NGL 2.2.2 doesn't show bond

[hainm]

Hi

Please see this issue: nglviewer/nglview#1008 (comment)
I am able to reproduce with NGL 2.2.2 too

File is here (need to change .txt to .cif): https://github.com/nglviewer/nglview/files/13765118/last_molecule_r_0-pos-1.txt

[ppillot]

The core cif file parsing functions expects the field atom_site_type_symbol to properly infer the element. This cif file does not have it: it only contains the atom_site_label.
It seems that, in this case, we could default to infer element based on the later, in case atom_site_type_symbol is missing.

[jump2cn]

hi team,
I noted that the browser console warns with 'more than 500 atoms...', it's come from this code.

https://github.com/nglviewer/ngl/blob/master/src/structure/structure-utils.ts#L606-L607

it will work well for me when I change the check limit to 99999. can we have a bigger count limit or the count limit can be a configured parameter?

[fredludlow]

@jump2cn - That code is inferring bonds between atoms in the same residue, and the 500 is there to avoid an explosion in complexity - I guess it's kind of an arbitrary number but 99999 seems a little high - your file has ~1500 atoms? We could set it to 2500.

I was trying to find an example where it would behave more poorly - it's doing a nearest neighbour search using a kdtree then for each pair looking up to see if they would reasonably be bonded. In your case where there's no disorder that nearest neighbour search isn't going to be a problem even if every atom is queries, most will return only a few neighbours. For something like an NMR structure with 20 models of a large ligand all overlaid I guess it could get expensive, but I couldn't find an example quickly where it was a problem.