Add support for editing markup in PDFs

[widdowquinn]

PDFs also support editing markup such as highlighting, strikethroughs, and text insertion. These could also be captured and would be useful, for example, when providing lists of corrections for large documents such as PhD theses.

(the code below assumes that the offset suggested in #5 is also implemented)

SEVERITY_NAMES = {0: "Minor comments", 1: "Major comments", 2: "Edits"}

# Edits to be reported
EDITS = ("Cross-Out", "Inserted Text")

def iter_edit_contents(page: PageObject) -> Iterator[str]:
    try:
        edit_indirects = page["/Annots"]
    except KeyError:
        return

    for edit_indirect in edit_indirects:
        edit = edit_indirect.getObject()

        try:
            if edit["/Subj"] in EDITS:
                if edit["/Subj"] == "Cross-Out":
                    yield edit["/Subj"], "-"
                else:
                    yield edit["/Subj"], edit["/Contents"]
        except KeyError:
            continue


def load_comments(filename: str, offset: int) -> SeverityDict:
    res: SeverityDict = defaultdict(list)

    reader = PdfFileReader(filename, STRICT)
    for page_num, page in enumerate(reader.pages, 1):
        for contents in iter_annot_contents(page):
            m_stars = re_stars.match(contents)
            assert m_stars is not None  # should always match

            stars = m_stars["stars"]
            comment = m_stars["comment"]

            # number of stars
            severity = len(stars)

            res[severity].append(f"p{page_num - offset}: {comment}")
        for edit_type, edit in iter_edit_contents(page):
            res[2].append(f"p{page_num - offset}: {edit_type} ({edit})")

    return res

[michaelmhoffman]

Something like this could work. "Edits" aren't really a severity level though, and certainly shouldn't have higher severity than "Major comments". It would be better to leave them at level 0—if a user wants non-edits to have a different severity level they can add extra asterisks.

For insertions, the comment should be Insert "text"

I take it that one cannot easily identify the text corresponding to a strikethrough? In a long-forgotten version of pdfcomments I had [ink] as the comment text to remind myself to look there for comments scribbled with a stylus. You could use [deletion] for strikeout annotations.

A test case for the system added in #7 would be necessary.

[michaelmhoffman]

PyPDF2 documentation now has extensive examples for how to deal with various kinds of annotations https://pypdf2.readthedocs.io/en/3.0.0/user/reading-pdf-annotations.html