Question or problem about Python programming:
I’d like to search a Word 2007 file (.docx) for a text string, e.g., “some special phrase” that could/would be found from a search within Word.
Is there a way from Python to see the text? I have no interest in formatting – I just want to classify documents as having or not having “some special phrase”.
How to solve the problem:
More exactly, a .docx document is a Zip archive in OpenXML format: you have first to uncompress it.
I downloaded a sample (Google: some search term filetype:docx) and after unzipping I found some folders. The word folder contains the document itself, in file document.xml.
After reading your post above, I made a 100% native Python docx module to solve this specific problem.
# Import the module from docx import * # Open the .docx file document = opendocx('A document.docx') # Search returns true if found search(document,'your search string')
The docx module is at https://python-docx.readthedocs.org/en/latest/
In this example, “Course Outline.docx” is a Word 2007 document, which does contain the word “Windows”, and does not contain the phrase “random other string”.
>>> import zipfile >>> z = zipfile.ZipFile("Course Outline.docx") >>> "Windows" in z.read("word/document.xml") True >>> "random other string" in z.read("word/document.xml") False >>> z.close()
Basically, you just open the docx file (which is a zip archive) using zipfile, and find the content in the ‘document.xml’ file in the ‘word’ folder. If you wanted to be more sophisticated, you could then parse the XML, but if you’re just looking for a phrase (which you know won’t be a tag), then you can just look in the XML for the string.
A problem with searching inside a Word document XML file is that the text can be split into elements at any character. It will certainly be split if formatting is different, for example as in Hello World. But it can be split at any point and that is valid in OOXML. So you will end up dealing with XML like this even if formatting does not change in the middle of the phrase!
You can of course load it into an XML DOM tree (not sure what this will be in Python) and ask to get text only as a string, but you could end up with many other “dead ends” just because the OOXML spec is around 6000 pages long and MS Word can write lots of “stuff” you don’t expect. So you could end up writing your own document processing library.
Or you can try using Aspose.Words.
It is available as .NET and Java products. Both can be used from Python. One via COM Interop another via JPype. See Aspose.Words Programmers Guide, Utilize Aspose.Words in Other Programming Languages (sorry I can’t post a second link, stackoverflow does not let me yet).
A docx is just a zip archive with lots of files inside. Maybe you can look at some of the contents of those files? Other than that you probably have to find a lib that understands the word format so that you can filter out things you’re not interested in.
A second choice would be to interop with word and do the search through it.
You can use
docx2txt to get the text inside the docx, than search in that txt
npm install -g docx2txt docx2txt input.docx # This will print the text to stdout
a docx file is essentially a zip file with an xml inside it.
the xml contains the formatting but it also contains the text.
OLE Automation would probably be the easiest. You have to consider formatting, because the text could look like this in the XML:
Looking for this phrase
There’s no easy way to find that using a simple text scan.
You should be able to use the MSWord ActiveX interface to extract the text to search (or, possibly, do the search). I have no idea how you access ActiveX from Python though.
You may also consider using the library from OpenXMLDeveloper.org