You can practice on https://leetcode.com/

leetcode has a bunch of dynamic programming problems listed as medium difficulty

Am I the only one who finds those problems extremely difficult? I have a hard time understanding the solutions once they're presented to me let alone solving them on my own.

I did everything on interview cake on a white board. He gives hints and explanations, works pretty well. You can get most for free off the Facebook page posts

I think these are great ideas. I think most people struggle with the issue of white boarding and being on the spot.

I hate name and shaming, but _ _ _ _ _ was easily the most low-ball I ever heard about from a recruiter contact.

- Relo to San Jose w/ no assistance.

- Do DBA work for multiple db instances. The list is very long for this stuff, and add in some Linux admin stuff. I'm guessing they have genuine big data, so very real problems here.

- ETL, etc.

- Write stored procedures, secure back-end code for programmers to connect to, etc.

- Swap 24 hr on-call shifts every few weeks.

- 3 to 6 month contract.

I know the numbers that are tossed around here for far less. I will only say that you should divide your number by 4 and you are still over-shooting.

It didn't matter since I wasn't able to do all of that.

Thanks for this insider summary - I've long suspected a lot of boot camps or DeVry-type courses were like this. Keeping that $10k/student revenue stream going >>>>> everything else.

The one I'm at apparently has a 15% acceptance rate (which I'm not sure I believe, but its been repeated by people in my class so i guess?), and I had absolutely no idea how one person was in it. This person just had absolutely no chance. Fortunately for them and the class they dropped out last week. We lost 2 out of 20 people, the other one was solid tho and definitely average, I think the person just freaked out at how difficult some stuff was and how independent you are expected to be.

When they drop out are they still on the hook for full tuition?

That's the scam at DeVry, I avoided it by being an overachiever!

Just the % of the course they completed. So last week was 1/4 of the way through the 15k.

A quick SQL question - specifically mysql maybe.

I have a huge table, and I want to generate counts of things that match a query, grouped by subtype. Like, say
select type, count(*) from mytable where (query)

This is WAY slower than it should be, partly because query contains a lot of things that can not be indexed. But really what I want to know is "are there results for each type", i.e. I am not concerned with the real number. I want to know if it's, like, less than 10 or more than 10.

Is there a way to do this? If it was just one count I wanted I'd do
select count(*) from table where (query) limit 11
and this would usually be super fast

I guess what I kind of want is a limit on each count. Preferably without having to break it out into N queries (one for each type).

I think you can do something like

You'll get one row for each type that exists. How fast it's going to be is going to depend on your where clause I think, but with no where clause it will be fast as long as there is an index on type.

I was about to say the same thing

seems like what you're looking for

That will just tell me if there's 0 or 1. I want to know the count of them, up to some arbitrary number N. So, like, 0, 1, 2 ... 10 or more than 10.

I left the group by out of my query, sorry, but yeah, doing a group by on the table is not an option. It's way too large for me to do a full table scan, which is what that would require.

Oh my... why?

Without seeing the query, you may want to consider using a pattern like this (assuming you can't do what was already suggested):

As far as I know, using where exists is universally faster, so should work with MySQL.

But my intuition tells me that you may be best off using pure joins all the way through, like so:

I see. I misunderstood what you meant here...

Dave's second example seems like it may be closer to what you want.

Let's say for the time being that there aren't other tables. And I feel like maybe I've expressed myself badly. I want to know the count for each type, up to N, as fast as possible. So let's say I have 3 Xs, 29 Ys and 152 Zs. I want
X | 3
Y | 11
Z | 11
(the 11s are because I want to know "more than 10")

Anything that requires me to do a group by on the full table is really kind of a non-starter. There could be 10s of millions of results.

Like the fastest way that I know of, I think, would be to do
(select 'X', count(*) from mytable where type='X' and (query) limit 11)
union
(select 'Y', count(*) from mytable where type='Y' and (query) limit 11)
union
(select 'Z', count(*) from mytable where type='Z' and (query) limit 11)

I guess I was hoping there was some semantic like
"select type, count(* up to 11) ... group by type"
where it would stop counting once it hit a threshold.

The problem with unions, especially as you don't have CTE's in MySQL, is that you will end up scanning the table 3x for each query. You can circumvent this issue by initializing into a temp table that is far smaller, but something like this should also do the trick, using stop-gap in each subquery:

Probably have to mess around with this a bit to get it really up to speed. You may have to change "and" to "or" if a value does not exist.

Couldn't you do something like
select type from table where count(type) > 10;

But you wanted all types, can you short circuit SQL and an OR count(type) <= 10?

ugh. I need to IP ban someone from my ubuntu webserver but through nginx reverse proxy to my node/express/passport all IPs are coming as 127.0.0.1, any ideas? Also need to figure out how a passport user is associated to an IP... I'm bad at this.

You should be able to add that to the log message in nginx.config, it's been over a year since I did any nginx'ing but I'm pretty sure we were able to get source ip into the log.

In your nginx config file add:

proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $remote_addr;

then on your box you should see these. node will automagically use this over the proxy ip address if its there.

or you can just block him right on your proxy server:
make sure ngx_http_access_module is installed

then just add

deny {ip address};

to your config file

For some reason 2p2 loads lighting fast in Argentina, but FB takes forever. You'd think it would be the other way around.

So, some follow up, they sent me the schedule, basically I need to be there from 10:00 to 3:30? Which means 5 and a half hours?? How can an interview ever last that long? Has anyone had a marathon interview like this before?