fix facebook feed zero rows after decoy filter - #4
Merged
lawmight merged 1 commit intoSep 10, 2026
Merged
Conversation
Live smoke after PR #3 hit no_rows with articles=2 because decoy filtering dropped scrambled posts in-page without surfacing why. Add French action-menu and metric labels, fix dedupe skipping author-only decoy containers, and record per-article rejection reasons (decoy/suggestion/messenger/etc). When all candidates are decoys, fail with anti-scrape error even at zero rows; otherwise include rejection breakdown in the generic no_rows hint. Co-authored-by: Tom Coustols <tom.coustols@tcdynamics.fr>
|
Important Draft PR not reviewedDraft PRs are not automatically reviewed by default.
To automatically review draft PRs, update your CodeRabbit configuration: reviews:
auto_review:
drafts: trueThanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Reviewer's GuideThe extractor now supports French Facebook feed markup, preserves candidates rejected by readability and scope checks for diagnostics, and distinguishes all-decoy zero-row results from other extraction failures with actionable error details. Sequence diagram for Facebook feed extraction diagnosticssequenceDiagram
participant FacebookPage
participant Extractor
participant FeedCommand
participant CLI
FacebookPage->>Extractor: buildFeedExtractScript(limit)
Extractor->>Extractor: dedupe containers with fallback keys
Extractor->>Extractor: extractPost(container)
alt readable post
Extractor-->>FeedCommand: rows and diagnostics
FeedCommand-->>CLI: extracted feed rows
else rejected candidate
Extractor->>Extractor: rejectionSample(container)
Extractor-->>FeedCommand: no_rows and diagnostics.rejections
FeedCommand->>FeedCommand: summarizeFeedRejections(diagnostics)
alt all candidates are decoys
FeedCommand-->>CLI: CommandExecutionError anti-scrape decoy
else other candidates rejected
FeedCommand-->>CLI: CommandExecutionError with rejection breakdown
end
end
Flow diagram for French Facebook feed extractionflowchart TD
A[Facebook feed markup] --> B[Recognize French labels and metrics]
B --> C[Find author and content blocks]
C --> D{Readable candidate?}
D -->|Yes| E[Build feed row]
D -->|No| F[Record rejection reason author and snippet]
E --> G[Return extracted rows]
F --> H{All rejections are decoys?}
H -->|Yes| I[Raise anti-scrape decoy error]
H -->|No| J[Raise no-rows error with counts and samples]
File-Level Changes
Tips and commandsInteracting with Sourcery
Customizing Your ExperienceAccess your dashboard to:
Getting Help
|
lawmight
marked this pull request as ready for review
September 10, 2026 20:48
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Hey - I've found 2 issues
Prompt for AI Agents
Please address the comments from this code review:
## Individual Comments
### Comment 1
<location path="clis/facebook/feed.js" line_range="902" />
<code_context>
+ ).length,
mainTextLength: textOf(document.querySelector('[role="main"]')).length,
feedFound: !!document.querySelector('[role="main"] [role="feed"], [role="feed"]'),
+ rejections: rejections.slice(0, 12),
surface,
},
</code_context>
<issue_to_address>
**issue (broader_impact):** The extractor truncates rejection diagnostics to the first 12 entries before Node classifies them with `allDecoy`. If the first 12 candidates are decoys but a later rejected candidate has another reason, the omitted candidate cannot prevent the command from incorrectly raising the all-decoy anti-scrape error.
**Triggers:** When more than 12 candidate articles are rejected and the first 12 are decoys.
**Suggested fix:** Compute the rejection summary from all candidates and truncate only the diagnostic payload presented to the caller.
</issue_to_address>
### Comment 2
<location path="clis/facebook/feed.js" line_range="897-899" />
<code_context>
actionMenuCount: actionMenuAnchors().length,
- fallbackActionCount: document.querySelectorAll('[role="main"] [aria-label="Like"], [role="main"] [aria-label="赞"], [role="main"] [aria-label="Comment"], [role="main"] [aria-label="评论"]').length,
+ fallbackActionCount: document.querySelectorAll(
+ '[role="main"] [aria-label="Like"], [role="main"] [aria-label="赞"], [role="main"] [aria-label="J\\'aime"], '
+ + '[role="main"] [aria-label="Comment"], [role="main"] [aria-label="评论"], [role="main"] [aria-label="Commenter"]',
+ ).length,
mainTextLength: textOf(document.querySelector('[role="main"]')).length,
feedFound: !!document.querySelector('[role="main"] [role="feed"], [role="feed"]'),
</code_context>
<issue_to_address>
**nitpick:** The French fallback selector counts `J'aime`, `Commenter`, and the other like/comment labels but omits the newly supported `Partager` label, even though `fallbackContainers` recognizes it. A French page whose visible fallback evidence includes share actions therefore reports an incomplete `actions` diagnostic and can lose that action evidence in no-row classification.
**Triggers:** When French feed markup exposes `Partager` as the relevant fallback action and the other counted action labels are absent.
**Suggested fix:** Add `[aria-label="Partager"]` to the `fallbackActionCount` selector, keeping it aligned with `fallbackContainers`.
</issue_to_address>Sourcery assessment
Approval pending. 1 finding to address first.
Blocking findings: clis/facebook/feed.js:902
| ).length, | ||
| mainTextLength: textOf(document.querySelector('[role="main"]')).length, | ||
| feedFound: !!document.querySelector('[role="main"] [role="feed"], [role="feed"]'), | ||
| rejections: rejections.slice(0, 12), |
There was a problem hiding this comment.
issue (broader_impact): The extractor truncates rejection diagnostics to the first 12 entries before Node classifies them with allDecoy. If the first 12 candidates are decoys but a later rejected candidate has another reason, the omitted candidate cannot prevent the command from incorrectly raising the all-decoy anti-scrape error.
Triggers: When more than 12 candidate articles are rejected and the first 12 are decoys.
Suggested fix: Compute the rejection summary from all candidates and truncate only the diagnostic payload presented to the caller.
Comment on lines
+897
to
+899
| '[role="main"] [aria-label="Like"], [role="main"] [aria-label="赞"], [role="main"] [aria-label="J\\'aime"], ' | ||
| + '[role="main"] [aria-label="Comment"], [role="main"] [aria-label="评论"], [role="main"] [aria-label="Commenter"]', | ||
| ).length, |
There was a problem hiding this comment.
nitpick: The French fallback selector counts J'aime, Commenter, and the other like/comment labels but omits the newly supported Partager label, even though fallbackContainers recognizes it. A French page whose visible fallback evidence includes share actions therefore reports an incomplete actions diagnostic and can lose that action evidence in no-row classification.
Triggers: When French feed markup exposes Partager as the relevant fallback action and the other counted action labels are absent.
Suggested fix: Add [aria-label="Partager"] to the fallbackActionCount selector, keeping it aligned with fallbackContainers.
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
After PR #3, live smoke on tom.coustols stopped returning scrambled Adolph decoys (good) but started failing with
no feed rows could be extracteddespitearticles=2, actions=1, mainTextLength=976. The in-page extractor was rejecting scrambled posts before rows were built, so Node never saw decoy rows and fell through to the generic error.Related: live smoke 2026-09-11, OpenCLI ae8e790
Type of Change
Changes
Actions pour cette publication par …, J'aime/Commenter/Partager, Tous:/commentaires/partages, Partagé avec Publicdiagnostics.rejections[]with reason/author/snippet per candidate (decoy_content, suggestion, messenger, insufficient_evidence, comment_thread, etc.)CommandExecutionErrorwithdecoy_content=Nbreakdownrejected=N, reason=count, samples=…when articles exist but nothing extractsChecklist
npm test -- clis/facebook/— 112 passed (includes Adolph/decoy + French fixtures)Post-merge verification
On tom.coustols desk:
opencli facebook feed --limit 5should either return readable posts (e.g. LVLUP with Lani) OR fail withdecoy_content=2/ rejection samples — not a barearticles=2hint with no explanation.Summary by Sourcery
Improve Facebook feed extraction to support French content and explain zero-row results caused by filtered or unreadable posts.
Bug Fixes:
Enhancements:
Tests:
Summary by cubic
Fixes the Facebook feed extractor returning zero rows after decoy filtering, so live feeds with scrambled posts now either return readable posts or fail with a clear anti-scrape error instead of a bare no-rows hint.
Changes
decoy_content,suggestion,messenger, etc.) in diagnostics.CommandExecutionErrorwhen all candidates are decoys, even at zero rows.Written for commit 61a0b99. Summary will update on new commits.