You are viewing a single comment's thread.

view the rest of the comments →

bikergang_accountant ago

Can I plug https://gitgud.io. It's run by the same guy that does catbox.

I'm going to try running the site on dotnet core and see what hangups come first.

We could go completely away from c#. This is how they get you on the db. You end up with two layers dependent on their environment and you can't simply write a replacement in one step.

I am shockingly ok with sticking with c#. I don't know it but I know you can't ask your main contributers to code in a language they don't know intuitively. I think core is the way to go even thought I could contribute the most on node.

bikergang_accountant ago

I'd really love to see a profile of the sql calls and their total execution time by template. Caching db is my jam.

@PuttItOut

For example this will produce a partner database with the same name +'.meta'. It stores data about the calls with a schema like this:
CREATE TABLE stat (template text,time int,calls int,past int, sig float);

time is total execution time.
calls is total calls.
past is how many times a different call shared execution time due to asynchronous.
sig is the standard deviation of the execution time (actually varience).

If we had data like that it's easy to identify one call to cache and that particular call speeds up 100x. I can even get past writes "invalidating the cache". I don't invalidate the cache. I write to the cached object and use a write to db when the key expires. Your most active data just lives in memory with no ipc or networking involved.

But the key is you need that profile information first. Your profiler needs write caching though so you aren't writing to a table for every read.