Please note that there is a rewrite of Unoconv called "Unoserver": https://github.com/unoconv/unoserver/
We are running Unoserver successfully in production, and it’s now the recommended solution.
Unoserver does not have all the features of Unoconv, which features it will get depends on a combination of what people want, and if someone wants to implement it.
Until Unoserver has all the major features people need, Unoconv is in bugfix mode, there will be no major changes. Once Unoserver has the major features of Unoconv, Unoconv will become unsupported.
Universal Office Converter (unoconv) is a command line tool to convert any document format that LibreOffice can import to any document format that LibreOffice can export. It makes use of the LibreOffice’s UNO bindings for non-interactive conversion of documents.
For practical reasons we mention LibreOffice, but OpenOffice is supported by unoconv as well.
unoconv can be installed using packages coming from your distribution, or simply by copying the unoconv python script to your system.
If you installed unoconv by hand, make sure you have the required LibreOffice or OpenOffice packages installed. A hard requirement is the UNO python bindings which are often inside a subpackage named libreoffice-pyuno or libobasis4.4-pyuno.
Various sub-packages are needed for specific import or export filters, e.g. XML-based filters require the xsltfilter subpackage, e.g. libobasis4.4-xsltfilter.
Important
|
Neglecting these requirements will cause unoconv to fail with unhelpful and confusing error messages. |
To find a good Python installation to use to run unoconv, do the following:
To find which Python to use to run unoconv, you can try a script I made.
cd /tmp wget -l https://gist.githubusercontent.com/regebro/036da022dc7d5241a0ee97efdf1458eb/raw/1bc0655423d196acd79a5d9fa60d2baada8dd534/find_uno.py python3 find_uno.py
It should list all Pythons that have Libreoffice libraries installed.
unoconv starts its own office instance (if it cannot find an existing listener) that it then uses. There are some challenges to do this correctly, but in general this works fine.
Typically you would convert an ODT document to PDF by running:
unoconv -f pdf some-file.odt
However, you can always start an instance yourself at the default port 2002 (or specify another port with -p/--port) and after use you can tear it down:
unoconv --listener & sleep 20 unoconv -f pdf *.odt unoconv -f doc *.odt unoconv -f html *.odt kill -15 %-
It is also possible to use a listener or LibreOffice instance that accepts connections on another system and use it from unoconv remotely. This way the conversion tasks are performed on a dedicated system instead of on the client system. This works only if you have a shared filesystem mounted at the same location.
Beware that the pyuno python module needs to be compiled with the exact same version of python that you are using to load it. A lot of people that run into problems loading pyuno are actually using a precompiled LibreOffice that they downloaded somewhere and is incompatible with the python version on their system.
To solve this issue, the project’s office suite ships with its own python interpreter located in the 'program' directory, this one should work flawlessly.
The most recent unoconv works around this issue by automatically detecting incompatibilities, and restarting itself using a compatible python (the same one that ships with LibreOffice).
You can influence the automatic detection by setting the UNO_PATH environment variable to point to an alternative LibreOffice installation, e.g.:
UNO_PATH=/opt/libreoffice4.4 unoconv -f pdf some-file.odt
But you can also force another python by using it to execute unoconv, e.g.:
/opt/libreoffice4.4/program/python.bin unoconv -f pdf some-file.odt
or on macOS:
/Applications/LibreOffice.app/Contents/MacOS/python unoconv -f pdf some-file.odt
or on Windows:
C:\Program Files (x86)\LibreOffice 4.4\program\python.exe unoconv -f pdf some-file.odt
Tip
|
If you plan to use unoconv extensively (or in an automated fashion) it is more efficient to use the correct python interpreter directly. Or event put it directly in the Shebang (the first line) of the unoconv script ! |
Since OpenOffice 2.3 you do not need an X display for starting ooffice. However you may need the openoffice.org-headless package from your distribution. Since LibreOffice 2.4 nothing special is needed, running in headless mode does not require X.
For any older OpenOffice releases, remember that ooffice requires an X display, even when using it in headless mode. One solution is to use Xvfb to create a headless X display for ooffice.
LibreOffice 3.6.0.1 or later is required to use unoconv under macOS. This is the first version distributed with an internal python script that works. No version of OpenOffice for macOS (3.4 is the current version) works because the necessary internal files are not included inside the application.
Some people have had difficulties using unoconv through webservices. Here is a list of probable causes and recommendations:
-
Use the latest version of unoconv (or GitHub master branch)
-
Use the most recent stable release of LibreOffice (less memory, more stable, fewer crashes)
-
Use the native LibreOffice python binary to run unoconv
-
Hardcode this native python path in the unoconv script shebang (or ensure PATH is set)
-
Ensure that the user running unoconv has write access to its HOME directory (ensure HOME is set)
-
Test with SELinux in permissive mode
It is recommended to open the unoconv script and modify the very first line to point directly to your installed LibreOffice python binary, so replace this:
#!/usr/bin/env python
with something like this:
#!/opt/libreoffice4.4/program/python
If you encounter problems converting files, it often helps to try again. If you are using a listener, restarting the listener may help as well.
The reason for conversion failures are unclear, and they are not deterministic. unoconv is not the only project to have noticed problems with import and export filters using PyUNO. We assume these are related to internal state or timing issues that under certain conditions fail to correctly work.
If you can reproduce the problem on a specific file, please take the time to open the file in LibreOffice directly and export it to the desired format. If this fails, it needs to be reported to the LibreOffice project directly. If that works, we need to know !
We are looking into this with the LibreOffice developers to:
-
Collaborate closer to find, report and fix unexpected failures
-
Allow end-users to increase debugging and improve reporting to the project
If you encounter a problem with converting documents using unoconv, please consider that this could be caused by a number of things:
-
incomplete LibreOffice installation
-
LibreOffice bug or regression specific to your version/distribution
-
LibreOffice import or export filter issue
-
problem related to stale lock files
-
problem related to the source document
-
problem related to permissions or SELinux
-
problem related to the python UNO bindings
-
problem related to the unoconv python script
It is recommended to follow all of the below steps to pinpoint the problem:
-
if this is the first time you are using LibreOffice/OpenOffice, make sure you have all the required sub-packages installed, depending on the distribution this could be the xsltfilter, headless, writer, calc, impress or draw sub-packages.
-
check if there is no existing LibreOffice process running on the system that could interfere with proper functioning
# pgrep -l 'office|writer|calc'
-
check that there are no stale lock files present, e.g. '.~lock.file.pdf#' or '.~lock.index.html#'
-
check that the LibreOffice instance handling UNO requests is not handling multiple requests at the same time
-
try using the latest unoconv release, or the latest version on Github at: https://github.com/dagwieers/unoconv/downloads
-
try the conversion by opening the file in LibreOffice and exporting it through LibreOffice directly
-
try unoconv with a different minor or major LibreOffice version to test whether it is a regression in LibreOffice
-
try to load the UNO bindings in python manually:
-
do this with the python executable that ships with the LibreOffice package/installer
# /opt/libreoffice4.4/program/python.bin -c 'import uno, unohelper'
-
or alternatively, run the distribution python (with the distribution LibreOffice)
# python -c 'import uno, unohelper'
-
-
try unoconv with a different python interpreter manually:
# /opt/libreoffice4.4/program/python.bin unoconv -f pdf test-file.odt
If you tried all of the above, and the issue still remains, the issue might still be related to import/export filters, LibreOffice or unoconv, so please report any information to reproduce the problem on the Github issue-tracker at: https://github.com/dagwieers/unoconv/issues
And do mention that you already tried the above hints to troubleshoot the issue.
If you’re interested to help out with development, here are some pointers to interesting sources:
-
[Tutorial] Import uno module to a different Python install http://user.services.openoffice.org/en/forum/viewtopic.php?f=45&t=36370&p=166783
-
UDK: UNO Development Kit http://udk.openoffice.org/
-
Python-UNO bridge http://www.openoffice.org/udk/python/python-bridge.html
-
Python and OpenOffice.org http://wiki.services.openoffice.org/wiki/Python
-
OpenOffice.org developer manual http://api.openoffice.org/DevelopersGuide/DevelopersGuide.html
-
Framework/Article/Filter/FilterList OOo 2 1 http://wiki.services.openoffice.org/wiki/Framework/Article/Filter/FilterList_OOo_2_1
-
Framework/Article/Filter/FilterList OOo 3 0 http://wiki.services.openoffice.org/wiki/Framework/Article/Filter/FilterList_OOo_3_0
Other implementations using python and UNO:
-
convwatch http://cgit.freedesktop.org/libreoffice/core/tree/bin/convwatch.py
-
oooconv https://svn.infrae.com/oooconv/trunk/src/oooconv/filters.py
-
officeshots.org http://code.officeshots.org/trac/officeshots/browser/trunk/factory/src/backends/oooserver.py
-
cloudooo http://svn.erp5.org/erp5/trunk/utils/cloudooo.handler/ooo/cloudooo/handler/ooo/
Other tools that are useful or similar in operation:
-
Text based document generation: http://www.methods.co.nz/asciidoc/
-
DocBook to OpenDocument XSLT: http://open.comsultia.com/docbook2odf/
-
Simple (and stupid) converter from OpenDocument Text to plain text: http://stosberg.net/odt2txt/
-
Another python tool to aid in converting files using UNO: http://www.artofsolving.com/files/DocumentConverter.py http://www.artofsolving.com/opensource/pyodconverter