3eff76f60 Merge pull request #53 from borglab/feature/refactor 13215dfa7 Merge pull request #52 from borglab/fix/tests 696913b11 install setuptools 9523328ba Merge branch 'master' into fix/tests 7c630b361 some more cleanup 656993a71 cleaned up Typename a16f6f38e move qualified and basis type outside to their own class scope 72ead8bd7 Merge pull request #51 from borglab/fix/test-interface-parser 6deefd4fc added tests for interface_parser 50d490a85 Merge pull request #50 from borglab/feature/refs-all-types be4511290 updated docs for BasisType 0e80b0d8c update MATLAB tests 0015d7397 added support for shared pointer and ref for basis types 86d2158f1 remove std::string from list of Basis types 94f928441 ignore code coverage reports 2033dd345 replace prints with log.debug statements ae98091b3 fix deprecation in doc tests 13a2f66c4 Merge pull request #46 from borglab/feature/new-shared-pointer 3c7d85865 updated docs 6d7897088 use @ for raw pointer, go back to * for shared pointer 1d6194c57 updated matlab wrapper to handle both raw and shared pointers 1448f6924 fix some failing tests 2ab1dae32 Merge branch 'master' into feature/new-shared-pointer 96f8a56bd Merge pull request #47 from borglab/fix/ci 6003203f3 run CI only for pull requests a8f29ead1 fix the python version yml key fcae17227 check if directory exists when testing f592f08c9 explicit pip3 so that we don't use Python2 d49c2f3c2 correct call for pip dfe360526 fix the CI 149b7c523 docs for templated functions f2189acc6 support typedefing for templated functions 965458a2b added test for templated functions eaff6e6ab made is_const common for all types 3d9c70b32 added tests and cleaned up a bit 010b89626 support for simple pointers on basis types 6b98fd80c new syntax for shared_ptr ff7ad0b78 support for templated functions a1a443c8d Merge pull request #43 from borglab/fix/cmake-and-matlab 2f3a055e4 remove accidentally committed file 770d055e2 set proper paths for cmake and eschew relative paths 773d01ae1 fix bug in matlab wrapper 721ef740f Merge pull request #41 from borglab/feature/type-hints 67aac9758 minor refactor of CI yml e6a63ae0c fix all mypy issues a3aaa3e7c remove a lot of linter issues from matlab_wrapper a96db522f static typing for interface_parser git-subtree-dir: wrap git-subtree-split: 3eff76f604b5ba9e71cf4947654e288142ed7a94 |
||
---|---|---|
.github/workflows | ||
cmake | ||
docs | ||
gtwrap | ||
pybind11 | ||
scripts | ||
sphinx | ||
tests | ||
utilities | ||
.gitignore | ||
CMakeLists.txt | ||
DOCS.md | ||
LICENSE | ||
README.md | ||
matlab.h | ||
pybind_wrapper.tpl.example | ||
requirements.txt | ||
setup.py |
README.md
WRAP
The wrap library wraps the GTSAM library into a Python library or MATLAB toolbox.
It was designed to be more general than just wrapping GTSAM. For notes on creating a wrap interface, see gtsam.h
for what features can be wrapped into a toolbox, as well as the current state of the toolbox for GTSAM.
Prerequisites
Pybind11
and pyparsing
- This library uses
pybind11
, which is included as a subdirectory in GTSAM. - The
interface_parser.py
in this library usespyparsing
to parse the interface filegtsam.h
. Please install it first in your current Python environment before attempting the build.
python3 -m pip install pyparsing
Getting Started
Clone this repository to your local machine and perform the standard CMake install:
mkdir build && cd build
cmake ..
make install # use sudo if needed
Using wrap
in your project is straightforward from here. In your CMakeLists.txt
file, you just need to add the following:
find_package(gtwrap)
pybind_wrap(${PROJECT_NAME}_py # target
${PROJECT_SOURCE_DIR}/cpp/${PROJECT_NAME}.h # interface header file
"${PROJECT_NAME}.cpp" # the generated cpp
"${PROJECT_NAME}" # module_name
"${PROJECT_MODULE_NAME}" # top namespace in the cpp file e.g. gtsam
"${ignore}" # ignore classes
${PROJECT_BINARY_DIR}/${PROJECT_NAME}.tpl # the wrapping template file
${PROJECT_NAME} # libs
"${PROJECT_NAME}" # dependencies
ON # use boost
)
For more information, please follow our tutorial.
Documentation
Documentation for wrapping C++ code can be found here.
Python Wrapper
WARNING: On macOS, you have to statically build GTSAM to use the wrapper.
-
Set
GTSAM_BUILD_PYTHON=ON
while configuring the build withcmake
. -
What you can do in the
build
folder:-
Just run python then import GTSAM and play around:
import gtsam gtsam.__dir__()
-
Run the unittests:
python -m unittest discover
-
Edit the unittests in
python/gtsam/*.py
and simply rerun the test. They were symlinked to<build_folder>/gtsam/*.py
to facilitate fast development.python -m unittest gtsam/tests/test_Pose3.py
- NOTE: You might need to re-runcmake ..
if files are deleted or added.
-
-
Do
make install
andcd <gtsam_install_folder>/python
. Here, you can:- Run the unittests:
python setup.py test
- Install
gtsam
to your current Python environment.python setup.py install
- NOTE: It's a good idea to create a virtual environment otherwise it will be installed in your system Python's site-packages.
- Run the unittests:
Matlab Wrapper
In the CMake, simply include the MatlabWrap.cmake
file.
include(MatlabWrap)
This cmake file defines functions for generating MATLAB wrappers.
wrap_and_install_library(interfaceHeader linkLibraries extraIncludeDirs extraMexFlags)
Generates wrap code and compiles the wrapper.
Usage example:
`wrap_and_install_library("lba.h" "" "" "")`
Arguments:
interfaceHeader
: The relative or absolute path to the wrapper interface definition file.linkLibraries
: Any additional libraries to link. Your project library (e.g.lba
), libraries it depends on, and any necessary MATLAB libraries will be linked automatically. So normally, leave this empty.extraIncludeDirs
: Any additional include paths required by dependent libraries that have not already been added by include_directories. Again, normally, leave this empty.extraMexFlags
: Any additional flags to pass to the compiler when building the wrap code. Normally, leave this empty.