Repository navigation

Fancyvoigtfit - #76

Merged

sbird merged 67 commits into

masterfrom

fancyvoigtfit

Aug 17, 2026

Merged

Fancyvoigtfit#76
sbird merged 67 commits into
masterfrom
fancyvoigtfit

Commits on Aug 5, 2026
- Remove the thresh argument to get_mean_flux
  sbird
  committed
  dde3a52
- Clean up the threshold for DLA filtering
  sbird
  committed
  3de89bf
- Replace the C mean flux rescaling with threaded numpy
  
  sbird
  and
  claude
  committed
  bba49fc
- Update gitignore
  sbird
  committed
  509e894
- flux_power: get the mean flux from the k=0 Fourier mode
  
  sbird
  and
  claude
  committed
  f0720b6
- flux_power: use scipy's threaded FFT
  
  sbird
  and
  claude
  committed
  6a1ff81
- Test flux_power directly and drop _powerspectrum
  
  sbird
  and
  claude
  committed
  29d6c43
- Don't need the threshold: small ffts are fast anyway
  sbird
  committed
  186c819
- Take the power spectrum bins from scipy too
  
  sbird
  and
  claude
  committed
  fddfcdb
- Share one thread pool, and thread the flux power batches
  
  sbird
  and
  claude
  committed
  db7db55
- flux_pdf: count the histogram in parallel
  
  sbird
  and
  claude
  committed
  40c0334
Commits on Aug 6, 2026
- Ensure that data where F is NaN does not loop in mean flux.
  sbird
  committed
  ab0639d
- Do not change input array in 3D power
  sbird
  committed
  a92f22a
- Mask DLA pixels rather than overwriting them, and rescale around them
  sbird
  committed
  b63e964
- Increase the accuracy of the mean flux calculation
  sbird
  committed
  bd65f44
- Only walk the sightlines which contain a strong absorber
  
  sbird
  and
  claude
  committed
  5736705
- Add a Voigt profile fitter for DLAs
  sbird
  committed
  5fcd9b8
- Estimate HCD amplitudes with a lower envelope rather than a fit
  
  sbird
  and
  claude
  committed
  8861d9a
Commits on Aug 7, 2026
- Blending HCDs are fit badly by the peak estimator: use the integral
  sbird
  committed
  0506fca
- Use np.trapezoid instead of the removed np.trapz
  
  sbird
  and
  claude
  committed
  46a3f4f
- Fix rho_DLA crash and replace_not_DLA array sizing
  
  sbird
  and
  claude
  committed
  61c915c
- Get the species density on the same particles as the ionic density
  
  sbird
  and
  claude
  committed
  8cc9907
- Check for a None cofm before converting it
  
  sbird
  and
  claude
  committed
  4818e2f
- Fix the noise seeds and load the temperature when we need it
  
  sbird
  and
  claude
  committed
  745ed66
- Fix wrong loop index when computing the curvature. Worked by accident before.
  sbird
  committed
  f895795
- Take the particles we want before converting them
  
  sbird
  and
  claude
  committed
  cbd7a35
- Join the segments once instead of appending in a loop
  
  sbird
  and
  claude
  committed
  8ce0bdf
- Stop doubling the memory in get_observer_tau, and preallocate the noise
  
  sbird
  and
  claude
  committed
  1c6e5bf
- Do not smooth the spectra when the resolution is perfect
  
  sbird
  and
  claude
  committed
  8f312e9
- Add more tests of increasingly complex DLA Voigt fits
  sbird
  committed
  f253456
Commits on Aug 8, 2026
- Fit the center of the profile as well as the amplitude.
  sbird
  committed
  cab6e8c
- Clean up iterate_new_spectrum somewhat.
  sbird
  committed
  ec2ce5d
- Test fit_profiles not fit_components
  sbird
  committed
  ebb3bcd
- Remove the instantly converging iteration in fit_components
  sbird
  committed
  53f44ae
- Switch to the new HCD voigt fitter when doing DLA masking
  sbird
  committed
  7223f69
Commits on Aug 9, 2026
- Do not copy the tau array, mutate it in place instead
  sbird
  committed
  86f356b
- Speed up the HCD Voigt fit by a factor of two
  
  sbird
  and
  claude
  committed
  fb2cf80
- Expose masktau to get_flux_power_1D
  sbird
  committed
  bcbf3c4
Commits on Aug 11, 2026
- Add the ability to load only a subset of sightlines from the disc for large sightline files
  sbird
  committed
  d2c7e60
- Compute the mean flux without a double precision copy of the spectra
  
  sbird
  and
  claude
  committed
  b0025ed
Commits on Aug 12, 2026
- Pass the thread pool into the flux statistics functions
  
  sbird
  and
  claude
  committed
  08906bc
Commits on Aug 13, 2026
- Replace the C++ sightline neighbour search with numpy and a KD-tree
  
  sbird
  and
  claude
  committed
  b66aee7
- Use a brute force search as the oracle for the sightline neighbour test
  
  sbird
  and
  claude
  committed
  8af1218
- Remove the C++ _near_lines
  
  sbird
  and
  claude
  committed
  91b50c3
- Convert positions to single precision after selecting the sightline particles
  
  sbird
  and
  claude
  committed
  5bc5cff
- Do the sightline mesh lookup in blocks, on the thread pool
  
  sbird
  and
  claude
  committed
  3a23642
Commits on Aug 14, 2026
- Gather the particles near each sightline without a lock
  
  sbird
  and
  claude
  committed
  688586b
- Remove get_near_lines, and test the search geometry through the interface
  
  sbird
  and
  claude
  committed
  667a62f
- Replace the GSL 2D interpolation with scipy's RectBivariateSpline
  
  sbird
  and
  claude
  committed
  1ccbf38
- Interpolate on the thread pool with scipy.ndimage.map_coordinates
  
  sbird
  and
  claude
  committed
  be21ef6
- Drop python 2 compatibility
  sbird
  committed
  e5c55f8
- Use a portable CPU count instead of os.sched_getaffinity
  
  sbird
  and
  claude
  committed
  0eaa124
- Move the build metadata to pyproject.toml rather than setup.py
  sbird
  committed
  bcf19cd
- Update the pypi classifiers
  sbird
  committed
  b0e8451
- Fix three memory errors in the interpolation
  
  sbird
  and
  claude
  committed
  7f91794
- Use 64 bit particle indices for the sightline neighbour lists
  
  sbird
  and
  claude
  committed
  625d0a8
- Check the shapes of the arrays passed to the interpolation
  
  sbird
  and
  claude
  committed
  f64be2a
Commits on Aug 15, 2026
- Speed up the optical depth computation by an order of magnitude
  
  sbird
  and
  claude
  committed
  2aebe44
Commits on Aug 16, 2026
- Find the particles near a sightline with a mesh rather than a sorted map
  
  sbird
  and
  claude
  committed
  c85c9fc
- Change absorption.cpp to do computations in double precision
  sbird
  committed
  f31df59
- Convert the pythin module to work in double-precision
  sbird
  committed
  e1097ec
- Convert the old-school C-extractor code to double as well
  sbird
  committed
  0fbc79e
- Cleanup unused argument in internal function
  sbird
  committed
  1a40cbe
- Remove the particle pre-filtering
  sbird
  committed
  fef9f40
- Clean up some left-over pre-filtering functions
  sbird
  committed
  21cd53d
- Drop an old compatibility name for python 2
  sbird
  committed
  34e86b6
- Fix a few pylint warnings
  sbird
  committed
  d8f75e6

Commits

Commits on Aug 5, 2026

Commits on Aug 6, 2026

Commits on Aug 7, 2026

Commits on Aug 8, 2026

Commits on Aug 9, 2026

Commits on Aug 11, 2026

Commits on Aug 12, 2026

Commits on Aug 13, 2026

Commits on Aug 14, 2026

Commits on Aug 15, 2026

Commits on Aug 16, 2026