docx (OOXML) to html converter
Convert a docx (OOXML) file to semantic HTML. All of Word formatting nonsense is stripped away and you’re left with a cleanly-formatted version of the content.
>>> from docx2html import convert >>> html = convert('path/to/docx/file')
Running Tests for Development
$ virtualenv path/to/new/virtualenv $ source path/to/new/virtualenv/bin/activate $ cd path/to/workspace $ git clone git://github.com/PolicyStat/docx2html.git $ cd docx2html $ pip install . $ pip install -r test_requirements.txt $ ./run_tests.sh
docx2html is designed to take a docx file and extract the content out and convert that content to html. It does not care about styles or fonts or anything that changes how the content is displayed (with few exceptions). Below is a list of what currently works:
- Nested lists
- List styles (letters, roman numerals, etc.)
- Nested tables
- Converting to smaller formats (for bitmaps and tiffs)
- There is a hook to allow setting the src of the image tag out of context, more on this later
- Simple headings
- Root level lists that are upper case roman numerals get converted to h2 tags
Handling embedded images
docx2html allows you to specify how you would like to handle image uploading. For example, you might be uploading your images to Amazon S3 eg: Note: This documentation sucks, so you might need to read the source.
import os.path from shutil import copyfile from docx2html import convert def handle_image(image_id, relationship_dict): image_path = relationship_dict[image_id] # Now do something to the image. Let's move it somewhere. _, filename = os.path.split(image_path) destination_path = os.path.join('/tmp', filename) copyfile(image_path, destination_path) # Return the `src` attribute to be used in the img tag return 'file://%s' % destination html = convert('path/to/docx/file', image_handler=handle_image)
There are two main naming conventions in the source for docx2html there are build functions, which will return an etree element that represents HTML. And there are get_content functions which return string representations of HTML.
- There was a bug with hyperlinks that had a break tag in them. The document would fail to convert. This issue has been fixed.
- There was a bug with hyperlinks that were missing text. The document would fail to convert. This issue has been fixed.
- If a list had an inconsistency in the ilvls, the content for the inconsistent ilvl would be lost. Now we roll that inconsistent list into the root, no longer losing the content.
- If a list had a numId that was not stored in the numbering dict, then a key error would be thrown. Now if either the numId or the ilvl for a given list tag is invalid it defaults to returning a list type of decimal.
- Sometimes in the OOXML an image will have a height or width of 0. If this happens we are now ignoring the height and width in the OOXML and using the full image instead.
- Added a user facing version
- There was a problem for some lists that would cause missing content if the list id’s were not well behaved. This issue has been addressed.
- Fixed missing content with hyperlinks with more than one run tag and smartTags.
- Certain image types are now being ignored. These include: emf, wmf and svg.
- If the indentation level of a set of lists (with the same list id) were mangled (Starting off with a higher indentation level followed by a lower) then the entire sub list (the list with the lower indentation level) would not be added to the root list. This would result in removing the mangled list from the final output. This issue has been addressed.
- Header detection was relying on case. However it is possible for a lower case version of headers to show up. Those are now handled correctly.
- Added a function to remove tags, in addition stripped ‘sectPr’ tags since they have to do with headers and footers.
- Hyperlinks with no text no longer throw an error
- Fixed a bug with determining the font size with an incomplete styles dict
- Fixed a bug with determining the font size of a paragraph tag
- Added a changelog
- Styles are now stripped from hyperlinks
- jinja2 is now used to render test xml
- Correctly handle tables and paragraphs in lists. Before if there was a table in a list it would break the list into two halves, the half before the table and the half after the table (with the table inbetween them). Now if there is a table or paragraph in a list those elements get rolled into the list.