uses serialized graph in tile model

[aarongundel]

Uses serialized graph in tile model. This was already used in quite a few places here, there were only a few changes to make.

[jacobtylerwalls]

Because load_serialized_graph() does a db query, we're probably giving back all the intended performance savings. I noticed the two implementations perform roughly the same (this is with a nodegroup from AfS):

python3 manage.py shell
>>> from timeit import timeit
>>> from arches.app.models.tile import Tile
>>> timeit("Tile.get_blank_tile_from_nodegroup_id('0434ccce-1d9f-11eb-a29f-024e0d439fdb')", globals=globals(), number=5000)  # pr
3.2668450839992147
>>> timeit("Tile.get_blank_tile_from_nodegroup_id('0434ccce-1d9f-11eb-a29f-024e0d439fdb')", globals=globals(), number=5000)  # dev branch
3.308565291983541

I can take a closer look at this tomorrow, but what do you think about that point in general, in other words only trying to do this (potential) optimization only if we already have the published graph in hand?

[aarongundel]

The idea here is that it's likely the caller of this method will not just grab the empty tile and be finished, but that they'll set some nodes in the tile and then save it back to the database - at which point this becomes a bit more valuable because then the serialized graph doesn't have to be loaded on save. So this is a prefetch of the the serialized graph for that tile object. However, it is true that this is not strictly needed here, and it's probably better left for when it is actually needed - the optimization is a bit premature.

[jacobtylerwalls]

Looking good, just a few suggestions.

[jacobtylerwalls]

Assuming we only have a queryset here, which might be the case after the outcome of the prior comment, this could be rewritten with .exclude() so that the exclusion of semantic nodes is done in the db and returns a smaller result set. Since we're only using the nodeid you could also select only that with values/values_list.

[jacobtylerwalls]

Thanks, looks great!