/tech/ - Technology★

>>852508 (OP)

SQLite, Postgres if it could scale a lot, for simple tasks just embedded key/value database library.

Mongo and Couch tend to be pretty fucking useless.

.txt files

Depends on the use case. No DB fits all.

>>852691

Don't talk to me or my database ever again

sqlite is literally the best

Need internets? (Y/N)

Y: Postgres

N: Sqlite

>>852825

this actually

>>852510

If one were, in fact, going to use this thread for data mining, which database would be best to store the mined data in?

I'm a big fan of SQLite, It's great for literally everything that isn't a website.

>What's your favorite dbs /tech/?

Mnesia

>>852825

Why would it matter if you need internet or not? No database should interface directly with the internet. They should be gated off by servers that sanitize and authenticate data before read/write occurs.

>>852952

It's for internet scale things, Postgres is much better at handling concurrent read and writes.

Sqlite can either read or write, not both.

Microsoft Pocket Access 97

>>852962

sqlite3 can handle concurrent reads and serialized writes. It works well if you have a website that just reads and a backend that just writes

>>852510

InfluxDB for time-series data, sqlite3 for most stuff, Percona MySQL in a Galera cluster for serious stuff.

>>852868

Oracle, obv. It's "Optimized for Evil"(TM)

I'v been writing my own graph database which uses GPGPU to do querying. Since graphs are a superset of relational databases, I'm planning using it to power most of my future projects.

I added predictive modeling features to it as well. There's an algorithm i'm working on which will dynamically create/prune edges between verticies and represent them as recurrent neural networks or baysian networks (depending on which is more applicable) and attempt to find the relationtion between the data based on known patterns and time-sequence data.

Preliminary testing has shown it can reduce the total amount of needed edges in the network and find previously unknown relationships between verticies. Although for mostly uncorrelated data it doesn't work at all.

Its as of yet unnamed and unpublished, but I plan to release open source it once it is stable, if anyone else is interested.

>>853244

>Although for mostly uncorrelated data it doesn't work at all

what does it mean?