2,185 karma · joined September 27, 2010
They were testing SQL behavior for Materialize and thought to check that both systems handle interval functions identically. Being thorough, they tried something like:
select interval '0.5 months 2147483647 days';
You can try it yourself on dbfiddle[0] Instead of erroring, Postgres returned a bogus value `{"days":-2147483634}` you can read why here[1]So naturally they decided to fix it in Postgres, which is why they contributed and why it's handled properly in 15+ [2]
[0] https://www.db-fiddle.com/f/ijT76fsmL99bHvXxhAtf7j/0 [1] https://git.postgresql.org/gitweb/?p=postgresql.git;a=commit... [2] https://www.db-fiddle.com/f/i3KikCb72AN1EZpywErZvr/1
Every morning when I exit the train to a blast of heat on cramped platforms, wait in line for my turn to weave up a tiny staircase, into the aging waiting halls with hoses routing ceiling leaks into trash cans, zig zag through a labyrinth of low-ceiling corridors, giving a wide berth to half-naked mentally ill people, to a side exit (since the main exit has been closed for months, and finally, as icing on the cake, I jump over a permanent puddle created by construction barriers RIGHT at the crosswalk where THOUSANDS of people exit the busiest train station in the entire country, I think about the line:
“One entered the city like a god. One scuttles in now like a rat.”
I guess I'm surprised that the approach is to use semantic search here, where in regular search, completely different algorithms (Lucene?) have "won".
I have questions now: It seems like everything hinges on the search step, where domain-specific content is gathered to serve as prompt input for the LLM.
1. In an area like technical docs, if I can't find what I need it's often because I don't know the right terms - e.g.: I am looking for "inline subquery" but they're really called "lateral joins" - Would the search step have any likelihood of finding the right context to feed the LLM here?
2. How much value is added by feeding the search results through the LLM with the user's prompt, vs just returning the results directly to the user?
3. Are there good techniques being developed for handling citations in the LLM output? IIRC Google had this in Bard
IIUC analytical = "what question are you trying to answer" and in analytics, RT/NRT is absolutely novelty/vanity. Operational = "what action are you trying to take" and it makes sense to want to have up-to-date data when, for example, running ML models, triggering notifications, etc...
Anyone here taking them up on it? I'm genuinely curious how it's going.
Oracle has decent support for incrementally updated materialized views, redshift has some too. Materialize.com is an entire snowflake-like platform built around incrementally maintained materialized views.
Here is a longer lecture on a similar idea from U Chicago Writing Program https://www.youtube.com/watch?v=vtIzMaLkCaM&t=3s
> However, how this plays out with browsers that allow extensions or are modified remains a grey area. As the proposal vaguely mentions, "Web Environment Integrity attests the legitimacy of the underlying hardware and software stack, it does not restrict the indicated application’s functionality."
That's not vague at all.
If I recall correctly, it was made available in response to another big wave of criticism directed towards Google about "Content Farms". It has been interesting to see the difference in response between the "Content Farm" debacle and what we are dealing with currently.
[0] https://googleblog.blogspot.com/2011/03/hide-sites-to-find-m...
> That, combined with the fact that these streams usually have large viewership, means we incur a not insignificant cost and have no one to bill. In 2021 alone, Mux had over $750,000 in unpaid invoices due to pirated streams. For an infrastructure company like Mux, this pirating comes with hard costs. Transcoding, storing, and delivering video is not cheap. If these pirated streams were not held in check, they could quickly spiral out of control and have a significant negative impact on our business. By identifying and shutting down these streams, we are able to reduce our costs.
So to actually saved $750k, they'd have to have some combination of:
1. 100% COGS or 0% operating margin. Youch. Probably not, likely closer to 30% COGS, making actual infra/bandwidth cost to them of $750k services closer to $225k 2. Massive growth in attempted abuse. E.g. Their level of attempted abuse grew 3.3x to $2.5M, they stopped all of it, that would have cost them $2.5M x 30% COGS = $750k
But at the end of the day I think they just wrote a clever headline to get upvotes on HN :)
As a vendor, if you want to participate, you have to have a public listing page with a price. But many vendors don't actually want you to initiate the buying process that way. They want you to talk to them first, negotiate a deal/contract, and only then handle the payment through AWS marketplace (presumably with pre-committed spend you already have on AWS) via something called "Private OFfers"
So lots of marketplace pages have an absurdly high public listing price to discourage people from using that route.
Everyone should watch it https://www.youtube.com/watch?v=GAXLHM-1Psk
I'd just look for what people say in the bad reviews. I generally use mine on the lowest temps, 80 or 90. I don't know what kind of maniac would use them at the higher temps. Every time I turn it on I have to press the button to cycle through the temperatures, a mild annoyance but NBD. It automatically turns off after a few hours which is a feature, not a bug. It used to be that I'd sit down at my desk and my hands would never warm up, now they warm up from the mat and stay warm even after the thing turns off.
IMO these are the best people because they intervene before a company or team veers too far off course.
What are the use cases that max out Redis speed/throughput?
Also, I bet there are SWE's from FanDuel and DraftKings on HN. No judgement I would love to hear how you see things from the inside.
If I understand correctly: you'd run the recursive query, it produces results for every step, effectively showing you the progression of output over time, and then once it hits "present day", it completes and stops?
How would you generate results going forward? I.E. A minute elapses after the results return, do you have to re-run the whole query for all time?
It's so obviously just that the other company was like checking to see how your product and there's integrates. They weren't actually using it. But now they're in the "Trusted by" panel on the homepage.
I think I understand the model here, but I guess I'm curious how you model in risk of suddenly getting stuck with a bunch of 3 year RI commitments?
> "Hugo sounds great! Let me try it out!" > "Okay, I made a new theme and a page in content/!!" > "Running hugo server -D now!!! ... Where is it!!!!"
It's not Hugo itself that is complicated, it's all the other JS and CSS dependencies that other getting started repos tend to layer in.
I think this can also be extended to data tools. I've had so many experiences where a SaaS tool that tries to automate away the manual SQL analysis shows something totally wrong. I tend to "Not trust, verify" anything coming from a SaaS UI when serious business decisions are about to be made based on it.
SQL interface on Differential/Timely Dataflow = https://materialize.com/
Python interface on Differential/Timely Dataflow = https://bytewax.io/
Grossly oversimplified, but directionally correct.
The core components of Materialize have always been licensed under the BSL [0], with a conversion to the Apache v2.0 License four years from the date of publication. These components (now called the storage, compute, and adapter layers) continue to be developed in the open under the same license, but are no longer packaged or supported for on-premise deployment.
Timely Dataflow and Differential Dataflow [1] are the open source frameworks that have always been at the heart of the compute layer in Materialize.
If you're referencing the downloadable binary when you mentioned "the open-source version", we decided the opportunity cost of splitting our attention and continuing to release updated binary versions by "rebundling" Materialize is too high. We have a small team and ambitious plans for Materialize, (maybe some `WITH RECURSIVE`?) so we don't plan to release updates to the binary.
[0] https://github.com/MaterializeInc/materialize/blob/main/LICE... [1] https://timelydataflow.github.io/differential-dataflow/intro...