Export to markdown of multi-sheet .xlsx file

[Fedezzab]

Question

I recently started using docling to export the content of an Excel (.xlsx) file to markdown (md).

I am using the standard DocumentConverter with the default pipeline for this format. However, I noticed that when processing a file with multiple sheets, the .export_to_markdown() method simply aggregates the content of all sheets into a single markdown output.

While debugging, I observed that the corresponding DoclingDocument does differentiate between sheets. For example, its tables attribute is a list containing as many TableItem objects as there are sheets in the file.

Is there a clean way to export the content of different sheets separately? I found one approach, but it's quite naïve—it involves creating deep copies of the DoclingDocument while restricting its attributes to individual tables, one by one.

I would like to achieve the same result in a more efficient way, similar to how it's possible to export a PDF page by page using the page_no parameter in the method .export_to_markdown().

[ceberam]

Thanks @Fedezzab for your interest in docling library.

Unfortunately, there is not a more elegant solution for what you are trying to achieve at the moment. The alternatives would be:

• Use the page_no parameter in method export_to_markdown, provided that we assign a sheet to a page. This is currently not the case. The sheets are added in a docling document as GroupItem objects with the SECTION label. We will consider creating a page for a sheet, specially since our backend parser (MsExcelDocumentBackend) claims to support pagination.
• Use the new DocSerializer, which allow you to serialize a specific document item to markdown. Check the thread in Apply export_to_markdown to individual document items #962 for more details. The idea would be to iterate over the group items and serialize them individually. However, this type of groups are not addressed by the serializer and you will get an empty text. We will address this as an issue.

  from docling.document_converter import DocumentConverter  
  from docling_core.experimental.serializer.markdown import MarkdownDocSerializer
  from docling_core.types.doc import GroupLabel
  
  converter = DocumentConverter()
  doc = converter.convert("/Path/To/My/Spreadsheet.xlsx").document
  serializer = MarkdownDocSerializer(doc=doc)
  sheet_md = [serializer.serialize(item=item).text for item in doc.groups if item.label == GroupLabel.SECTION]  

[ceberam]

@Fedezzab in the PR #1332 that we just merged, we have added pagination to the Excel backend. You can now use the method .export_to_markdown() and select the desired worksheet number (1-based index) through the page_no parameter.
In addition, we have also added implemented the concepts of page size and bounding boxes for the parsed tables and images, using the cell grid as a metric space.