Update pgvector to 0.38/latest in self-hosted about supabase HOT 7 OPEN

The format string would be %d/%b/%Y:%T %z.

1. I assume that log source is PostgREST
   

   supabase/docker/volumes/logs/vector.yml

   Line 95 in e559600

   # PostgREST logs are structured so we separate timestamp from message using regex

2. Sample timestamp would be 06/May/2024:08:16:11 -0500 according to PostgREST observability documentation

3. parse_timestamp function accepts strptime format string and is implemented in Rust chrono crate.

Here is a reproducible example for timestamp parsing:

use chrono::DateTime;
use chrono::offset::FixedOffset;
use chrono::format::ParseError;

fn parse_date(date_str: &str) -> Result<DateTime<FixedOffset>, ParseError> {
    DateTime::parse_from_str(date_str, "%d/%b/%Y:%T %z")
}

fn main() {
    let timestamp = "06/May/2024:08:16:11 -0500";
    println!("Original timestamp: {}", timestamp);
    
    match parse_date(timestamp) {
        Ok(date) => println!("Parsed date: {}", date),
        Err(e) => println!("Failed to parse date: {}", e),
    }
}

which outputs

$ cargo run
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.02s
     Running `target/debug/parse-pgrest-timestamp`
Original timestamp: 06/May/2024:08:16:11 -0500
Parsed date: 2024-05-06 08:16:11 -05:00

from supabase.

Thanks for the help.

I forgot to mention that the version of vector I was trying to use is 0.38 while in the master branch we use 0.28.1 here.

Are there any reasons not to upgrade to 0.38? 0.2x version is quite old now.

from supabase.

@stefanstidlffg Could you change the issue so it matches the issue template? https://github.com/supabase/supabase/issues/new?assignees=&labels=bug&projects=&template=1.Bug_report.md

from supabase.

I don't think it's a bug in the current code. It's just a breaking change if you ever go to vector 0.3x.

I could do a PR with the update to vector 0.38 if you prefer that. But I don't know if anything else is affected by the update.

from supabase.

I have modified the title to reflect the end goal, @stefanstidlffg if you can modify the issue description to follow the template, that would be great. 🙏
feel free to open a PR if anyone does want to tackle this!

from supabase.

I have modified the title to reflect the end goal, @stefanstidlffg if you can modify the issue description to follow the template, that would be great. 🙏
feel free to open a PR if anyone does want to tackle this!

I would if you tell me which template matches the situation. As said before this is not a bug.

from supabase.

@stefanstidlffg Yes it is not a bug for the current version. But eventually it will break when someone decide to upgrade pgvector and it is not forward compatible. Assuming parse_timestamp is available for current version you could only fix the timestamp parsing code

from supabase.