Files

417 lines
18 KiB
Python

from dataclasses import dataclass, field
from io import BytesIO
from pathlib import Path
from html import escape
import re
class ImportFailure(ValueError):
pass
@dataclass
class ImportResult:
markdown: str
warnings: list[str] = field(default_factory=list)
body_format: str = "markdown"
@dataclass
class ImportedImage:
name: str
content: bytes
content_type: str
cell: str = ""
@dataclass
class ExcelSheetResult:
title: str
markdown: str
images: list[ImportedImage] = field(default_factory=list)
body_format: str = "markdown"
source_index: int = 0
def import_document(upload, flatten_excel=False) -> ImportResult:
suffix = Path(upload.name).suffix.lower()
content = upload.read()
upload.seek(0)
if suffix == ".docx":
return _docx(content)
if suffix in {".xlsx", ".xlsm"}:
return _xlsx(content, flatten=flatten_excel)
if suffix == ".pdf":
return _pdf(content)
if suffix in {".md", ".txt"}:
return ImportResult(content.decode("utf-8-sig"))
if suffix in {".doc", ".xls"}:
raise ImportFailure("Alte .doc/.xls-Dateien bitte zuerst als .docx/.xlsx speichern.")
raise ImportFailure("Unterstützt werden DOCX, XLSX, XLSM, PDF, Markdown und Text.")
def _docx(content):
import mammoth
result = mammoth.convert_to_markdown(BytesIO(content))
warnings = [message.message for message in result.messages]
return ImportResult(result.value.strip(), warnings)
def _render_excel_rows(rows, flatten=False):
from tabulate import tabulate
width = max(len(row) for row in rows)
rows = [row + [""] * (width - len(row)) for row in rows]
header, body = rows[0], rows[1:]
if not flatten:
return tabulate(body, headers=header, tablefmt="github")
# Document-style flattening: retain every cell in reading order without
# inventing record headings or field/value labels. Cells in one Excel row
# become lines in one paragraph; Excel rows are separated by a blank line.
blocks = []
for row in rows:
values = [_escape_markdown_text(value) for value in row if value != ""]
if values:
blocks.append(" \n".join(values))
return "\n\n".join(blocks) or "_Keine Inhalte_"
def _escape_markdown_text(value):
value = str(value).replace("\\", "\\\\")
for character in ("*", "_", "[", "]", "<", ">", "#", "|"):
value = value.replace(character, f"\\{character}")
return value.replace("\r\n", " \n").replace("\r", " \n").replace("\n", " \n")
def _xlsx(content, flatten=False):
from openpyxl import load_workbook
workbook = load_workbook(BytesIO(content), read_only=False, data_only=True)
sections = []
for sheet in workbook.worksheets:
rows = [["" if cell is None else str(cell) for cell in row] for row in sheet.iter_rows(values_only=True)]
while rows and not any(value for value in rows[-1]):
rows.pop()
if not rows:
continue
rendered = _render_flattened_worksheet(sheet, workbook) if flatten else _render_excel_worksheet_html(sheet, workbook)
sections.append(f"<h2>{escape(sheet.title)}</h2>\n{rendered}")
if not sections:
raise ImportFailure("Die Arbeitsmappe enthält keine Daten.")
return ImportResult("\n\n".join(sections), body_format="html")
def import_excel_sheets(
upload, flatten=False, *, metadata_only=False, preview_max_rows=None,
include_image_data=True, selected_indexes=None,
) -> list[ExcelSheetResult]:
"""Convert each non-empty worksheet into an individual document payload."""
from openpyxl import load_workbook
from openpyxl.utils import get_column_letter
suffix = Path(upload.name).suffix.lower()
if suffix not in {".xlsx", ".xlsm"}:
raise ImportFailure("Der Mehrblattimport unterstützt XLSX- und XLSM-Dateien.")
upload.seek(0)
workbook = load_workbook(upload, read_only=False, data_only=True)
results = []
selected_indexes = set(selected_indexes) if selected_indexes is not None else None
for source_index, sheet in enumerate(workbook.worksheets):
if selected_indexes is not None and source_index not in selected_indexes:
continue
images = []
for number, image in enumerate(getattr(sheet, "_images", ()), 1):
image_format = (getattr(image, "format", None) or "png").lower()
if image_format == "jpg":
image_format = "jpeg"
if image_format not in {"png", "jpeg", "gif", "webp"}:
continue
anchor = getattr(image, "anchor", None)
marker = getattr(anchor, "_from", None)
cell = f"{get_column_letter(marker.col + 1)}{marker.row + 1}" if marker else ""
image_content = b""
if include_image_data and not metadata_only:
try:
image_content = image._data()
except (AttributeError, OSError, ValueError):
continue
images.append(ImportedImage(
name=f"{slugify_filename(sheet.title)}-{number}.{image_format if image_format != 'jpeg' else 'jpg'}",
content=image_content, content_type=f"image/{image_format}", cell=cell,
))
has_content = _worksheet_has_content(sheet)
if not has_content and not images:
continue
if metadata_only:
markdown = ""
elif has_content:
markdown = (
_render_flattened_worksheet(sheet, workbook, max_rows=preview_max_rows) if flatten else
_render_excel_worksheet_html(sheet, workbook, max_rows=preview_max_rows)
)
else:
markdown = ""
results.append(ExcelSheetResult(
title=sheet.title, markdown=markdown, images=images,
body_format="html", source_index=source_index,
))
workbook.close()
upload.seek(0)
if not results:
raise ImportFailure("Die Arbeitsmappe enthält keine Daten oder unterstützten Bilder.")
return results
def _worksheet_has_content(sheet):
"""Check for a value without materializing the complete worksheet in memory."""
return any(
cell.value not in (None, "")
for row in sheet.iter_rows()
for cell in row
)
def slugify_filename(value):
value = re.sub(r"[^A-Za-z0-9._-]+", "-", value).strip("-.")
return value[:80] or "arbeitsblatt"
def _excel_color(color, workbook):
"""Resolve RGB, indexed and theme-based openpyxl colors to a CSS hex value."""
if not color or not getattr(color, "type", None):
return None
value = None
if color.type == "rgb" and color.rgb:
value = str(color.rgb)[-6:]
elif color.type == "indexed" and color.indexed is not None:
from openpyxl.styles.colors import COLOR_INDEX
index = int(color.indexed)
if 0 <= index < len(COLOR_INDEX):
value = COLOR_INDEX[index][-6:]
elif color.type == "theme" and color.theme is not None and workbook.loaded_theme:
from xml.etree import ElementTree
try:
root = ElementTree.fromstring(workbook.loaded_theme)
scheme = root.find(".//{http://schemas.openxmlformats.org/drawingml/2006/main}clrScheme")
entries = list(scheme) if scheme is not None else []
entry = entries[int(color.theme)]
color_node = next(iter(entry))
value = color_node.attrib.get("val") or color_node.attrib.get("lastClr")
except (ElementTree.ParseError, IndexError, StopIteration, TypeError, ValueError):
value = None
if not value or not re.fullmatch(r"[0-9A-Fa-f]{6}", value):
return None
rgb = [int(value[index:index + 2], 16) for index in (0, 2, 4)]
tint = float(getattr(color, "tint", 0) or 0)
if tint:
rgb = [round(component * (1 + tint) if tint < 0 else component + (255 - component) * tint) for component in rgb]
return "#" + "".join(f"{max(0, min(255, component)):02x}" for component in rgb)
def _excel_cell_style(cell, workbook):
declarations = []
if cell.fill and cell.fill.fill_type:
fill = (
_excel_color(getattr(cell.fill, "fgColor", None), workbook) or
_excel_color(getattr(cell.fill, "bgColor", None), workbook)
)
if not fill and cell.fill.fill_type == "linear":
stops = getattr(cell.fill, "stop", ())
fill = _excel_color(stops[0].color, workbook) if stops else None
if fill:
declarations.append(f"background-color: {fill}")
font_color = _excel_color(cell.font.color, workbook)
if font_color:
declarations.append(f"color: {font_color}")
if cell.font.bold:
declarations.append("font-weight: bold")
if cell.font.italic:
declarations.append("font-style: italic")
if cell.font.sz:
declarations.append(f"font-size: {float(cell.font.sz):g}pt")
if cell.alignment.horizontal in {"left", "center", "right", "justify"}:
declarations.append(f"text-align: {cell.alignment.horizontal}")
if cell.alignment.vertical in {"top", "center", "bottom"}:
declarations.append(f"vertical-align: {'middle' if cell.alignment.vertical == 'center' else cell.alignment.vertical}")
if cell.alignment.wrap_text:
declarations.append("white-space: normal")
return "; ".join(declarations)
def _tint_hex(value, tint):
import colorsys
rgb = tuple(int(value[index:index + 2], 16) / 255 for index in (1, 3, 5))
hue, lightness, saturation = colorsys.rgb_to_hls(*rgb)
lightness = lightness * (1 + tint) if tint < 0 else lightness + (1 - lightness) * tint
tinted = colorsys.hls_to_rgb(hue, max(0, min(1, lightness)), saturation)
return "#" + "".join(f"{round(component * 255):02x}" for component in tinted)
def _excel_table_cell_styles(sheet, workbook, max_row, max_column):
"""Create visual fallbacks for Excel's built-in 'Format as Table' styles."""
from openpyxl.styles import Color
from openpyxl.utils.cell import range_boundaries
styles = {}
for table in sheet.tables.values():
info = table.tableStyleInfo
match = re.fullmatch(r"TableStyle(Light|Medium|Dark)(\d+)", info.name or "") if info else None
if not match:
continue
family, number = match.group(1), int(match.group(2))
palette_slot = (number - 1) % 7
theme_index = 0 if palette_slot == 0 else 3 + palette_slot
base = _excel_color(Color(theme=theme_index), workbook) or "#4472c4"
min_col, min_row, table_max_col, table_max_row = range_boundaries(table.ref)
table_max_col, table_max_row = min(table_max_col, max_column), min(table_max_row, max_row)
if min_col > table_max_col or min_row > table_max_row:
continue
header_fill = _tint_hex(base, 0.55) if family == "Light" else base
header_color = "#000000" if family == "Light" else "#ffffff"
stripe_fill = _tint_hex(base, 0.88 if family == "Light" else 0.82)
for column in range(min_col, table_max_col + 1):
styles[(min_row, column)] = f"background-color: {header_fill}; color: {header_color}; font-weight: bold"
data_start = min_row + (1 if table.headerRowCount else 0)
data_end = table_max_row - (1 if table.totalsRowShown else 0)
if info.showRowStripes:
for row in range(data_start, data_end + 1):
if (row - data_start) % 2 == 1:
for column in range(min_col, table_max_col + 1):
styles[(row, column)] = f"background-color: {stripe_fill}"
if table.totalsRowShown and table_max_row >= min_row:
for column in range(min_col, table_max_col + 1):
styles[(table_max_row, column)] = f"font-weight: bold; border-top: 2px solid {base}"
return styles
def _render_excel_worksheet_html(sheet, workbook, max_rows=None, bounds=None):
"""Render worksheet values and the most relevant visual Excel cell formatting."""
from openpyxl.utils import get_column_letter
if bounds:
start_column, start_row, bound_max_column, bound_max_row = bounds
else:
start_column, start_row, bound_max_column, bound_max_row = 1, 1, sheet.max_column, sheet.max_row
last_row = last_column = 0
for row in sheet.iter_rows(
min_row=start_row, max_row=bound_max_row,
min_col=start_column, max_col=bound_max_column,
):
for cell in row:
if cell.value not in (None, ""):
last_row = max(last_row, cell.row)
last_column = max(last_column, cell.column)
if not last_row:
return "<p><em>Keine Inhalte</em></p>"
max_row = min(last_row, start_row + max_rows - 1) if max_rows else last_row
max_column = last_column
table_styles = _excel_table_cell_styles(sheet, workbook, max_row, max_column)
merged_starts, merged_children = {}, set()
for merged in sheet.merged_cells.ranges:
if merged.min_row > max_row or merged.min_col > max_column:
continue
merged_starts[(merged.min_row, merged.min_col)] = (
min(merged.max_row, max_row) - merged.min_row + 1,
min(merged.max_col, max_column) - merged.min_col + 1,
)
for row in range(merged.min_row, min(merged.max_row, max_row) + 1):
for column in range(merged.min_col, min(merged.max_col, max_column) + 1):
if (row, column) != (merged.min_row, merged.min_col):
merged_children.add((row, column))
columns = []
for column in range(start_column, max_column + 1):
width = sheet.column_dimensions[get_column_letter(column)].width
columns.append(f'<col style="width: {max(3, min(float(width or 13), 80)):.2f}ch">')
output = ['<table class="doc-table-bordered"><colgroup>', *columns, "</colgroup><tbody>"]
for row in range(start_row, max_row + 1):
row_style = ""
if sheet.row_dimensions[row].height:
row_style = f' style="height: {float(sheet.row_dimensions[row].height):g}pt"'
output.append(f"<tr{row_style}>")
for column in range(start_column, max_column + 1):
if (row, column) in merged_children:
continue
cell = sheet.cell(row=row, column=column)
attributes = []
rowspan, colspan = merged_starts.get((row, column), (1, 1))
if rowspan > 1:
attributes.append(f'rowspan="{rowspan}"')
if colspan > 1:
attributes.append(f'colspan="{colspan}"')
style = _excel_cell_style(cell, workbook)
if not style and (row, column) in table_styles:
style = table_styles[(row, column)]
elif (row, column) in table_styles and "background-color" not in style:
style = "; ".join(filter(None, (style, table_styles[(row, column)])))
if style:
attributes.append(f'style="{style}"')
value = "" if cell.value is None else escape(str(cell.value)).replace("\n", "<br>")
output.append(f"<td{' ' if attributes else ''}{' '.join(attributes)}>{value}</td>")
output.append("</tr>")
output.append("</tbody></table>")
if max_rows and last_row > start_row + max_rows - 1:
output.append(f"<p><em>Vorschau auf {max_rows} Zeilen begrenzt.</em></p>")
return "".join(output)
def _render_flattened_worksheet(sheet, workbook, max_rows=None):
"""Flatten ordinary cells while preserving explicitly defined Excel tables."""
from openpyxl.utils.cell import range_boundaries
table_ranges = []
for table in sheet.tables.values():
min_col, min_row, max_col, max_row = range_boundaries(table.ref)
table_ranges.append((min_col, min_row, max_col, max_row))
table_ranges.sort(key=lambda item: (item[1], item[0]))
def containing_table(row, column):
for bounds in table_ranges:
min_col, min_row, max_col, max_row = bounds
if min_row <= row <= max_row and min_col <= column <= max_col:
return bounds
return None
blocks = []
rendered_tables = set()
rendered_max_row = min(sheet.max_row, max_rows) if max_rows else sheet.max_row
for row_number in range(1, rendered_max_row + 1):
ordinary_values = []
tables_starting_here = []
for column_number in range(1, sheet.max_column + 1):
bounds = containing_table(row_number, column_number)
if bounds:
if bounds[1] == row_number and bounds not in rendered_tables:
tables_starting_here.append(bounds)
rendered_tables.add(bounds)
continue
cell = sheet.cell(row=row_number, column=column_number)
value = cell.value
if value not in (None, ""):
rendered_value = escape(str(value)).replace("\n", "<br>")
style = _excel_cell_style(cell, workbook) if cell.has_style else ""
if style:
rendered_value = f'<span style="{style}">{rendered_value}</span>'
ordinary_values.append(rendered_value)
if ordinary_values:
blocks.append("<p>" + "<br>".join(ordinary_values) + "</p>")
for min_col, min_row, max_col, max_row in tables_starting_here:
blocks.append(_render_excel_worksheet_html(
sheet, workbook,
bounds=(min_col, min_row, max_col, min(max_row, rendered_max_row)),
))
if max_rows and sheet.max_row > max_rows:
blocks.append(f"<p><em>Vorschau auf {max_rows} von {sheet.max_row} Zeilen begrenzt.</em></p>")
return "\n".join(blocks) or "<p><em>Keine Inhalte</em></p>"
def _pdf(content):
from pypdf import PdfReader
reader = PdfReader(BytesIO(content))
pages = []
for number, page in enumerate(reader.pages, 1):
text = (page.extract_text() or "").strip()
if text:
text = re.sub(r"[ \t]+\n", "\n", text)
pages.append(f"## Seite {number}\n\n{text}")
if not pages:
raise ImportFailure("Das PDF enthält keinen extrahierbaren Text. Für Scans ist OCR erforderlich.")
return ImportResult("\n\n".join(pages), ["PDF-Layout und Bilder können nicht vollständig übernommen werden."])