getting different results with pyFAI==2024.9.0 and pyFAI==2025.3.0

[MehmetTopsakal]

Hi,

I think there is a bug here:

pyFAI/src/pyFAI/integrator/azimuthal.py

Line 720 in cf4658f

empty = self._empty

I changed

empty = self._empty

as

empty = dummy if dummy is not None else self._empty

and my problem was sloved.

[MehmetTopsakal]

I also believe, this issue is present in many other places!

[kif]

The migration should have been done everywhere to have pyFAI consistent (within a given version)

[kif]

Dear Mehmet,

This change of behaviour is known and described in the changelog of v2025.03:
https://github.com/silx-kit/pyFAI/releases/tag/v2025.03
The rational is to enable automatically dynamic mask for Dectris detector data:

• Automatically recognizes and masks out bad pixel
• dummy and delta_dummy values are already known from the manufacturer
• empty bins are no more marked with the dummy value (which can be very large for Eiger) but with the empty value which is fully under control of the user, 0.0 by default.

Basically we are making a complete distinction between empty and dummy values.

• dummy value is the value of pixels of the input image which are dynamically masked out
• empty is the value given to the mean value of a bin which does not have any contribution.

Mathematically, the default value should be empty = NaN but most downstream software dislike NaNs. With the new implementation, the user has the ability to define this value freely and independently from the empty value.

This is why I don't think it is a bug to be fixed.

Jerome

[MehmetTopsakal]

Thanks Jerome for explanation. Unfortunately, I need more help.

import numpy as np
import matplotlib.pyplot as plt
plt.rcParams["figure.constrained_layout.use"] = True # For better plots.
import fabio
import pyFAI

from silx.resources import ExternalResources


## https://pyfai.readthedocs.io/en/stable/usage/tutorial/Separation/Separate.html
downloader = ExternalResources("pyfai", "http://www.silx.org/pub/pyFAI/testimages")
image_file = downloader.getfile("Pilatus6M.cbf")
geometry_file = downloader.getfile("Pilatus6M.poni")
img = fabio.open(image_file).data
ai = pyFAI.load(geometry_file)
mask = None



i2d = ai.integrate2d(
    data=img,
    npt_rad=2000,
    npt_azim=360,
    filename=None,
    correctSolidAngle=True,
    variance=None,
    error_model=None,
    radial_range=(1,5),
    azimuth_range=None,
    mask=None,
    dummy=np.nan,
    delta_dummy=None,
    polarization_factor=None,
    dark=None,
    flat=None,
    method="bbox",
    unit="q_A^-1",
    safe=True,
    normalization_factor=1.0,
    metadata=None,
)


import xarray as xr
da_i2d = xr.DataArray(
                data=i2d.intensity.astype("float32"),
                coords=[i2d.azimuthal.astype("float32"), i2d.radial.astype("float32")],
                dims=["azimuthal_i2d", "radial_i2d"],
            )

fig = plt.figure()
ax = fig.add_subplot(2,1,1)
da_i2d.plot.imshow(ax=ax,robust=True)

ax = fig.add_subplot(2,1,2,sharex=ax)
da_i2d.mean(dim='azimuthal_i2d').plot(ax=ax)

produces this:

with pyFAI==2025.3.0

and produces this:

with pyFAI==2024.9.0

I don't see empty keywork in ai.integrate2d. What should I do in this case so that the new implementation is correct?

Thanks
Mehmet

[kif]

The empty value is not a kwarg of integrate[12]d but an attribute of the azimuthal integrator class:

ai = pyFAI.load(...)
ai.empty = np.nan
i2d = ai.integrate2d( .....)
[....]

I choose to set this as attribute because I could not see a use-case where one would like to have different empty value for different images.