Generated columns are not supported by older PG versions. For example if you used AWS AuroraDB with Postgres compatibility, they didn’t support generated columns till a few years ago.
In my experience using PG FTS, there are often some preprocessing tasks you might like to do before storing the index. One common thing is the generated index has to fit inside the column length limit. So if your user provides a long text field with a lot of unique words, then the index will be very long and might lead to a database error. So here you can check for that on the fly, and truncate the result.
Used Car prices have increased a lot in the pandemic due to the chip shortage, logistics bottlenecks, and rising demand. So most car loans that predate 2020 are in the money - the car can be sold to fully cover the loan. But cars purchased in 2021 are not necessarily in positive equity. This is why repos are growing and why they are mostly for car loans originated in the last two years.
For a typical suburban American home, 80% of the cost is labor and 20% is materials. This rule of thumb even applies in small cases - replacing the floor will likely cost you 5x more to have it done for you vs what the materials cost at Home Depot
Cruise ship arrivals were only 7% of Venice visitors in 2018. They have already been diverted to a nearby industrial port since reopening after the pandemic
As someone who’s used Bloomberg in a professional setting, this is more like 1% of the data for 5% of the cost. This doesn’t even have an Excel plugin or any other API. For people looking for an alternative, I recommend Refinitiv Eikon, which actually is 90% of the data for 50% of the cost and it has a native Python API and no rate limits on data.
This case was about users public profiles. LinkedIn makes users profiles accessible to the Googlebot and other public visitors without the need to login. There was an additional wrinkle in that HiQ actually had a commercial agreement with LinkedIn for some time that explicitly allowed them to scrape. They were arguing that canceling this agreement was anticompetitive and would destroy their entire business. Because of that pre-existing agreement, they likely didn't dwell on the distinction of public or private as they already set the precedent of signing up to an agreement that authorized them in the past, so they can't now say they never needed one in the first place.
Scraping is legal, but the legal framework requires scraping to abide by three rules:
1. The page is not behind a login. If you have to create an account to access the content, then you have to agree to abide by the terms of service which may ban scraping.
2. What you scrape is still protected by copyright. If you scrape a clip from Star Wars from a web page, you still can't redistribute it without a license.
3. Your activity may not impede access to the site in anyway i.e. no DDOSing. If there is a robots.txt file, you are supposed to abide by it but the courts do not literally say that. Must rate meter your requests, scrape at night, so on. Gray area if the robots.txt bans every page and you end up having to ignore it.
Those other economic activities generally have a regulator standing by, a gambling commission, the FDA, SEC, and FTC. Most of them aren’t anywhere near funded or have necessary powers or maybe susceptible to regulatory capture. But at least there is someone. In the world of private angel investing, there is truly no one but you and your legal team to go after crooks.
I used to do this but switched to iCloud Beta for custom domain this year. It’s $1/month (free with any paid iCloud plan) and integrates with iPhones/Mac, supports push email for iPhone, and is free of Google spying.
Not to mention the 10M barrels we sold just to pay for the upkeep (it's like selling gas from my car to pay for new brakes - we shouldn't be bartering oil for maintenance expenses).
But a trade is a trade - had they not bought the 77M the SPR would have dropped by 74M rather than the increase of 3M.
The US bought 77M barrels of crude oil in March 2020 under a Trump executive order. The purchases were directly from US producers, but WTI was trading around $20-25/barrel in that period. Selling 50M barrels now is going to profit the country by $50/barrel (before holding and transport costs), so by all measures this has been a good trade for the country!
While researching this I was also reminded that we are already selling 20M+ barrels per year from the SPR. Many major budget bills over the past 7 years included a sale from the SPR as part of the funding. We also need to pay $450M for overdue maintenance on the SPR itself - and you guessed it, that was "paid for" by selling 10M barrels from the SPR this year. This is not a petroleum reserve anymore, this is a budget reserve for Congress to balance its bills.
price reductions are seasonal in real estate. Peak new listings are around Super Bowl Sunday and Labor Day. There are very few price reductions in that period. The best quality homes at attractive prices sell first. By late October, the total number of listings on the market starts to decline, and price reductions become common.
My SaaS business serves corporate law firms and investment banks that invest in large corporate bankruptcies (think Hertz). When I launched, I first applied for a merchant license from Stripe, and was quickly denied by an automated system, citing violation of TOS (no reference to what violation). But given there are a ton of alternatives, I just used Braintree instead and it's been a great experience so far.
California and Quebec carbon allowances markets are now linked (trading as CCA), and Washington State is expected to join. I researched that market thoroughly and it’s not perfect but very well structured. It’s mostly carbon allowances, not offsets (offsets don’t really reduce total emissions). CCAs trade on the ICE exchange and there are regular scheduled auctions. Arguably more liquid than most stocks.
While that might be true, that's kind of saying that if I go around breaking windows, I'm helping the economy as people are going to go out and spend money to buy new windows. Our society produces a limited amount of goods and services, and any allocation to one project means we have a little bit less for everything else. Whether the beneficiary is American or not isn't really relevant as it doesn't change the end result that everyone else has to tighten their belts to make room for the war.
Those are state by state decisions. Many states choose to fund the prisons by taxing phone conversations. None of it is encouraged by the private industry - the prisons have a contest between the main vendors (GTL and Securus) and whoever offers to pay the highest percentage and amount of revenue wins. They of course know who is going to pay, but the public has a hard time associating “let’s raise taxes and spend more on prisons” with a reduction in telecom cost for the prisoners. And even those who do get it, view it as yet another way to be tough on crime. Ironically, having frequent contact with family is the best way to reduce recidivism.
If you factor in the resale value of a 2 year old iPhone and a 2 year old Pixel, I doubt this is true. And one of the biggest reasons for the high depreciation is the loss of software updates...
I moved from ElasticSearch to PG FTS in production, and here are the things I had to give up:
1. PostgreSQL has a cap on column length, and the search index has to be stored in a column. The length of the column is indeterminate - it is storing every word in the document and where it's located, so a short document with very many unique words (numbers are treated as words too) can easily burst the cap. This means you have to truncate each document before indexing it, and pray that your cap is set low enough. You can use multiple columns but that slows down search and makes ranking a lot more complicated. I truncate documents at 4MB.
2. PostgreSQL supports custom dictionaries for specific languages, stemmers, and other nice tricks, but none of those are supported by AWS because the dictionary gets stored as a file on the filesystem (it's not a config setting). You can still have custom rules like whether or not numbers count as words.
Minus all of the data though. That’s why people pay for Ancestry, is because they go in to countries around the world and pay to digitize archives and make them available to subscribers. Usually data like the census is made available on a rolling basis so every year there is new data to process and release
A few insights I've learned from my meetings with the company:
* The magic number is $40. If you send a large box of clothes to them, they will pay you some amount that's very near to $40. That's the amount that stops you from complaining.
* They effectively sell 6mm SKUs because almost every item is unique. This makes efficient logistics key. They used to operate out of 2 warehouses, and they would charge the same for shipping from either warehouse, until some engineer pointed out that they would actually lose money if the customer ordered a low value item from the warehouse farthest from them. So then they shifted to subtly only show/promote you items that are located in the nearest warehouse to you.
* Customers typically browse ThredUp while waiting in line. This is something you do to fill the time, like Twitter or Facebook. They find most purchases are impulse and average time on site/app is very low. This might be why growth isn't as good right now
Those policies often work through gradual tightening, rather than a sudden change. So it might start with a 7% + CPI cap, then a few years later they'll change it to 5%, then remove the CPI, and so on. Sometimes the cap is allowed to be changed by a simple committee vote or determined by a "Rent Board," which is easier than passing a new law. That way the initial law can get passed without too much opposition.
The best part of Zoom for deaf people is the automatic captioning.
There is one big downside of using generic video chat for signing - it’s usually optimized for clear audio, not clear video. When the connection gets weak, Zoom will drop frames but try to preserve audio. For the deaf, dropping audio is of course not a problem but dropping frames means they miss a key word in the sentence.
I’ve looked pretty hard at moving my fairly large Postgres RDS to Aurora and the pricing is just very scary. I have a single 100GB instance running on fairly minimal hardware and serving a Django site with 6k monthly visitors. My usage is pretty concentrated during 9-6pm EST as I mainly serve US law firms and investment funds. Any idea if that will end up costing me a fortune to go Aurora? I’m worried about all of those requests in a short amount of time leading to it spinning up a lot of instances.
My estimate is the death toll from Covid is saving trillions in future benefit obligations for the US overall. From a GAAP perspective, the epidemic might come out a wash. This contrasts with the 1918 pandemic, which mainly killed the young and productive.