Updates from NIAC 2016

[phyy-nx]

Action items from minutes completed:
-H,K,L change to NX_number for incommensurate structures
-Discussion on flags, change to bitfield, no enum for efficiency
-Expand the description of partiality
-Need to keep mm and pixel positions because of parallax
-Write out predicted and observed
-Drop the mm in names, rather use units
-bounding_box as array, document the usage
-Spell out background
-Drop the val on the intensities, keep the _var version
code is not in place we defer it for now.
-Add polar_angle and azimuthal_angle

TODO
-Need documentation for polar_angle and azimuthal_angle
-Need to add some examples of flags. Would like to use the DIALS flags.

Deferred:
-overlaps is a list of overlapped reflections. This is a ragged array. As the

[zjttoefs]

The bbx could have the correct dimensions enforced.

Otherwise this looks like a nice improvement (without checking the NIAC minutes). I'll defer merging so you can add the commits to the remaining TODOs here, though. If you prefer an early merge, let me know.

[zjttoefs]

I like the new "bounding_box" name and the added documentation for the flags and angles.

Since the first dim "n" for the bounding box is never used elsewhere (as far as I see), that might be dropped. But it might come in handy as well.

[zjttoefs]

This should have addressed all NIAC comments.

[prjemian]

In the flags fields, there is bit 17: bad_shoebox. It is not obvious what is a shoebox in this context.

[phyy-nx]

dials/dials#363 aims to document all these flags, including bad_shoebox. When those doc strings are in place, I'll put them here.

Question about the first dim "n" for the bounding box. Should all the other columns have 'n' as a dimension? For example, 'h', 'k', and 'l' should all be arrays of dimension n.

[prjemian]

Should all the other columns have 'n' as a dimension? For example, 'h', 'k', and 'l' should all be arrays of dimension n.

The existing NXDL files are not consistent in this regard. Yours is a good idea, fits the current circumstances, and would inspire us to revise NXDL files similarly. One reason for the inconsistency in other NXDL files is that some fields have a variable shape.

[phyy-nx]

Cool. I'll get that going then.

…

On Tue, Jun 13, 2017 at 9:15 AM, Pete R Jemian ***@***.***> wrote: Should all the other columns have 'n' as a dimension? For example, 'h', 'k', and 'l' should all be arrays of dimension n. The existing NXDL files are not consistent in this regard. Yours is a good idea, fits the current circumstances, and would inspire us to revise NXDL files similarly. One reason for the inconsistency in other NXDL files is that some *fields* have a variable shape <https://duckduckgo.com/?q=hdf5+dataset+shape&t=ffsb&ia=web>. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#568 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AM2Oyrh6ZYyg0B9iat66pTGL7HNxBlZzks5sDrWpgaJpZM4Nwgv7> .