packaged ptychography software (frontend + backend)

recipes-tag/nsls2ptycho/build.sh

@@ -0,0 +1,19 @@
+# TODO: git clone submodule
+# TODO: detect CUDA, MPI, and FFTW. See below --- 
+# For CUDA, three issues: 1. If users want to run the GPU version, CUDA
+# driver and toolkit must exist. The latter might be taken care of by conda, 
+# but users must handle the former. 2. Conda has no up-to-date CuPy, and 
+# users are encouraged to pip install a pre-built one based on the toolkit
+# version, which makes the automation a bit more difficult. 3. User may only 
+# need the CPU version, then we don't need to touch anything GPU-related!
+#
+# For MPI (and mpi4py), the problem is to build mpi4py on top of whatever
+# MPI in the system. Conda is unable to do a customized build, and it's 
+# better to install a specified MPI first (either from Conda or by oneself,
+# if, say, cuda-aware MPI is needed), and then pip install mpi4py.
+#
+# Currently this script assumes 1. CuPy is either not needed or installed 
+# separately, and 2. "conda install mpi4py" works. Obviously we need to 
+# polish this when more beamlines need this pacakge...
+
+$PYTHON -m pip install . --no-deps --ignore-installed -vv

recipes-tag/nsls2ptycho/meta.yaml

@@ -0,0 +1,43 @@
+package:
+  name: nsls2ptycho
+  version: "1.0.0"
+
+source:
+  git_rev: packaging
+  git_url: https://github.com/leofang/ptycho_gui.git
+
+build:
+  number: 0
+  skip: True # [not py36 or win]
+
+requirements:
+  build:
+    - python 
+    - cython
+    - pip
+    - fftw >=3
+    - qt >=5
+    - setuptools
+  run:
+    - python
+    - fftw >=3
+    - pyfftw
+    - mpi4py 
+    - qt >=5
+    - pyqt >=5
+    - numpy
+    - scipy
+    - matplotlib
+    - pillow
+    - h5py
+    - posix_ipc
+
+about:
+  home: https://github.com/leofang/ptycho_gui
+  license: MIT
+  license_file: LICENSE
+  summary: NSLS-II in-house ptychography software
+
+extra:
+  recipe-maintainers:
+    - leofang