetterBToEightRegex and keep the pattern Regex("[bB]")),
and then update all usages in extractOcrNumber to replace the old mapping
logic/name so the parser substitutes 'B' -> '8' instead of '6'.
- Around line 475-477: The fallback ID generation in FuzzyAttributeParser (the
block checking FuzzySearch.ratio for "match_parent"/"wrap_content") is
non-deterministic and low-cardinality; replace the Math.random() usage with a
thread-safe incremental counter (e.g., an AtomicInteger field in the class) and
produce IDs like "view_${counter.incrementAndGet()}" to ensure deterministic,
unique IDs and avoid test flakiness and collisions.
- Around line 502-511: The extractOcrNumber function currently normalizes the
entire input string (using ocrLetterOToZeroRegex, ocrLetterIToOneRegex,
ocrLetterZToTwoRegex, ocrLetterSToFiveRegex, ocrLetterBToSixRegex) which
corrupts unit suffixes like "sp"/"dp"; change the logic to detect and strip any
trailing unit suffix before normalization (e.g., check for common units such as
"sp", "dp", "px"), apply the regex replacements only to the stripped numeric
part, run numberExtractionRegex on that normalized numeric substring, and then
reattach/return the numeric value (or use it downstream) without having altered
the original unit suffix; update extractOcrNumber to operate on numericPart and
leave unitPart untouched.

🪄 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: 1f8c18e6-85e4-44d2-a113-07314d69b577

📥 Commits

Reviewing files that changed from the base of the PR and between bc85cf0 and 5727d0d.

📒 Files selected for processing (2)

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

🚧 Files skipped from review as they are similar to previous changes (1)

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

[jatezzz]

@coderabbitai

[coderabbitai]

The PR description images show:

• UI_sketch_18: A hand-drawn layout sketch with two EditText fields ("Your name", "User name"), a "Finish" Button (B-1, B-2), and a Switch (SW-1), each annotated with layout attributes like match_parent, 150dp, inputType: text, id: btm_finish, etc.
• photo_5116610229700332736_y: The "Generate XML" screen on an Android tablet showing the sketch with green bounding boxes overlaid on detected UI elements (EditTexts, Button, Switch), with OCR-recognized labels and margin guides visible.
• photo_5116610229700332737_y: The generated activity_main.xml in AndroidIDE, showing a LinearLayout containing two EditText elements (hints: "Your name", "User name"), a Button (text: "Finish"), and a Switch (checked: true) — all with pixel-scaled dp dimensions and tools:ignore="HardcodedText".

---

@jatezzz, it looks like you tagged me without a specific request! Here's what I can help with for PR #1185:

• 🔍 Trigger a full review of the changes in FuzzyAttributeParser.kt and YoloToXmlConverter.kt
• 💬 Answer questions about the PR changes
• 🐛 Investigate specific concerns about the OCR fuzzy parsing or YOLO-to-XML refactoring

Would you like me to proceed with a full review, or is there something specific you'd like me to look at?

---

🧠 Learnings used

Learnt from: hal-eisen-adfa
Repo: appdevforall/CodeOnTheGo PR: 1176
File: app/src/androidTest/kotlin/com/itsaky/androidide/helper/DevicePermissionGrantUiHelper.kt:47-76
Timestamp: 2026-04-15T01:58:44.608Z
Learning: In `appdevforall/CodeOnTheGo`, `clickFirstAccessibilityNodeByText` in `app/src/androidTest/kotlin/com/itsaky/androidide/helper/DevicePermissionGrantUiHelper.kt` pre-filters via `findAccessibilityNodeInfosByText` (case-insensitive text containment) before applying the caller's `matchBy` predicate. This is intentional and correct for all current callers because the targeted buttons (NEXT, accept, grant) have visible rendered text that is matched by the pre-filter. Do not flag the pre-filter as missing contentDescription-only nodes — that scenario does not apply to these UI elements.

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.