A friend at LinkedIn who would prefer to remain nameless tells me they reduced their storage costs by 7% just by manually interning the phrases "Congratulations!", and "Congrats, <name>!" in their comment database schema.
I came across a really cute algorithm for count-distinct yesterday and couldn't believe I hadn't seen it before, so I did a little write-up on how it works:
*first contact with an alien*
me: hey btw can you guys do SAT in polynomial time
<alien jeff👽 has sent >
me: is this c++?
them: yeah just run make
me: it says im missing something called "libyorbo"
them: oh really? thats required for posix here
NULL BITMAP: I wrote down some thoughts on the problems you run into when trying to represent non-treelike expressions for query languages, and some of the solutions I've seen:
Today is my first day at
@MaterializeInc
!
I also have a new blog post—in preparation for starting, I read some of the papers for the tech underlying the system, which led me to an interesting problem which I wrote about here:
This week I wrote a bit about some of the ways I have seen columns represented in query optimizers, in particular a nonobvious (IMO) way that is vastly superior (IMO) to the obvious way:
this week i wrote about a reductive but helpful (to me) way of thinking about how datalog works. and also followed up on some threads from last week. i hope you like it
I really enjoyed
@TylerGlaiel
's post the other day, and while it wasn't the point, I thought his pathfinding problem was interesting enough that I should write down some thoughts on it:
Unfortunately, this same unnamed friend has just gotten fired for unsportsmanlike conduct during a company water polo game. As a result I can no longer verify the validity of these claims
me: can you just explain the algorithm
them (getting flustered): oh uh well i haven't done this since college man.
<alien jeff👽 has sent >
me: this is getting an index out of bounds when i run it
them: im not used to doing this without my IDE
hello... the first issue of my newsletter NULL BITMAP is out and we are starting off with a classic but surprising issue that comes up with implementing SQL: the halloween problem. you can read it and give me your email (🔫) here
fixing a typo in the readme of a github repo with 35 stars and then putting
Coder | Dreamer | Open Source Evangelist
Location: ✈️ -> The World
in my twitter bio
i used to work at a company where the IT staff would regularly send out fake phishing emails and you'd get a talking to if you clicked on one. this was my favourite one they ever sent
I saw
@irfansharifm
at a Go meetup in Toronto today. I told him how cool it was to meet him in person, but I didn’t want to bother him and ask him for photos or anything.
He said, “Oh, like you’re doing now?” and went back to making his slides.
DB VENDORS: you still have to handle aborts and issue retries at lower isolation levels, right? so it shouldnt matter if theyre more frequent at serializable
DEVELOPERS: well the way we handle retries is to display an error and tell the user “something went wrong”
Fascinating disconnect I’ve seen on this one:
- academic researchers: SERIALIZABLE isn’t the default? But everyone enables it first thing, right?
- database users: SERIALIZABLE exists?
Big news! Like, REALLY big news.
We're popping the lid on
@MaterializeInc
today. You can check out our sweet new web page: , and click through to the first in a series of blog posts:
There will be more in the coming days, but ...
HELLO!!! for the month of february i decided i was going to make a ZINE. so I made a strategy guide for a game that doesn't exist. i hope you enjoy FISHERGIRL
important annoucement: i have written a program that solves the halting problem. annoyingly, my little brother Bungo unplugged my mac so i am currently powering it back on. luckily i ran fsync on the source file before i lost power
hey ✌️ this is my last week at cockroach labs. excited for something new, but sad to be leaving. it's been a true privilege to work with some of the smartest, most thoughtful people i've known
I wrote up some thoughts about the process of generating SQL queries for testing. It includes a plea for ideas on how to unify semantic analysis and query generation!
it’s too bad query languages don’t really seem to capture people’s imagination like programming languages do. it would be cool to have a lot more random crappy half baked hobby query languages floating around (sincere)
i was very excited to join my friends
@conor_power23
and
@lisabdunlap
to talk about writing and databases and writing about databases! we had a great chat
are there any languages that have type-level debuggers? like if you have some nontrivial trait/template level stuff, some way to step through unification?
guy emails me to say he likes my blog, offers a chat with an eng about their product. im not looking right now but i hop on the call anyway to see whats up. its with a recruiter instead. they ask why they dont have my resume, i say because i never applied. 2 weeks later:
*guy who has had two decades of in-office work to make industry connections and friendships voice* yeah i dont see why we should ever go back into the office, really. remote is the future
db correctness bugs: 0.001% of my data validity errors
hardware problems: 0.01% of my data validity errors
application logic: 99% of my data validity errors
someone who is good at applications please help me budget this my software is dying
hello, one more life update, as of this month i am a "research software engineer" working with some cool people at Sutter Hill. please update your spreadsheets
We introduce Ternary Logic Partitioning (TLP), a general, effective approach for finding logic bugs in DBMS. It nicely complements our work on PQS and NoREC. TLP already found 175 bugs in widely-used DBMS, and can detect bugs in a wide range of features.
Using overly technical language in communication can be alienating. Here are some jargon terms nerds love to use that I think are problematic:
"Computer" instead of "thingy"
"Program" instead of "thingy"
"Project" instead of "thingy"
"Office" instead of "thingy"
CODD DEFINED "RELATIONALLY COMPLETE" TO MEAN "EQUIVALENT TO RELATIONAL CALCULUS" DESPITE NOT EVEN HAVING A NOTION OF RECURSION, IF YOU ASK ME, DATALOG IS A MUCH MORE NATURAL STARTING POINT,