c2.transform.ooxml

日本語

Make Office Open XML files (Word, Excel and PowerPoint files such as .docx, .xlsx and .pptx) searchable in Plone.

The add-on registers one transform, ooxml_to_text, in portal_transforms. Plone's search (and any add-on that calls portal_transforms.convertTo("text/plain", ...)) then gets the text of these files. The text is read with Python's zipfile and lxml only: no Java, no LibreOffice, no other service and no other library.

Only the text is extracted; the structure of the document (headings, tables, layout) is not kept.

Formats

Extension	MIME type
.docx	application/vnd.openxmlformats-officedocument.wordprocessingml.document
.dotx	application/vnd.openxmlformats-officedocument.wordprocessingml.template
.docm	application/vnd.ms-word.document.macroEnabled.12
.dotm	application/vnd.ms-word.template.macroEnabled.12
.xlsx	application/vnd.openxmlformats-officedocument.spreadsheetml.sheet
.xltx	application/vnd.openxmlformats-officedocument.spreadsheetml.template
.xlsm	application/vnd.ms-excel.sheet.macroEnabled.12
.xltm	application/vnd.ms-excel.template.macroEnabled.12
.pptx	application/vnd.openxmlformats-officedocument.presentationml.presentation
.potx	application/vnd.openxmlformats-officedocument.presentationml.template
.ppsx	application/vnd.openxmlformats-officedocument.presentationml.slideshow
.pptm	application/vnd.ms-powerpoint.presentation.macroEnabled.12
.potm	application/vnd.ms-powerpoint.template.macroEnabled.12
.ppsm	application/vnd.ms-powerpoint.slideshow.macroEnabled.12

Documents in the Strict form of ISO/IEC 29500 are read as well. Macros are never read or run.

Not supported: .xlsb (a binary format, not XML) and the old binary formats .doc, .xls and .ppt.

What text is extracted

Format	Extracted	Not extracted
Word	The body, tables, headers and footers, footnotes and endnotes, inserted tracked changes	Comments, deleted tracked changes, field codes, SmartArt
Excel	Sheet names, cell values (text and numbers, the saved results of formulas)	Formulas, comments, phonetic readings
PowerPoint	The text of the slides (also in tables and shapes), the speaker notes	Comments, SmartArt

Embedded documents (such as a workbook in a presentation) and the text in images are not extracted.

• Word: the body first, then the headers and footers, the footnotes and the endnotes. Every paragraph ends a line; tabs and line breaks are kept.
• Excel: the sheets in the order of the workbook, hidden sheets too. Each sheet starts with its name on a line, then each row is one line with its cell values separated by tabs. Dates stay numbers, as they are saved.
• PowerPoint: the slides in the order of the presentation, hidden slides too, each followed by its notes. Every paragraph ends a line.
• A line break is never written after another one: empty paragraphs and empty rows give no empty lines.

Limits and errors

A malicious or broken file must not stop or slow down Plone. These limits apply to one document:

Limit	Value
Size of one part after decompression	64 MiB
Size of all the parts read, after decompression	256 MiB
Number of parts read	5,000
Length of the text	10,000,000 characters
Time	30 seconds
Size of the central directory of the zip (about 45,000 entries)	2 MiB

• The time limit usually stops the reading first: at the speed measured below, reading 64 MiB parts up to 256 MiB in all would take about a minute.
• The sizes are counted while the parts are read; the sizes that the zip declares are not trusted.
• The size of the central directory is checked before the zip is opened, as zipfile will read it; it also limits the number of entries.
• Only parts that are stored or compressed with deflate (as Office writes them) are read. A part compressed with another method (such as bzip2, LZMA or Zstandard) is treated as a broken part.
• When a limit is reached, the text read so far is returned and a warning is logged.
• The shared strings of a workbook are kept in memory while its sheets are read; they count toward the length of the text.
• A file that is not a zip (a password-protected document is not one), a broken zip or a broken main part gives no text, with a warning. A broken part after the main part (a header, a sheet, a slide) keeps the text read so far, with a warning. Saving the file in Plone still works.
• The XML is read without entities, DTDs or network access; a part with a DOCTYPE is refused.
• A single run of text over about 10 MB is refused by libxml2 as broken XML: in the main part, the document gives no text; in a later part, the text read so far is kept.
• The log never contains the content of a document.

Measured on 2026-11-08 (Python 3.14, Plone 6.2) with one ordinary 52 MB part of about 35 million characters: with the default limits, the reading stopped at the length of the text after about 3 seconds, with a peak of 29 MB of Python memory; with the limits raised, all of its text was read in about 11 seconds, with a peak of 102 MB. For that document the memory grew with the text, not with the size of the part. These numbers describe that document only; for any input, it is the limits above that bound the reading.

Measured again on 2026-11-09 (Python 3.14, Plone 6.2) with made-up documents, one at a time, in a process that only reads the file and extracts its text. The peak is the maximum resident set size of that process, not only the memory of Python objects as above. The time is that of the extraction only, without starting the process and reading the file:

Document	Peak memory of the process	Extraction time	Stopped by
One body part of about 60 MB, full of text	73.3 MiB	0.7 s	The length of the text
One part of 70 MiB of small empty elements	29.0 MiB	8.6 s	The size of the part (64 MiB)
A short body and five headers of about 60 MB each, without text	28.8 MiB	30.0 s	The time (30 seconds), before the total size (256 MiB)
Shared strings of about 9 million characters, used by 2 million cells	48.8 MiB	0.1 s	The length of the text (the shared strings and the text together)
About 40 MB of comments and processing instructions	27.5 MiB	0.2 s	Nothing
A central directory just under 2 MiB (38,003 entries)	50.7 MiB	0.1 s	Nothing

Installation

Add c2.transform.ooxml to the dependencies of your project, then activate "c2.transform.ooxml" in the Add-ons control panel. The install registers the MIME types if the site does not know them yet, and the transform.

Install either this package or c2.transform.msoffice, not both. Both handle these MIME types, and which one Plone uses would then be hard to predict. Choose c2.transform.msoffice if you also need the old formats (.doc, .xls, .ppt); choose c2.transform.ooxml if Office Open XML is enough and you do not want Java.

Index the existing files

Installing does not index the files that are already in the site (on a large site that could take hours). Index them again afterwards, in one of two ways:

• The whole catalog: in the ZMI, portal_catalog > Advanced > "Clear and Rebuild" (manage_catalogRebuild).

• Only the Office Open XML files, for example with bin/zconsole run <zope.conf> reindex_ooxml.py:

  from c2.transform.ooxml.config import ooxml_mimetypes
  from zope.component.hooks import setSite
  
  import transaction
  
  portal = app.Plone  # the id of your site
  setSite(portal)
  types = [m["mimetypes"][0] for m in ooxml_mimetypes]
  for brain in portal.portal_catalog.unrestrictedSearchResults(mime_type=types):
      brain._unrestrictedGetObject().reindexObject(idxs=["SearchableText"])
  transaction.commit()

Uninstall

Deactivate the add-on in the Add-ons control panel. The transform is removed; the MIME types stay, because other add-ons may use them.

Compatibility

Plone 6.0, 6.1 and 6.2 on Python 3.10 to 3.14 (Plone 6.0 and 6.1 up to Python 3.13).

License

GPL version 2 or later.