[WIP]feature: support more general image extraction #672
Conversation
…age using base url 3.support normalizing base url
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Summary of Changes
Hello @alcholiclg, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request significantly enhances the image extraction capabilities within the docling HTML backend. The primary goal is to improve the handling of relative image URLs by accurately capturing and utilizing the base URL of the document being processed, and by centralizing and refining the image fetching and URL resolution logic.
Highlights
- Improved Base URL Handling: The system now captures the
current_urlof a document during conversion and makes it available to the HTML backend, allowing for more accurate resolution of relative image paths. This is achieved by patchingDocumentConverter._convertto inject thecurrent_urlinto the document's backend attributes and updatingDocLoader.loadto set this URL for each document. - Refactored URL Resolution Logic: The
validate_urlutility function has been renamed toresolve_urland significantly enhanced. It now accepts an explicitbase_urlparameter and includes a newnormalized_basehelper to correctly append a trailing slash to base URLs that represent directories, ensuring more robust URL concatenation. - Centralized Image Extraction: A new
extract_imageutility function has been introduced. This function consolidates the logic for fetching images from both data URIs and standard URLs, leveraging the updatedresolve_urlfor non-absolute paths. Thehtml_handle_figureandhtml_handle_imagemethods in thedoclingbackend now utilize this centralized function. - Expanded Unit Tests: The unit tests for image extraction (
tests/doc/test_img.py) have been updated to include more comprehensive scenarios. Test variables for URLs are now lists, allowing for multiple URLs to be tested per category, and extracted images are saved into task-specific directories for better organization and verification.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
| need_append_slash = ('.' not in parsed.path.split('/')[-1] | ||
| or 'arxiv.org' in parsed.netloc) |
There was a problem hiding this comment.
This logic includes a hardcoded special case for arxiv.org. This makes the function less generic and might require adding more special cases in the future as you encounter other sites with similar URL structures. Consider adding a code comment to explain why this special case is necessary. If a more general solution is feasible, that would be even better for long-term maintainability.
tests/doc/test_img.py
Outdated
| os.path.join(save_dir, 'picture_' + str(idx) + '.png')) | ||
| assert len( | ||
| doc_results[0].pictures) > 0, 'No pictures found in the document.' | ||
| doc_results = doc_loader.load(urls_or_files=self.absolute_path_img_url) |
There was a problem hiding this comment.
The load method is called with self.absolute_path_img_url, which is a list. However, the loop iterates through doc_results without using the task_id to access the corresponding URL from the list. This could lead to incorrect assertions or errors if the number of documents loaded doesn't match the number of URLs.
tests/doc/test_img.py
Outdated
| os.path.join(save_dir, 'picture_' + str(idx) + '.png')) | ||
| assert len( | ||
| doc_results[0].pictures) > 0, 'No pictures found in the document.' | ||
| doc_results = doc_loader.load(urls_or_files=self.relative_path_img_url) |
There was a problem hiding this comment.
The load method is called with self.relative_path_img_url, which is a list. However, the loop iterates through doc_results without using the task_id to access the corresponding URL from the list. This could lead to incorrect assertions or errors if the number of documents loaded doesn't match the number of URLs.
tests/doc/test_img.py
Outdated
| os.path.join(save_dir, 'picture_' + str(idx) + '.png')) | ||
| assert len( | ||
| doc_results[0].pictures) > 0, 'No pictures found in the document.' | ||
| doc_results = doc_loader.load(urls_or_files=self.figure_tag_img_url) |
There was a problem hiding this comment.
The load method is called with self.figure_tag_img_url, which is a list. However, the loop iterates through doc_results without using the task_id to access the corresponding URL from the list. This could lead to incorrect assertions or errors if the number of documents loaded doesn't match the number of URLs.
tests/doc/test_img.py
Outdated
| os.path.join(save_dir, 'picture_' + str(idx) + '.png')) | ||
| assert len( | ||
| doc_results[0].pictures) > 0, 'No pictures found in the document.' | ||
| doc_results = doc_loader.load(urls_or_files=self.data_uri_img_url) |
There was a problem hiding this comment.
The load method is called with self.data_uri_img_url, which is a list. However, the loop iterates through doc_results without using the task_id to access the corresponding URL from the list. This could lead to incorrect assertions or errors if the number of documents loaded doesn't match the number of URLs.
initialize need_append_slash to False Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
alcholiclg
changed the title
Change Summary
Related issue number
Checklist
pre-commit installandpre-commit run --all-filesbefore git commit, and passed lint check.