Skip to content

Implementing total neutron scattering data reduction using the Mantid Framework

License

Notifications You must be signed in to change notification settings

neutrons/mantid_total_scattering

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Health Main Release (w/ Mantid Framework) Dev Release (w/o Mantid Framework)
Build Status Anaconda-Server Badge Anaconda-Server Badge
codecov Anaconda-Server Badge Anaconda-Server Badge
PyPI version

Total Scattering Data Reduction using Mantid Framework

This project is trying to implement total scattering data reduction for neutron time-of-flight diffractometers using the algorithms currently available in the Mantid framework

This entails taking raw neutron counts from detectors in the diffraction experiment and turning them into the reciprocal-space structure factor patterns, F(Q) or S(Q), and applying a Fourier Transform to real-space to give the pair distribution fuction, PDF.

This is the future backend for the ADDIE project and hopes to support multiple diffractometers performing total scattering measurements.

Structure factor S(Q) -> Pair Distribution Function G(r)

alt text

Running mantidtotalscattering will generate the total scattering data in reciprocal space saved as NeXus file, if one is using the multiple bank mode. Another alternative mode is the single bank mode which will merge spectra from all detectors into a single pattern. The single bank mode will be taken in the auto reduction implementation since otherwise manual efforts are needed to merge the different banks data. When using the single bank mode, no Bragg data will be saved. With the multiple bank mode, both the total scattering data and the Bragg diffraction data will be saved. For the Bragg data output, there are two available formats -- the GSS format and the XYE format. The GSS data file could be loaded into the ADDIE interface (see the link above) for visualization. Here, it is noteworthy that if one is loading in the GSS data in Mantid, one has to rebin the loaded in workspace first (since the output GSS data is ragged, i.e., different banks are not sharing the common x-axis, for the purpose of removing the non-sense data in each bank), followed by running the Mantid algorithm ConvertToPointData.

Installation

Pre-requiste: Installing conda

The following commands can get you setup (on Linux machine) to get conda installed (as miniconda):

CONDA_PYTHON=3
MINICONDA_URL="https://repo.continuum.io/miniconda";
MINICONDA_FILE="Miniconda${CONDA_PYTHON}-latest-Linux-x86_64.sh";
wget "${MINICONDA_URL}/${MINICONDA_FILE}";
bash ${MINICONDA_FILE} -b -p $HOME/miniconda;
export PATH="$HOME/miniconda/bin:$PATH";

NOTE: You can change the python version number via the CONDA_PYTHON variable

You will have to excute the last command on every new bash session (export PATH...). Adding this last line to your ~/.bashrc will automatically add it on every bash session startup.

Mantid Framework Included

Anaconda (Recommended)

Setup (w/ pure conda)

Add channels with dependencies, create a conda environment with python_version set to either 2.7.14 or 3.6, and activate the environment

conda config --add channels conda-forge --add channels mantid --add channels mantid/label/nightly
conda create -n mantidts_env python=${python_version}
source activate mantidts_env

Setup (w/ conda + mamba)

NOTE: Mamba is still in "beta".

python_version=3.6
conda config --add channels conda-forge --add channels mantid --add channels mantid/label/nightly
conda install mamba -c conda-forge
mamba update mamba -c conda-forge
mamba create -n mantidts_env python=${python_version}
source activate mantidts_env

Simply replace conda -> mamba in "Install" instruction commands

Install (or Update)

conda install -c neutrons mantid-total-scattering

Go here for how to delete an environment or use:

conda remove --name mantidts_env --all

Notes

If you have an error (see below for example) related to the libGL library, you may not have it installed for the Mantid Framework to work. See instructions here for installing the necessary libraries for different OS

Example error: ImportError: First import of "._api" failed with "libGL.so.1: cannot open shared object file...

If you have an error that another version of Mantid is installed on the machine and being imported via PYTHONPATH, you can use the following as a workaround for CLI tool:

PYTHONPATH="" mantidtotalscattering

Usage (CLI reduction tool)

To launch the total scattering script, complete the input JSON file (found in examples directory), and run:

mantidtotalscattering examples/sns/nomad_simple.json

If you need to specify the path to Mantid build, use:

MANTIDPATH=/path/to/mantid/build/bin PATH=$MANTIDPATH:$PATH PYTHONPATH=$MANTIDPATH:$PATH mantidtotalscattering <json input>

Mantid Framework Not Included (for development)

This is mainly for development if you want to use a local development build of Mantid Framework instead of one included.

PyPI (Recommended)

Install

pip install mantid-total-scattering

Anaconda

Setup

Add channels with dependencies, create a conda environment with python_version set to either 2.7 or 3.6, and activate the environment

conda config --add channels conda-forge
conda create -n mantidts_env python=${python_version}
conda activate mantidts_env

Install

conda install -c neutrons mantid-total-scattering-python-wrapper

Development

Clone the repository to a local directory

git clone https://github.com/neutrons/mantid_total_scattering.git
cd mantid_total_scattering

Located in the main directory of the repo, we have several ways to do the local testing for mantidtotalscattering using a selected version of mantid build. The new way of Mantid building has been changed to use conda, and detailed information can be found here, https://developer.mantidproject.org/GettingStarted/GettingStartedCondaLinux.html#gettingstartedcondalinux. Suppose we are following exactly the instruction in the link above to build Mantid, we will have a conda environment mantid-developer, in which case we can execute the command below to use the local Mantid build to work with the local version of mantidtotalscattering,

conda activate mantid-developer
python MANTID_REPO_DIR/build/bin/AddPythonPath.py
python total_scattering/cli.py INPUT_JSON_FILE

where MANTID_REPO_DIR refers to the full path of the mantid repo directory.

The second way we can try is to launch mantidtotalscattering from within mantidworkbench. To do this, again, assuming we are located in the main mantidtotalscattering directory, we can execute,

MANTID_REPO_DIR/build/bin/launch_mantidworkbench.sh

to start up the local version of mantidworkbench. Then we can load the following script into mantidworkbench and execute it,

import json
from total_scattering.reduction import TotalScatteringReduction
from total_scattering.reduction import validateConfig

with open("FULL_PATH_TO_INPUT_JSON_FILE", "r") as handle:
    config = json.load(handle)

# validate the config
validateConfig(config)

# Run total scattering reduction
TotalScatteringReduction(config)

where we need to replace FULL_PATH_TO_INPUT_JSON_FILE with the full path to our input json file.

N. B. To build a local version of mantid framework, we can check out this link, https://developer.mantidproject.org/GettingStarted/GettingStartedCondaLinux.html#gettingstartedcondalinux.

The third way is to set up a local virtual environment for mantidtotalscattering and add in the python path of local Mantid build. To do this, follow the steps below,

  1. virtualenv -p MANTID-DEVELOPER_ENV_LOCATION/bin/python --system-site-packages .venv

  2. source .venv/bin/activate

  3. python MANTID_REPO_DIR/build/bin/AddPythonPath.py

  4. pip install -r requirements.txt -r requirements-dev.txt

  5. python setup.py develop

where MANTID_REPO_DIR refers to the full path of the mantid repo directory.

N.B. With the third way of setup, one could then execute python SCRIPT_NAME.py to run mantidtotalscattering reduction. The script can be with the contents as shared above. Meanwhile, one can also execute mantidpython SCRIPT_NAME.py if on ORNL analysis cluster. One interesting observation is if using the former way (i.e., run python SCRIPT_NAME.py), the SavePlot1D algorithm cannot take the OutputType as being plotly-all, whereas using the latter command will be able to do that.

Tests

To build and run the tests via pytest, use:

python setup.py test

N. B. This is assuming that the mantid-developer conda environment mentioned above is active.

To build and run tests via Docker, use:

docker build -t unit-test-env -f .ci/Dockerfile.nightly_ubuntu16.04_python3 . && docker run -t unit-test-env /bin/bash -c "mantidpython -m pytest"

Tagging a New Version

Mantid Total Scattering uses versioneer. These are the instructions to create a new version, working on a local clone,

git branch --track main origin/main  #  create a local main branch set to follow remote main
git checkout main
git fetch -p -t  # fetch all changes from the remote repo
git rebase -v origin/main  # sync with remote main branch
git merge --ff-only origin/qa  # merge in all of the changes in branch next
git tag v1.0.7  # create the tag in the format that versioneer has been configured
git push origin v1.0.7  # push the tag to remote to kick off the deploy step

The steps above will tag a new main release (the production release) and will kick off the pipeline action which will further trigger the connected GitLab action to deploy the built conda package to SNS analysis cluster (to the central mantidtotalscattering conda environment located at /opt/anaconda/envs). If we want to tag a release candidate, follow the steps below,

git branch --track qa origin/qa  #  create a local main branch set to follow remote main
git checkout qa
git fetch -p -t  # fetch all changes from the remote repo
git rebase -v origin/qa  # sync with remote main branch
git merge --ff-only origin/next  # merge in all of the changes in branch next
git tag v1.0.8rc1  # create the tag in the format that versioneer has been configured
git push origin v1.0.8rc1  # push the tag to remote to kick off the deploy step

This will kick off the pipeline action which will further trigger the connected GitLab action to deploy the built conda package to SNS analysis cluster (to the central mantidtotalscattering-qa conda environment located at /opt/anaconda/envs).