Architecture

Architecture

There are three main parts, the executable, and IO library, and the core analysis library.

xrf_maps.exe : Main executable

• Dependencies: libxrf_io, libxrf_fit, hdf5, netcdf
• Optional Dependencies: zeromq, qt-charts
• Source:
  • src/core/main.cpp
    • Reads and parses command line arguments and creates an Analysis_Job object which is passed to be processed.

libxrf_io: IO library and general processing logic

• Dependencies: libxrf_fit, hdf5, netcdf
• Optional Dependencies: zeromq, qt-charts
• Sources processing whole dataset in memory:
  • src/core/process_whole
    • Functions to process an Analysis_Job.
    • Each one will load the whole dataset in memory before processing it.
    • Functions to optimize fit parameters and per pixel fitting.
  • src/io/file/hl_file_io
    • High level file io.
    • Logic for checking if analyzed hdf5 files exist to read them instead of raw data.
    • Logic to load MDA files for meta data and netcdf files for spectra
  • src/io/file/mda_io
    • For loading MDA files.
    • All APS scans require an MDA file.
    • For APS step scans, Spectra, Scalers and other meta data are saved in MDA file.
    • For APS fly scans, Scalers and other meta data are saved in MDA file, Spectra is saved per row in NetCDF files.
  • src/io/file/hdf_io
    • Loads Analyzed and raw hdf dataset.
    • There is an APS HDF5 raw spectra dataset that can be generated by converting NetCDF rows into a compressed 3d array in HDF5
    • Loads Confocal XRF datasets from APS 20 beamline.
    • Loads HDF5 electron microscopy FEI EMD datasets
    • Saved Analyzed dataset
  • src/io/file/netcdf_io
    • Loads Netcdf files that save a spectra row
  • src/io/file/csv_io
    • For saving fit_parameters to csv format after optimizing them.
  • src/io/file/aps/aps_fit_params_import
    • For loading aps specific maps_fit_parameters_override.txt files
• Sources for streaming datasets:
  • src/core/process_streaming
    • Functions to process an Analysis_Job using Stream_Blocks
  • src/workflow/source
    • Abstract class to define a stream source.
    • Connects to distributor or sink.
  • src/workflow/distributor
    • Abstract class to define a stream process distributor. Typically thread pool to process the spectra.
    • Connects to Sink
  • src/workflow/sink
    • Abstract class to define a stream sink
  • src/workflow/xrf/spectra_file_source
    • Loads dataset and passes the Spectra by a Stream_Block which contains meta information on which pixel it is
  • src/workflow/xrf/integrated_spectra_source
    • Integrates the detector dataset into one spectra
  • src/workflow/xrf/detector_sum_spectra_source
    • Adds up the spectra for N detectors. So if you have 4 detectors it will sum up all four into one spectra
  • src/workflow/xrf/spectra_net_streamer
    • Class for streaming Stream_Blocks over the network. Spectra and Analyzed Counts.
  • src/io/net/basick_serializer
    • Basic spectra encoder/decoder for sending and receiving sptectra and analyzed counts over the network.

libxrf_fit: Data structures, models, and fitting logic

• Dependencies: src/support folder
• Optional Dependencies: zeromq, qt-charts
• Sources:
  • src/data_struct/analysis_job
    • Stores information for a job to run, dataset directory, number of detectors, files to process, ect
  • src/data_struct/quantification_standard
    • Storing quantification calibration curves
  • src/data_struct/element_quant
  • src/data_struct/element_info
    • Class to hold element information from reference/xrf_library.csv
  • src/data_struct/fit_parameters
    • Class used by optimizers and gauss_model.
    • With mpfit you can put contraints on the min and max values for the fit parameters
  • src/data_struct/fit_element_map
    • Dictionary <string, 2d array> that hold element counts after fitting
  • src/data_struct/params_override
    • Structure that holds maps_fit_parameter_override.txt info
  • src/data_struct/spectra
    • Data structure that hold the spectra, elapsed live time, elapsed real time, input counts, and output counts
    • template class that default to float32 but can be compiled as double
  • src/data_struct/spectra_line
    • One row of spectras
  • src/data_struct/spectra_volume
    • Rows and Col's of spectras
  • src/data_struct/stream_block
    • Used for streaming spectra. Holds info about the current pixel index, width, height of the dataset being loaded
  • src/quantification/models/quantification_model
    • Model used to generate the quantification calibration curve
  • src/fitting/models/base_model
    • Abstract class for all models
  • src/fitting/models/gaussian_model
    • Model used to generate model spectra
  • src/fitting/routines/base_fit_routine
    • Abstract class for all fit routines
  • src/fitting/routines/param_optimized_fit_routine
    • This class is used for integrated datasets, pass only 1 sepctra and optimize fit parameters to it.
  • src/fitting/routines/matrix_optimized_fit_routine
    • Per pixel matrix fitting.
  • src/fitting/routines/roi_fit_routine
    • Per pixel roi fitting.
  • src/fitting/routines/nnls_fit_routine
    • Per pixel non negative least squares fitting.
  • src/fitting/routines/svd_fit_routine
    • Per pixel singluar value decomposition fitting.
  • src/fitting/optimizers/optimizer
    • Abstract class for all optimizers
  • src/fitting/optimizers/mpfit_optimizer
    • Optimize fit parameters with constraints
  • src/fitting/optimizers/lmfit_optimizer
    • Optimize fit parameters ( does not support constraints)
  • src/support
    • Support libraries for MDA, NNLS, PyBind11
  • src/workflow/thread_pool
    • Thread Pool class