Thanks @Connor1y for your fix!

@lan496, I think this script has not been used. I don't remember why I wrote it. If you don't think this is useful, maybe can we remove it?

That does raise the question, how did @Connor1y stumbled upon this?

I think we do not use this script anymore, but I also would like to know how @Connor1y has found this bug.

This is also written in the python2 manner.
[-101] must be an eigenvector of the corresponding matrix, though probably this is not the answer @lan496 expects.

Since the get_symmetry_dataset doesn't return symbols of operations, I need a standard list of point group operations and their corresponding symbols. Thanks for your work, I just happened to find that you have the list. Then someone found that in some cases -4 was always recognized as 2. That's why I found the bug :)

Thanks @Connor1y.

I am still confused about the long term action for this. Is the data not available in the C-API? Shouldn't we expose it on the python side?

Ideally we should have only 1 central source for all hard-coded data. We can keep it in python for readability and synchronise it with the C-api using cog

I think this file can be removed in the long term. The use case of this data that I supposed was providing symbols of Wyckoff positions, but now I think those derived information should be provided by other software package.

I don't think we should offload too much to third-party software. I believe this would be within the scope of the project and we could open a help wanted issue with some reference. I am always available to do reviews of the technical parts.

I see, but at least about the information of this issue can be removed. What I meant is the site symmetry symbols of Wyckoff positions, and the symbols defined in this file are not enough for that. Currently, I don't have a good idea how to implement the site symmetry symbols of Wyckoff positions.