Question 1

Why does opendataloader-pdf print output twice when not using the --quiet flag?

Accepted Answer

A bug causes stdout output to be duplicated when the --quiet option is not set. The fix resolves the promise with an empty string in streaming mode to prevent the CLI from double-writing. Update to the latest version of opendataloader-pdf, or apply the changes from PR #399 manually: in src/index.ts, in the executeJar function, change the resolve call to `resolve(streamOutput ? '' : stdout)`.

Question 2

Why does deleting a PDF after OpenDataLoaderPDF.processFile fail with 'The file cannot be deleted because it is being used by another process'?

Accepted Answer

In OpenDataLoader 2.2.1, the PDF PDDocument and related resources are not properly closed after processing, leaving an active file lock. To fix this, add a resource cleanup method and wrap the existing processFile logic in a try-finally block. Implement the following in DocumentProcessor.java:

```java
// Add this method
private static void closePdfResources() {
    try {
        StaticLayoutContainers.closeContrastRatioConsumer();
    } catch (Exception e) {
        LOGGER.log(Level.WARNING, "Unable to close contrast ratio consumer: " + e.getMessage());
    }
    PDDocument document = StaticResources.getDocument();
    if (document != null) {
        try {
            document.close();
        } catch (Exception e) {
            LOGGER.log(Level.WARNING, "Unable to close PDF document: " + e.getMessage());
        }
    }
}

// Wrap processFile body in try-finally
public static void processFile(String inputPdfName, Config config) throws IOException {
    try {
        // ... original code ...
    } finally {
        closePdfResources();
    }
}
```

After applying this fix, files can be deleted immediately after processing.

Question 3

Why do images fail to display in the generated markdown when the original PDF filename has spaces?

Accepted Answer

The markdown converter produces image links with unescaped spaces, causing renderers to truncate the path at the first space (e.g., `![image](my paper_images/image.png)`). To fix, wrap the image path in angle brackets: `![image](<my paper_images/image.png>)`. This is compliant with CommonMark §6.4 and ensures images render correctly in VS Code, GitHub, and most renderers. The fix was included in commit 4b87a30 (PR #468).

Question 4

How to fix "subprocess.CalledProcessError: returned non-zero exit status 1" when using opendataloader_pdf to convert PDF?

Accepted Answer

This error is caused by a bug in the image extraction logic when processing certain PDFs (e.g., files generated by QuarkXPress). The bug has been fixed in newer releases. Upgrade opendataloader_pdf to the latest version (v2.0.3 or later) using pip: pip install --upgrade opendataloader_pdf.

Question 5

Why do I get an error when using numpy >= 2.0 with docling or hybrid libraries?

Accepted Answer

The project already supports numpy >= 2.0 natively (verified with numpy 2.2.6/2.4.2 in lockfiles). If you encounter an error, it is likely due to a stale installation of transitive dependencies like docling or easyocr that previously pinned numpy < 2. To fix, reinstall from scratch: pip install --upgrade --force-reinstall docling or pip install --no-cache-dir docling[easyocr]. Ensure all packages are updated. If the issue persists, check your resolved dependency versions with pip list and report the traceback, Python version, and installation method.

opendataloader-pdf

Overview

README Preview

FAQ (5)

同类型项目

puppeteer

PaddleOCR

crawl4ai

prisma