Incorrect extracted text from annotation

[reyjexter]

Hi again! We have an example PDF document which when using for_annotation to get the text, it returns weird characters. However when showing logs of page objects text, this doesn't happen.

Here's the PDF affected:

https://github.com/reyjexter/pdfium-render-wasm/blob/master/www/dynamic-template.pdf

And the example:

https://github.com/reyjexter/pdfium-render-wasm

Here's how the console log on Chrome looks like:

Thanks and appreciate any help with this issue.

[ajrcarey]

Hi @reyjexter , thank you for reporting the issue.

I can reproduce the problem, but only when compiling to WASM; when running the following reproduction snippet from the console, the annotation text is returned correctly. This suggests the problem is in the WASM implementation of FPDFText_GetBoundedText. Checking now.

let pdfium = Pdfium::new(Pdfium::bind_to_library(
    Pdfium::pdfium_platform_library_name_at_path("./"),
)?);

let document = pdfium.load_pdf_from_file("./dynamic-template.pdf", None)?;

for page in document.pages().iter() {
    let text = page.text()?;

    for annotation in page.annotations().iter() {
        println!(
            "Text: {:?}, bounds: {:?}",
            text.for_annotation(&annotation),
            annotation.bounds()?
        );
    }
}

[ajrcarey]

Yes, the WASM implementation of FPDFText_GetBoundedText was the problem; the return value of FPDFText_GetBoundedText is the number of characters (in UTF16-LE format) copied into a byte buffer, not the number of bytes. (This is rare difference from the usual Pdfium pattern, which is to return the number of bytes.) This difference was not accounted for in the WASM implementation of this function. I have pushed a fix for the problem and it will be included in crate release 0.8.0.

There will be a delay before the release of 0.8.0, so in the meantime you may want to set pdfium-render as a git dependency in your project.

[reyjexter]

I can confirm that the change on master works correctly on document I sent. Thanks again and we shouldn't have any issue using pdfium-render as a git repository.

[ajrcarey]

Good to hear, closing as resolved.