Amazing explainer for something so basic as indexes. I'm always surprised how can people learn this so late. If there's another video I would recommend going into multiple column indexes too.
I used to work at EDF (electricite de France ) and we performed for PL/SQL script to update the 40 millions of line of customer to make the database faster. Kinda forgot about all of that. But that was satisfying to do.
SUPABASE is AWESOME. The Dev Team is teaching the product, how it should be. I am new to Supabase and these videos are very useful for me. If it's not important to you, try being overwhelmed with textwall documentations and useless videos.
Can you please update the docs for when to use which distance operator in HNSW and IVFFlat vector indexes. Currently it is unclear weather to use cosine distance, Euclidean distance or negative inner product. Also I would love to have detailed videos/blogs/docs on which embedding model to use for how many words(tokens) of data and also on hybrid search as well.
When I tried to index a table, it did improve the query time, but only by a couple of seconds. It was noticeable but not as much as in this video. Does it have anything to do with my supabase plan? I'm using the free tier.
In this video, the Explain Analyze tool reported a duration of 3 ms for the first query with username and is_active, and 0.02 ms for the second query with is_active and username. Can you explain why there is such a difference?
I would guess that either the machine is running multiple threads and the thread got swapped to inactive for 1 ms or that it cached the result from the first one, so it already knew the answer for the second one (because it processed it as the same query)
I left you database service , just because you don't have the query rows count features for the application codes . You have only in SQL but not in dart , javascript or any other languages
You can get the row count using the count option: supabase.com/docs/reference/javascript/select Also we have count() aggregate function option for counting rows of a referenced table: supabase.com/blog/postgrest-aggregate-functions#the-basics-of-aggregate-functions
Amazing explainer for something so basic as indexes. I'm always surprised how can people learn this so late.
If there's another video I would recommend going into multiple column indexes too.
I used to work at EDF (electricite de France ) and we performed for PL/SQL script to update the 40 millions of line of customer to make the database faster. Kinda forgot about all of that. But that was satisfying to do.
SUPABASE is AWESOME. The Dev Team is teaching the product, how it should be. I am new to Supabase and these videos are very useful for me. If it's not important to you, try being overwhelmed with textwall documentations and useless videos.
Loving the quote at the end!
Thanks excited to know more about Indexes.
Perfectly explained!
42069x faster
Missed opportunity! 😂
Please explain how the pgvector index works
Can you please update the docs for when to use which distance operator in HNSW and IVFFlat vector indexes. Currently it is unclear weather to use cosine distance, Euclidean distance or negative inner product. Also I would love to have detailed videos/blogs/docs on which embedding model to use for how many words(tokens) of data and also on hybrid search as well.
Tbh this isn't really something you should be looking at the supabase documentation for.
Fantastic video!
When I tried to index a table, it did improve the query time, but only by a couple of seconds. It was noticeable but not as much as in this video. Does it have anything to do with my supabase plan? I'm using the free tier.
How big is your dataset? In the video he's dealing with 10M entries
Nice, thanks!
Glad you enjoyed it! 🙌
maybe a noob question but does the advantage of the postgres query planner also work when I use it from the client side like in flutter? like in 08:05
Nice one jon!
Incredible! thanks
In this video, the Explain Analyze tool reported a duration of 3 ms for the first query with username and is_active, and 0.02 ms for the second query with is_active and username. Can you explain why there is such a difference?
I would guess that either the machine is running multiple threads and the thread got swapped to inactive for 1 ms or that it cached the result from the first one, so it already knew the answer for the second one (because it processed it as the same query)
will this index will work with iLike queries?
very didactic
Hot Damn!
I left you database service , just because you don't have the query rows count features for the application codes . You have only in SQL but not in dart , javascript or any other languages
You can utilise DB triggers and functions to update the count and query it in the frontend with ease
You can get the row count using the count option: supabase.com/docs/reference/javascript/select
Also we have count() aggregate function option for counting rows of a referenced table: supabase.com/blog/postgrest-aggregate-functions#the-basics-of-aggregate-functions