RT @duckdb: We are happy to announce DuckDB v1.2.0 “Histrionicus”! The new release has several usability, security and performance improve…

RT @mehd_io: Once again, I was faced with the unfortunate reality of parsing CSVs. Fortunately, @duckdb's excellent work on their CSV sniff…

RT @yusuktan: 仕事でCSVをちょこっと解析する必要があって、DuckDBを初めて使ってみたらとても体験が良かった

@matsonj @duckdb How else should I know is time to buy spx6900

RT @jamesacowling: Got to sit down with Professor @andy_pavlo on the latest @convex_dev Databased podcast for his 2024 databases year in re…

RT @duckdb: New blog post: Query Engines: Gatekeepers of the Parquet File Format In this post, Laurens Kuiper argues that we are wasting…

I had a little adventure investigating a SIGSEGV when loading DuckDB with RocksDB/TensorFlow on Linux. This led to a spiral of checking leaked symbols, and learning that random_device has two different interfaces, one in libc++ and another in libstdc+, which can cause symbol collisions. 😄

@caiocgomes @kaitou_renegade Interessante! Por curiosidade, você tem um exemplo de plano de consulta do seu modelo, ou um benchmark que possa compartilhar? 😄 Se for melhor em pvt, pedro@duckdblabs.com

Sem duvidas Caio! Eu só queria fazer um teaser de que muitas vezes o que parece ser um volume de dados necessário de uma solução distribuída, pode não ser exatamente o que parece. Ambas ferramentas tem seus propósitos, vantagens e desvantagens :-) Mas eu tive que subir um cluster spark em 2015 no meu mestrado, talvez dai venha meu trauma 😂.

@kaitou_renegade @caiocgomes A gente tem suporte tanto pra iceberg ( quanto pra delta ( e 2025 promete pra ambos 😄

@kaitou_renegade @caiocgomes E mesmo em uma tabela wide, com várias colunas, por conta de projection pushdown, a quantidade de dados que você carrega é bem menor, IO de disco desce muito, alguns operadores podem até executar direto no dado comprimido, etc. Pra precisar subir um cluster de spark vai chão 😅

@wileycwj After I got a CSV file exported from a big Dutch bank with unescaped quotes in quoted values I lost all my hope. And then I sold the shares I had 😂.

Last year, I added a new integer type to DuckDB called VARINT, capable of storing up to 1,262,612 digits. I initially implemented a quadratic algorithm to convert it to VARCHAR, which was slow. (My plan was to eventually get inspired by CPython.) However, an external contributor beat me to it: They even referenced "The Art of Computer Programming" and implemented a solution that's more than an order of magnitude faster than my naive approach. Just excellent!

@wileycwj I kid you not, after supporting encodings, I think this is the most requested option. At least it’s not multibyte quotes/escapes (yet). 😅

RT @duckdb: New blog post by @__AlexMonahan__ Vertical Stacking as the Relational Model Intended: UNION ALL BY NAME Alex takes us on a jo…

“DuckDB has entered the zeitgeist as the default choice for someone wanting to run analytical queries on their data. Pandas previously held DuckDB's crowned position.”

RT @duckdb: The DuckDB repository just reached 25,000 stars on GitHub. We used this occasion to stop for a moment and reflect on the projec…