How to remove hidden text in pdf file, which created with func show_pdf_page?

[Herrifly]

Hi! I am creating pdf files with important rectangles grouped, but the resulting pdf contains not only my notes, but also hidden text (of other parts of the target pdf) that I can copy and it prevents me from further processing, can I somehow leave only the visible text or create this pdf somehow Any other way? Thanks

[JorjMcKie]

When you use show_pdf_page(), the full original page will be present in the target - even when you use clip. We are considering to offer an option that removes source page parts that are not in clip, but you currently have to "manually" remove it yourself.
Here is a way to do that using redactions. We are creating 4 redaction rectangles representing the source page parts outside clip, execute source_page.apply_redactions() and then use show_pdf_page().

for src_page in source:
    left = src_page.rect
    left.x1 = clip.x0
    if not left.is_empty:
        src_page.add_redact_annot(left)
    right = src_page.rect
    right.x0 = clip.x1
    if not right.is_empty:
        src_page.add_redact_annot(left)
    top = src_page.rect
    top.y1 = clip.y0
    if not top.is_empty:
        src_page.add_redact_annot(left)
    btm = src_page.rect
    btm.y0 = clip.y1
    if not btm.is_empty:
        src_page.add_redact_annot(left)
    src_page.apply_redactions()  # optionally select what to remove

    tar_page.show_pdf_page(rect, src, src_page.number, clip=clip)

[Herrifly]

Ok, I'll try it, thanks

[Herrifly]

Your advice helped to overcome the problem, thank you very much.

[Herrifly]

Hi, I noticed that apply_reductions eats up a lot of memory, is there a way to beat this? This is especially evident when processing 200-300+ pages.

[JorjMcKie]

There are a few ways to influence what is happening in this method.

1. Especially when doing this for many pages, execute pymupdf.Tools.store_shrink(100) after each page to empty MuPDF's internal cache.
2. Consider whether you really need to punch holes in images where they overlap some redaction rectangle. Either ignoring images or completely deleting them is a huge saving in both, execution time and memory consumption. If an image is being processed for redactions, it always is replaced by a new image in PNG format with one or more white areas wherever there was an overlap.
3. Ver important also is to execute page.apply_redactions() only once per page! I saw many cases, where this was done immediately after each inserted redaction.

[Herrifly]

Using your advice, the amount of memory used has decreased tenfold! Thank you so much, I've been thinking about how to solve it all day)

[JorjMcKie]

Glad it helped! Did you implement all 3 recommendations?

[Herrifly]

First + second recommendations. Third I have already used