Andrius Merkys

This issue has not been addressed by #376. However, I wonder whether the granularity is right. Suppose a database is serving both fractional and Cartesian coordinates. Some entries have only...

I think `:` in property names will break the filter syntax, unless they are not supposed to be used in filters, but I guess having them queryable would be a...

> Yes, ":" would require us to extend the filter syntax - but it is a quite minor extension. I disagree, `:` is already used in zip comparisons, thus allowing...

> As far as I know, the "." is used to address subfields of dictionaries. This is true. However, since we probably will not place property data and meta in...

> I'm reluctant to overload "." to mean both data dictionary subfield and metadata suffix. I get how one can see ".meta" as a generalized subfield, but especially for a...

Since metadata placement discussion has branched to #462 and #463, I suggest bringing back the discussion about specifying the origin of a property. For simplicity I would vote for binary...

> @merkys But, there is a PR for this, #455 ? PR #455 is about marking structure origin, i.e., experimental vs. DFT-derived structures. The current issue is about marking properties...

I am mostly interested in chemical connectivity. However, I would expect the definition of chemical bond and its types to be quite involving. Could we adopt some already existing convention?...

Thanks all for interesting responses. I agree that choosing the right representation for bond type/order will require a lot of thought. Thus I find @vaitkus's suggestion really appealing: > *...

> I would prefer a more succinct format. Why duplicate "site" a zillion times? Maybe just array of arrays. I understand the pros of a more succinct representation, but I...