search

aiidateam / aiida-tutorials

AiiDA tutorials web site
http://aiida-tutorials.readthedocs.org
23 stars 37 forks source link

✨ NEW: explain computer PK better #202

Open ltalirz opened 4 years ago

ltalirz commented 4 years ago

Question arose while working on https://aiida-tutorials.readthedocs.io/en/latest/pages/2019_ISSP_Chiba_Japan/sections/calculations.html#computer-setup

Indeed - verdi code list shows PKs, while verdi computer list doesn't. The problem is, of course, that computers are stored in a different table.

I do think we need to find a more prominent place for information on the entities available in aiida than this https://aiida-core.readthedocs.io/en/latest/concepts/provenance.html#other-entities In particular, since Code are now stored in the Node table, but still have a number of dedicated commands (verdi code, load_code, ...).

mentioning @sphuber for info

ltalirz commented 4 years ago

Maybe also worth mentioning @CasperWA since from the export you've become an expert on all the different entities ;-) An "entity overview" would be very useful I believe

mbercx commented 3 years ago

Good point! I'm not sure if we have anything like this in the current material. Maybe something to add to the "Data" section? What do you think, @CasperWA, expert of all thing entity? 😁

sphuber commented 3 years ago

The title of this issue is and the statement "The problem is, of course, that computers are stored in a different table." are incorrect: Computer instances do have a pk. Any entity in AiiDA has a pk, not just nodes. The discussion in this issue doesn't really seem to address this question though. I suspect the original confusion by the user was why verdi computer list does not show the pk. So I doubt that an entity overview would have addressed the original question, still might be useful though. Probably we should think about maybe adding a flag to verdi computer list that will show the pk and UUID in addition to the label that is being shown by default

ltalirz commented 2 years ago

While adding a PK to verdi computer list solves part of the issue, another aspect of my point here was that a computer and a node can be assigned the same PK, i.e. there are "computer PKs" and "node PKs".

Without an overview of different entities in AiiDA, a new user has no chance of knowing about this structure (e.g., there is no intuitive reason that would allow you to deduce that Code PKs are the same as Node PKs but Computer PKs are different).