Table cell annotations are misplaced (or crash) when the cell is vertically padded

[jimbofreedman]

Fixes #121.

The multi-line branch of TableCell.get_annotations double-counts a cell's top padding, so annotations inside a cell that is shorter than its row either land on the wrong line or index past the end of annotation_lines.

Cause

height.setter pads short cells with empty lines, and width.setter then records line_width from the padded blocks — so line_width carries one zero-length entry per padding line. get_annotations accumulates line_break_pos over that padded list, which makes the loop variable no a padded line index, and then adds self.vertical_padding on top of it:

line_break_pos = list(accumulate(self.line_width))
...
for no, line_break in enumerate(line_break_pos):
    if a.start <= (line_break + no):
        annotation_lines[no + self.vertical_padding].append(a)

Annotation start offsets reference the pre-padding cell content, so no has to be a content-line index for the + vertical_padding shift to be correct.

valign	effect
top	none — vertical_padding == 0
middle	annotations silently shift onto the bottom padding lines, so text[start:end] returns whitespace or text from the next column
bottom	IndexError: list index out of range as soon as annotations reach the second content line

Minimal reproducer on master (2.7.3):

from inscriptis import get_annotated_text
from inscriptis.model.config import ParserConfig

html = ('<table><tr>'
        '<td style="vertical-align:bottom">a1<br><b>a2</b></td>'
        '<td>b1<br>b2<br>b3<br>b4</td>'
        '</tr></table>')

get_annotated_text(html, ParserConfig(annotation_rules={'b': ['b']}))
# IndexError: list index out of range

See #121 for the valign=middle case and full output.

Fix

Record the pre-padding row count as _content_height in normalize_blocks, and accumulate line_break_pos over only the content slice of line_width. no is then a content-line index 0..rows-1, and annotation_lines[no + top_pad] addresses the output line that actually holds the content. The single-line branch and valign=top are untouched.

Tests

tests/test_table_annotation_vertical_padding.py adds four regression tests (bottom-align crash, middle-align offset shift, annotation on the last content line, padded two-column list). All four fail on master — two with IndexError, two with wrong slices — and pass with the fix.

Locally clean: uv run ruff check, uv run ruff format --check, uv run ty check, uv build, and uv run pytest (76 passed) on each of Python 3.10–3.14.

Relation to #92

As noted in #121, this is not the bug reported in #92 — that snippet's two columns are equal height, so no padding is applied and its output is unchanged by this PR. #92 remains open.