401 karma · joined June 2, 2011
Netezza/IBM had a reasonably efficient implementation. Instead of broadcasting individual locks, they shared ordering relations between transactions as they were discovered.
Essentially, when transaction A reads a record and concurrent transaction B modifies the record, then B serializes after A (even if B's modification occurs before A's read in real time). These ordering relationships form a a graph. If a new ordering relationship between transactions causes the formation of a cycle, then some transaction must be aborted.
That said, the effect is exactly as the author claims. The transaction on the right will ignore the change made by the one on the left.
I believe AS is required for self-joins. A common expository query is to find all employees who earn more than their managers, which can be expressed via "select name from staff as employee join staff as manager on (employee.manager_id = manager.id) where employee.salary > manager.salary".
Also, SQL allows a sub-select to be listed in the FROM clause. But in such cases, the sub-select must be given a name (tables have a name, but sub-selects don't).
As an aside, the keyword "AS" is allowed by not required.
To make the analogy work the backlog needs to get rated on how much each open issue costs the company. Most organizations use some stratified measure of "severity". Few actually try to put a price on a problem, but I think it's worth the effort to try.
So... is it possible that inflation and dark energy are related? When we see distant/ancient galaxies accelerating away from us, is there any chance we're looking at the ramping down of the original inflation of the big bang?
I'm a die-hard C++ user, and have come to love functional programming for a very C++y reason: improved performance that comes from lock-free parallelizability (if you never modify values, you don't have to synchronize access).
But I also love the simplicity that comes from not having to think so deeply about lifetimes and which portions of code can see which objects in which states. The abstraction of data-structure from algorithm is nice, the terseness is nice, the regularity is nice, but it's the lifetime and visibility simplification that helps me the most.
I don't know Haskell or OCaml. I wouldn't know a monad if it slapped me in the face. But functional C++ makes me happy
PG 9.5 introduces some support for parallelization along with custom scans. So this capability seems like a natural fit.
As for utility, big data queries tend to do heavy aggregations, and parallelization is a nice performance enhancement where it could be sorely needed.
Redshift has good name recognition, but under the covers, it's ParAccel, which started with an early (8.2?) version of Postgres and modified it.
* empty string == NULL in oracle
* ROWNUM: it's kinda like a limit, but not
* DUAL: hah. postgres eat your heart out.
And of course there are business practice differences:
http://www.businessinsider.com.au/oracle-customer-explains-a...
... but tongue out of cheek, if you can afford it, Oracle has solutions for just about every database problem in just about every vertical domain. They have really good engineers, really good testing, and they're #1 for a reason
Perhaps I missed it, but I don't think the post specified how many rows were in the source. Could have been more than a million. Having a relatively expensive filter (regex) eliminating all results keeps the performance picture clearer. Anything done with a tuple that passes the filter would apply to both the regular and the parallel scan cases, so eliminating the results highlights just the difference between regular and parallel scans
You might be surprised how many queries are bound by memory bus bandwidth and CPU, rather than I/O ... but even I/O bound queries may benefit:
Remember that postgres is a "demand-pull" system. A sequential scan node returns a single tuple at a time, whenever its parent node asks for a tuple. So there can be considerable time between requests for an actual disk block.
Parallelizing the scan should result in more frequent requests to the disk controller, which could then have better opportunity to optimize reads - even for a single disk. Since most of the latency for spinning disks is seek time, having more outstanding block-read requests should help performance.
(a) at-home day care has SEVEN times the risk of infant mortality of commercial day care centers... but the mortality rate is SO low, that driving an extra ten minutes to the "safer" choice increases the overall chance of mortality (due to risk of car accidents)
(b) under Scott Walker, Wisconsin created HALF of all jobs created in the USA... ~18,000 net jobs were created, and Wisconsin created ~9,000. But neighboring Michigan created 13,000. Some states lost jobs.
More sophisticated examples of faulty reasoning using statistics. Don't know whether the book is really helping me to listen more critically, but it's fun to think I might be wiser, or at least less gullible, after reading it.
States license things all the time, and the conditions of their licenses block certain people from doing certain things. Why are the courts blocking the right of states to license this activity (marriage) in this particular case?
Again, I'm personally happy with the outcome. It feels as wrong to me to say the LGBT can't marry as it would to say an interracial couple can't marry. But Scalia asks what the legal reasoning is, not whether it's the right outcome. What's different about this minority or this situation?
Can anybody counter Scalia, and say why the issue of gay marriage couldn't wait to be resolved by the states? Why is this class of license inequity different than other classes, where the states' right to license something is not resolved by SCOTUS?
This doesn't mesh with my experience, but I love the phrasing. And it makes me wonder whether and how to reward walking away from a deal that could be closed, but would be problematic. I'm not smart enough to know how to do that.
There are research frameworks that are quite good. One is Factorie from U.Mass Amherst (factorie.cs.umass.edu) that supports latent dirichlet allocation and lots more. Stanford also has wonderful tools.
Yes, you can dump text into SQL, and postgres has some text analytics. But my guess is that you'll soon want capabilities that RDBs don't have. Mongo has some support for text, but not nearly as much as postgres. I think both SQL and NoSQL are currently round pegs in square holes for deep text analytics. They're barely ok for search.
OTOH, why should OP have to know what's interesting a-priori? The mantra of big data is "listen to the data, let it tell you, leave your preconceived notions at the door." The fault is with our current tool chain. We need something that tells us what about the text is interesting before we dive in for a closer look.
I'm imagining a tool that told me: "this text seems to have a lot of opinions and sentiment," "it's about a product that was returned," "a number of people's names are mentioned," "this text was loaded along with some structured data that appears to reference a price, a location, and a date."
Why is it such a stretch to combine the tools we already have to generate and push summarizations? Maybe it's just the cost of computation? If you know you're looking for topic and don't care about sentiment, then you can avoid paying for it?
His name was Matthew Looks, and his point was that it doesn't matter whether you're right or wrong. It only matters that you have a visceral opinion.
If you believe something or care about something strongly, it gets you to engage. If it turns out that you're objectively wrong, the clash will focus your attention and you'll learn. Just having an opinion, just caring, will bring out a better game.
This is how I interpret PG's post on taste. Having good taste is wonderful. But having taste, even bad taste, is better than not caring at all.
Or take a design that consists of functional transformations of a sequence of ASTs. What organizational communication structure is stateless? ("Don't tell me you already told me that a minute ago! I refuse to remember anything you say. ever. You want to tell me something, start from the beginning every time. And furthermore, I refuse to talk to you at all unless we're given an entirely new language in which to express our conclusions. Besides, you're nothing but a monad to me") Ridiculous for human communication, but arguably essential for large-scale distributed parallel computation.
Or take event-bubbling, or model-view-controller. Does any organizational communication system use those in anything other than a once-in-a-blue-moon, accidental way?