Bug in Iqxqy plotting non rectangular / square matrices?

[ricardo]

My users are complaining about SASView giving negative Y Axis for the following dataset:
​https://pastebin.com/raw/buBiU5JG

This is what sasviews gives:

​https://postimg.cc/R3TzMjSy

This is what I get with Matplotlib:

​https://postimg.cc/nsxJfHJq

This data is not a squared Maxtrix (looks more a sparse matrix):

In [3]: data = np.genfromtxt(filepath, skip_header=2, names="qx, qy, iqxqy, err"
   ...: )

In [4]: qx, qy, iqxqy, err = data['qx'], data['qy'], data['iqxqy'], data['err']

In [5]: len(np.unique(qx))
Out[5]: 78

In [6]: len(np.unique(qy))
Out[6]: 94

In [7]: len(qx)
Out[7]: 6978

In [8]: len(qy)
Out[8]: 6978

In [9]: 78*94
Out[9]: 7332

[pkienzle]

Confirmed the behaviour in current master.

The bug is in the assignment of qy values in the 2D data, not in the plotting. Plotting uses equal axes. Since the qx range of the displayed data is twice the qy range, the image appears horizontally stretched and vertically squashed.

The error is could be in the loader not preserving the qy values correctly, or in the conversion of the data into uniform steps in qx/qy, or most like to the final assignment of the qy axis to the 2D array.

[ricardo]

Just did some debugging.

The problem appears to be in sas/sascalc/dataloader/file_reader_base_class.py FileReader::_remove_nans_in_data (line 218).

Before calling this function (line 273) both ymin and ymax are good.

[pkienzle]

Look at line 278:

dataset.Ymax = np.max(dataset.qX_data)

[ricardo]

Thanks @pkienzle!

Anyone can merge this?

​https://github.com/SasView/sasview/pull/197