v3.02

2022-05-19

André Offringa <of…nga@gmail.com>

Version 3.2 of aoflagger allows concattenating measurement sets in frequency and process these in parallel. One of the use-cases for which this is useful, is the Prefactor/LINC pipeline for processing LOFAR data. Apart from that, a few smaller Lua options were added, and various bugs were squashed and performance improvements were made.

The full list of changes is given below.

New features

Generic improvements

• Faster vertical masked sumthreshold algorithm.

• New Lua functions:

  • aoflagger.upsample_mask()

  • Data.invert_mask()

  • Data.set_masked_visibilities()

• Allow specifying penalty in aoflagger.scale_invariant_rank_operator_masked().

• Add support for Sdhdf files. This adds the requirement to have the hdf5 library installed.

• Add the Apertif 2021-03-09 strategy and set it as default for Apertif.

• Allow downsampling in aoquality query_f.

• Support for Function Multi Versioning, which makes the packaged versions run faster.

• Spectrally concatenate different measurement sets and perform IO in parallel.

New rfigui features

• New ‘open file’ window that allows opening multiple files and setting options.

• Ask confirmation when closing rfigui without saving changes.

• Add time/frequency options to rfigui.

• Add a new simulate-data window to rfigui with more simulations options.

• Add ‘select zoom’ option to rfigui.

Bug fixes & refactoring

• Data with NaNs would cause some of the sumthreshold algorithms to behave incorrectly.

• Various fixes for newer compilers (gcc 11/12) or different environments.

• Fix return value of Data.get_complex_state(), which always returned complex.

• Report clear error when the Lua version is older than 5.3.

• Fix default strategy when exclude_original_flags = false.

• Several plot-drawing improvements.

• Fix loading of ATCA strategy in rfigui.

• Let ATCA strategy work when not all polarizations are available.

• Make Apertif strategy also work when no bandpass is available.

• Remove libxml2 dependency.

• Add CMake option for skipping GTKMM builds.

• Add HINTS to find FFTW3 in cmake.

• Make sure all python examples run.

• Improve performance of calculating median window (time threshold operation).

• Require C++17, use std instead of Boost where possible.