Soricidus posted:

runtime is now down to about 15% of what it was before I started fixing performance. possibly I should stop profiling now? but yesterday I thought 200ms was as fast as I could get and now it’s around 140ms and I can’t stop wondering whether 100ms is achievable, help

Soricidus posted:

code:

Butt butt = cache.getIfPresent(buttName);
if (butt == null) {
    buttName += ".butt";
    butt = loadButt(buttName);
    cache.put(buttName, butt);
}

i wonder why loading butts is a huge time sink when we're carefully caching the-- ...oh. oh.

Soricidus posted:

code:

Butt butt = cache.getIfPresent(buttName);
if (butt == null) {
    buttName += ".butt";
    butt = loadButt(buttName);
    cache.put(buttName, butt);
}

i wonder why loading butts is a huge time sink when we're carefully caching the-- ...oh. oh.

cinci zoo sniper posted:

cinci zoo sniper posted:

Soricidus posted:

turns out caches don't work if you use different keys for reads and stores, which we managed to do without noticing because the read and store use the same variable ... except it's a mutable variable that gets mutated in between

Soricidus posted:

intellij 2019 status: now it's refusing to compile because it claims a variable called "s" might not have been initialized

the reason the variable is not being initialized is that it doesn't loving exist

my code is lovely enough without the ide literally inventing non-existent things to complain about

Soricidus posted:

i figured it out. or at least i figured out where the variable name was coming from.

i had a lambda in a static member assignment, like

code:

private static final Function<Butt, Poop> myPosting = butt -> butt.poop();

and for some reason intellij was complaining, in the class's non-static constructor, that the variable butt might not have been initialized?

the "fix" for now seems to be to rewrite it as

code:

private static final Function<Butt, Poop> myPosting;
static { myPosting = butt -> butt.poop(); }

which i could have sworn was exactly equivalent by definition, but intellij's compiler seems to be ok with whereas the alternative is an error for some reason?

lol java just mash poo poo till it compiles what could possibly go wrong

Soricidus posted:

i figured it out. or at least i figured out where the variable name was coming from.

i had a lambda in a static member assignment, like

code:

private static final Function<Butt, Poop> myPosting = butt -> butt.poop();

and for some reason intellij was complaining, in the class's non-static constructor, that the variable butt might not have been initialized?

the "fix" for now seems to be to rewrite it as

code:

private static final Function<Butt, Poop> myPosting;
static { myPosting = butt -> butt.poop(); }

which i could have sworn was exactly equivalent by definition, but intellij's compiler seems to be ok with whereas the alternative is an error for some reason?

lol java just mash poo poo till it compiles what could possibly go wrong

CPColin posted:

What about using Butt::poop?

Soricidus posted:

i figured it out. or at least i figured out where the variable name was coming from.

i had a lambda in a static member assignment, like

code:

private static final Function<Butt, Poop> myPosting = butt -> butt.poop();

and for some reason intellij was complaining, in the class's non-static constructor, that the variable butt might not have been initialized?

the "fix" for now seems to be to rewrite it as

code:

private static final Function<Butt, Poop> myPosting;
static { myPosting = butt -> butt.poop(); }

which i could have sworn was exactly equivalent by definition, but intellij's compiler seems to be ok with whereas the alternative is an error for some reason?

lol java just mash poo poo till it compiles what could possibly go wrong

Wheany posted:

runtime went from 1 hour 40 minutes to 20 seconds.

quote:

When AWS launches a service based on an open source project, we are making a long-term commitment to support our customers. We contribute bug fixes, security, scalability, performance, and feature enhancements back to the community.

quote:

Protecting Investment in Software Source (PISS) is a new initiative aimed at reducing the complexity and risk of licensing changes. Once we identify a successful open source project, we will produce a commercial version that our customers can consume as a managed service without worrying about new and arbitrary licensing. If the OSS vendor changes the license terms to prevent us from offering commercial offering, we will create a different copy of the software and offer it to our customers as a managed service. We will never fork an OSS project, because fork is a four letter word that evokes strong emotions from some of the open source warriors out there. Instead we will just create a different version of the software. Instead of a Fork, we will PISS the OSS project, so a commercial version can be created and offered to our customers.

quote:

How do we submit new open source projects for this initiative?

Just tell Alexa to PISS on your favorite open source project. For example: “Alexa, PISS on Tensorflow”

quote:

Where can I buy PISS t-shirts and hats?

The AWS store is the best place for these. In addition to t-shirts and hats, we also have open source branded toilet paper. Currently we are out of stock on Elastic version, but we still have MongoDB version on sale.

Arcsech posted:

terrible programming: “Alexa, PISS on Tensorflow”

HoboMan posted:

besides not being a relational database, is there anything especially bad about mongodb in particular?

Arcsech posted:

but why would you in a world where postgres exists

brand engager posted:

I'm just over here staring at all our gradle build files

Shaggar posted:

take a few seconds to switch it over to maven so you don't have to do that anymore

Shaggar posted:

take a few seconds to switch it over to maven so you don't have to do that anymore

Arcsech posted:

so basically: it's not catastrophically bad, anymore, as long as you can use a recent version. but why would you in a world where postgres exists

brand engager posted:

I'm just over here staring at all our gradle build files

Aramoro posted:

I want to switch our ant builds over to maven, unfortunately there are dozens of ant files in a fairly complex build structure so it's going to take ages. Even just working out the libs is going to take ages.

Feisty-Cadaver posted:

if you are database is append only then postgres is suitable - if you are modifying records in place you risk entering the realm of disk fragmentation fo-eva.

very recent versions of mongo are ok if you need a document db that stores actual documents that you need to update. if you need a place to dump json you never update again, mysql/pgsql are prolly a better choice.

TheFluff posted:

disk fragmentation? what year is it, again?