re: DEV Community Is Open-Sourcing Its Performance Metrics via Skylight VIEW POST

re: Queries that load users are perfectly fine, they use primary key to do so. Like you mentioned, I would just limit columns that it selects. Now ind...

Are indexes on boolean columns that beneficial? I'm all for adding it if they are.

The amount of "that" unfortunately needs to be measured. I don't think a full index would make a difference, but a partial one might (if published = false is significantly less used).

Something like:

create index articles_published on articles (published) where published;

I'm trying to generate a massive articles table to see if I can have PostgreSQL activate the index.

I think there is one already or am I missing something?

There is one, but it's on "name", not on lower(name):

PracticalDeveloper_development> \d+ tags;
    "index_tags_on_name" UNIQUE, btree (name)

with something like this:

create index tags_lower_name on tags(lower(name));

in reality this index might be ignored, I don't think the tags table is that big, PostgreSQL might do a sequential scan if it deems it faster

All in all I still want to bring up the question of performance vs memory. My thought was that the includes(:user) is what is eating up a good deal of memory but is not necessarily killing the query. Its purpose here is to eager load user associations but in doing so without selecting columns, it loads the whole objects into memory, which can be costly.

Have you tried the performance vs memory difference by forcing eager loading in the includes?


Article.eager_load(:user).where(published: true).select(:published_at, :processed_html, :user_id, :organization_id, :title, :path).order("published_at DESC").page(2).per(12)
code of conduct - report abuse