NB! The information below is up-to-date as of HFST 3.11.
Presently, HFST3 supports three different backends, all enabled by default:
HFST3 abstracts over the different backends, and will make it easy to add support for new backends in the future, to support even more FST source code and transducer conversion options.
Run these commands:
curl http://apertium.projectjj.com/osx/install-nightly.sh > install-nightly.sh chmod a+x install-nightly.sh sudo ./install-nightly.sh
On Linux ubuntu:
wget https://apertium.projectjj.com/apt/install-nightly.sh -O - | sudo bash sudo apt-get -f install apertium-all-dev
On Linux fedora (e.g. on gtlab):
curl https://apertium.projectjj.com/rpm/install-nightly.sh |sudo bash sudo apt-get -f install apertium-all-devel
This downloads a shell script (1), makes it executable (2), and runs it (3). The shell script in turn will download and install prebuilt binaries for:
You get the latest version of all required tools in one go, no compilation required! :) Rerun with regular intervals to get the latest updates.
If you want to configure and compile the source code, do as follows:
We assume installation from svn here. You can downloaded a tarball from the
Hfst Github site. Then
unzip and untar, and go directly to the
svn co https://github.com/hfst/hfst.git/trunk hfst cd hfst/ ./autogen.sh ./configure --enable-all-tools --with-unicode-handler=glib
The configuration above is the default used by the Giella infra. To check what
other options there are, do
To avoid the very wordy output of
make, and get only essential messages,
add the following
The final steps to build and install
hfst after the configuration are as
make sudo make install
NB!! The following info is most likely outdated. See the Hfst readme file for further details.
If you want to run hfst from Python, you need to install the Python binding
provided through SWIG. First you need to install
swig and the Swig submodule
for Python using MacPorts:
sudo port install swig sudo port install swig-python
When done, you are ready to build and install the Python bindings:
python3.3 setup.py build_ext --inplace
Due to a new default compiler on MacOSX 10.9 (with which the hfst code
./configure --without-foma --with-unicode-handler=glib --enable-all-tools \ LDFLAGS=" -stdlib=libstdc++"
./configure --without-foma --with-unicode-handler=glib --enable-all-tools \ CXXFLAGS=" -stdlib=libstdc++" LDFLAGS=" -stdlib=libstdc++"
./configure --without-foma --with-unicode-handler=glib --enable-all-tools \ CXXFLAGS="-I/usr/include/c++/4.2.1" LDFLAGS=" -stdlib=libstdc++"
NB! Remember to
make clean after each
./configure run, and before
NB2! This does NOT apply to MacOSX 10.10 - in this case, just configure without any extra flags, as described earlier. This might be because the hfst build instructions are improved, or because of changes in Apple’s compilers. If it is because of changes in the Hfst code base, the above workaround is probably obsolete also on MacOSX 10.9, but this needs to be tested.
When building from the github source code,
OSX 10.6 and earlier requires newer versions of
than provided by the system, and
configure. Also the
bison versions installed breaks
compilation. Ensure you have MacPorts installed, then do the following
before anything else:
sudo port install automake autoconf flex bison pkgconfig
After this, do
./autogen.sh etc. as above.