Ask General Programming Questions Not Worth Their Own Thread

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Cavern of COBOL > Ask General Programming Questions Not Worth Their Own Thread

«‹›672 »

Volguus: Mar 3, 2009

Hmm, that could work. To be honest though, rather than doing that, for my case and my usage so far, I'd rather inconvenience those 3 users. The Bar message would look ugly and scream "Bad decisions were made, now we live with regrets".
Declaring that a Bar is invalid when there are no Foos present is enough for me right now, as it makes sense from a business logic point as well. And when a single Foo is sent, then there are no Foos in the Bar.

Would gRPC solve this problem better? I've never used it (and I believe there aren't C embedded libraries for it, but i could be wrong there), but could one declare something like:

code:

Response processFoo(Foo foo);

and later on just come and say:

code:

Response processBar(Bar bar);

as a brand new method while leaving the old one intact?

# ? Jan 21, 2018 01:20

Adbot: ADBOT LOVES YOU

# ? Jun 7, 2024 22:17

Jabor: Jul 16, 2010; #1 Loser at SpaceChem

Worth noting that it wouldn't necessarily be a long-term thing - after enough time that you expect all clients to have updated (and/or you're willing to stop supporting the ones that haven't), you delete the backcompat logic and change the definition to:

code:


message Bar {
  repeated Foo foo = 3;
  reserved 1, 2;
}

It's not unusual to see reserved tag numbers from previous api versions - it's just an acknowledgment that apis change over time. If you get in the habit of making backwards-compatible api changes now when it's largely inconsequential, you'll find it easier to do later on when it's important to get right.

That said if you control all the code that calls this api, or you're willing to immediately break backwards compatibility (and can ensure that all clients are going to update), you totally can just start fresh.

# ? Jan 21, 2018 01:34

Volguus: Mar 3, 2009

Jabor posted:

Worth noting that it wouldn't necessarily be a long-term thing - after enough time that you expect all clients to have updated (and/or you're willing to stop supporting the ones that haven't), you delete the backcompat logic and change the definition to:
code:
message Bar {
  repeated Foo foo = 3;
  reserved 1, 2;
}
It's not unusual to see reserved tag numbers from previous api versions - it's just an acknowledgment that apis change over time. If you get in the habit of making backwards-compatible api changes now when it's largely inconsequential, you'll find it easier to do later on when it's important to get right.

That said if you control all the code that calls this api, or you're willing to immediately break backwards compatibility (and can ensure that all clients are going to update), you totally can just start fresh.

Yes, that's good advice, thanks, I'll keep that in mind. While I do not control all the code that calls this API, and I wanted to not inconvenience those few clients ... theyr'e very few clients and they know they're testing my API (beta and all that) so I just may start fresh and tell them to change their stuff. 2 developers are chinese and one is indian, I have no idea how they're gonna react, but I can tell the business guy to just make it work.

# ? Jan 21, 2018 01:41

TooMuchAbstraction: Oct 14, 2012; I spent four years making
Waves of Steel
Hell yes I'm going to turn my avatar into an ad for it.; Fun Shoe

Volguus posted:

Would gRPC solve this problem better? I've never used it (and I believe there aren't C embedded libraries for it, but i could be wrong there), but could one declare something like:
code:
Response processFoo(Foo foo);
and later on just come and say:
code:
Response processBar(Bar bar);
as a brand new method while leaving the old one intact?

That was my original suggestion: make a new API method with the desired signature, and delete the old method once people have had sufficient time to switch over.

...oh, was the reason you were having problems because you were just getting a binary blob on a socket and had to determine what its contents were at runtime? Yeah, gRPC would handle automatically dispatching to the appropriate server method depending on what proto you receive. At least, it does that for me in Python.

# ? Jan 21, 2018 02:16

Volguus: Mar 3, 2009

TooMuchAbstraction posted:

That was my original suggestion: make a new API method with the desired signature, and delete the old method once people have had sufficient time to switch over.

...oh, was the reason you were having problems because you were just getting a binary blob on a socket and had to determine what its contents were at runtime? Yeah, gRPC would handle automatically dispatching to the appropriate server method depending on what proto you receive. At least, it does that for me in Python.

No, I am not using gRPC. One of the reasons is because I had to write an implementation in FreeRTOS on an ARC board using both lwip and bluetooth as communication mechanisms, and while looking around it didn't seem that gRPC and nanopb were working together. So, I abandoned the idea completely and I just read the message directly now.

The https://github.com/nanopb/nanopb/pull/248 pull though does gives me hope of this being implemented in the future, at which point (with a different endpoint than now, of course) i can probably provide a gRPC server.

# ? Jan 21, 2018 05:42

NihilCredo: Jun 6, 2011; iram omni possibili modo preme:
plus una illa te diffamabit, quam multæ virtutes commendabunt

Speaking of APIs:

Say I have an endpoint where users POST commands. My web application takes that command and ships it to a separate service, which will usually process it immediately but sometimes it will be slow.

My first idea will be having my endpoint do this:

1) Ship command, start idling while listening to the service for 'processing complete!' notification
2) If the notification arrives within, let's say, 1000ms (maybe there's a standard header to let the user set the timeout?), return HTTP 200 with the outcome
3) If 1000ms pass without outcome, return HTTP 303 pointing to the /commandStatus/{commandId} endpoint where the user can check at his leisure for the outcome of his command

But then I thought, why not just always create the resource at the status endpoint and always return HTTP 303, even immediately

Are there advantages to either approach? Any aspects I haven't considered?

# ? Jan 23, 2018 16:57

TooMuchAbstraction: Oct 14, 2012; I spent four years making
Waves of Steel
Hell yes I'm going to turn my avatar into an ad for it.; Fun Shoe

What's the advantage of the 303 redirect? How is it preferable to just having an HTTP POST that takes a long time? Especially if the user can know which requests will take a long time in advance, they can just configure the timeout on their request accordingly, right?

# ? Jan 23, 2018 17:09

Thermopyle: Jul 1, 2003; ...the stupid are cocksure while the intelligent are full of doubt. �Bertrand Russell

NihilCredo posted:

Speaking of APIs:

Say I have an endpoint where users POST commands. My web application takes that command and ships it to a separate service, which will usually process it immediately but sometimes it will be slow.

My first idea will be having my endpoint do this:

1) Ship command, start idling while listening to the service for 'processing complete!' notification
2) If the notification arrives within, let's say, 1000ms (maybe there's a standard header to let the user set the timeout?), return HTTP 200 with the outcome
3) If 1000ms pass without outcome, return HTTP 303 pointing to the /commandStatus/{commandId} endpoint where the user can check at his leisure for the outcome of his command

But then I thought, why not just always create the resource at the status endpoint and always return HTTP 303, even immediately

Are there advantages to either approach? Any aspects I haven't considered?

If you're going to do a solution where you don't have a long-running POST request, I would always return an url or whatever for the client to check for results. Or you could do push notifications with websockets.

By always returning the url to check for results you simplify your code and your clients code. Neither has to branch based upon whether we're doing a real result or an async result...it's always an async result.

# ? Jan 23, 2018 17:10

Blinkz0rz: May 27, 2001; MY CONTEMPT FOR MY OWN EMPLOYEES IS ONLY MATCHED BY MY LOVE FOR TOM BRADY'S SWEATY MAGA BALLS

How long is the request going to take? If it's on the order of seconds > 5 I'd suggest returning a 202 while the external request is outstanding is probably the preferred method.

Once data is returned, persist it and make the response available with a 200 the next time the same resource is requested.

Blinkz0rz fucked around with this message at 18:03 on Jan 23, 2018

# ? Jan 23, 2018 18:01

Mr Shiny Pants: Nov 12, 2012

NihilCredo posted:

Speaking of APIs:

Say I have an endpoint where users POST commands. My web application takes that command and ships it to a separate service, which will usually process it immediately but sometimes it will be slow.

My first idea will be having my endpoint do this:

1) Ship command, start idling while listening to the service for 'processing complete!' notification
2) If the notification arrives within, let's say, 1000ms (maybe there's a standard header to let the user set the timeout?), return HTTP 200 with the outcome
3) If 1000ms pass without outcome, return HTTP 303 pointing to the /commandStatus/{commandId} endpoint where the user can check at his leisure for the outcome of his command

But then I thought, why not just always create the resource at the status endpoint and always return HTTP 303, even immediately

Are there advantages to either approach? Any aspects I haven't considered?

Like above mentioned, something is to be said about being consistent. So always returning a status that the client can check, even if the command has already completed, sounds wise.

It will also make it easier for the consumers of your API , there is just one flow in the program instead of branches that will need different logic.

# ? Jan 23, 2018 19:10

NihilCredo: Jun 6, 2011; iram omni possibili modo preme:
plus una illa te diffamabit, quam multæ virtutes commendabunt

TooMuchAbstraction posted:

What's the advantage of the 303 redirect?

It seemed like the most appropriate response code to me.

https://httpstatuses.com/303 posted:

The server is redirecting the user agent to a different resource, as indicated by a URI in the Location header field, which is intended to provide an indirect response to the original request.

quote:

How is it preferable to just having an HTTP POST that takes a long time? Especially if the user can know which requests will take a long time in advance, they can just configure the timeout on their request accordingly, right?

I thought that having a slow-rear end POST would be bad UX? You'd have no idea if it's slow because it's actually working on it or because the web application just froze up somewhere, and you wouldn't have a confirmation that your command was actually received.

Thermopyle posted:

If you're going to do a solution where you don't have a long-running POST request, I would always return an url or whatever for the client to check for results. Or you could do push notifications with websockets.

By always returning the url to check for results you simplify your code and your clients code. Neither has to branch based upon whether we're doing a real result or an async result...it's always an async result.

Yeah, that was my thinking basically.

I've only glanced at websockets before, and mentally filed them away in a folder labelled 'try this if the calls grow numerous enough to make the client lag'.

Blinkz0rz posted:

How long is the request going to take? If it's on the order of seconds > 5 I'd suggest returning a 202 while the external request is outstanding is probably the preferred method.

Once data is returned, persist it and make the response available with a 200 the next time the same resource is requested.

These are POST RPC commands that result in changes to the system, and the user provides a unique UUID for each command, so not really applicable. I'll keep this suggestion in mind though in case I ever end up with a GET resource request that takes forever.

# ? Jan 23, 2018 19:15

redleader: Aug 18, 2005; Engage according to operational parameters

The convention seems to be to return 202 with a Location for the actual resource. You can return 303 from the 'queue job' endpoint once your job has finished and the resource is available.

# ? Jan 23, 2018 20:44

Pollyanna: Mar 5, 2005; Milk's on them.

What's the best way to work with tabular data and data analysis without relying on a database like SQLite or a big ol' framework like Rails? I had an interview question today that in retrospect would have been much more easily answered by using SQL, except working with tab-separated values in a file. The problem was basically a couple of tables and some questions that amounted to AVG, MAX, etc. in SQL along with some join trickery and grouping, but I opted for a brute force Ruby method since that's what I'm most comfortable answering questions like those in.

# ? Jan 24, 2018 23:58

TooMuchAbstraction: Oct 14, 2012; I spent four years making
Waves of Steel
Hell yes I'm going to turn my avatar into an ad for it.; Fun Shoe

Can't sqlite be backed by a flat file? So you can say "here, have this CSV, now I want to do SELECT foo, bar FROM my_file.csv WHERE foo > 10".

# ? Jan 25, 2018 00:07

spiritual bypass: Feb 19, 2008; Grimey Drawer

Loading to in-memory SQLite is usually pretty quick, but there's also this Python thing if you really don't want to do that: https://pythonhosted.org/querycsv/

# ? Jan 25, 2018 00:10

Nippashish: Nov 2, 2005; Let me see you dance!

Pollyanna posted:

What's the best way to work with tabular data and data analysis without relying on a database like SQLite or a big ol' framework like Rails?

My goto for things like this is "use python and load it in pandas". After that its just sql by different syntax.

# ? Jan 25, 2018 00:10

Pollyanna: Mar 5, 2005; Milk's on them.

I don't think I've ever loaded tab-separated values into SQLite before. Maybe SQL really is the best option here...my solution ultimately worked, but I think I beefed it due to how long it took. If it had been a take-home, then I would have been calmer and been more open to exploring the SQL option instead of feeling pressured to brute-force it in Ruby. Oh well, next time.

# ? Jan 25, 2018 00:27

Mr. Apollo: Nov 8, 2000

Not programming per se but does anyone have any recommended reading material for dealing with Bluetooth beacons (iBeacon, Eddystone, etc.). I�m looking for stuff like best practices, figuring out how many you need to cover a space to act as an indoor positioning system (I�m guessing it depends on how accurate you want it to be), security, management, stuff like that.

Everything I�ve come across so far is just a marketing pitch for products or services.

# ? Jan 27, 2018 04:24

downout: Jul 6, 2009

Mr. Apollo posted:

Not programming per se but does anyone have any recommended reading material for dealing with Bluetooth beacons (iBeacon, Eddystone, etc.). I�m looking for stuff like best practices, figuring out how many you need to cover a space to act as an indoor positioning system (I�m guessing it depends on how accurate you want it to be), security, management, stuff like that.

Everything I�ve come across so far is just a marketing pitch for products or services.

I can't help you but your av is pretty sweet.

Edit: to ask a question ->

Does anyone have suggestions for an application that would provide forms that can provide validation (formio style) and allow users to go back and reference entries and edit them? Basically an easily modifiable/editable validation grid?

downout fucked around with this message at 06:04 on Jan 27, 2018

# ? Jan 27, 2018 06:02

Volguus: Mar 3, 2009

downout posted:

I can't help you but your av is pretty sweet.

Edit: to ask a question ->

Does anyone have suggestions for an application that would provide forms that can provide validation (formio style) and allow users to go back and reference entries and edit them? Basically an easily modifiable/editable validation grid?

First time I've heard about form.io. Looking at their page ... what is it exactly that they're providing here? A form designer? They call it "application"? Whose application? Where is it hosted? Where's the database? Who sees the data? How ... what? It looks extremely confusing.

# ? Jan 27, 2018 06:43

downout: Jul 6, 2009

Volguus posted:

First time I've heard about form.io. Looking at their page ... what is it exactly that they're providing here? A form designer? They call it "application"? Whose application? Where is it hosted? Where's the database? Who sees the data? How ... what? It looks extremely confusing.

Formio is a drag-n-drop form builder tool. Each section on the form can be customized with validation rules and conditional functionality based on previous sections or entries. If you get the free account, then you can see all the tools including the data which is just records of form entries. Each record is displayed in a grid. When you edit the entry it takes you back to the form with the fields filled with the entries details. The forms can be embedded in an external sites page (Hubspot style) with custom css to change the visual.

I think they provide an API to access the records from external sources, and a lot of other stuff I haven't explored yet.

So I need pretty much that, but a way to allow entry users to view their entries and edit them. Formio is pretty close, but I'm trying to explore options.

# ? Jan 27, 2018 14:42

Jaded Burnout: Jul 10, 2004

What's the state of system languages these days? Is Go still bad because generics? Is Rust still for weirdos? Did JBlow release his language yet?

Educate a poor Ruby dev please.

# ? Jan 31, 2018 14:47

leper khan: Dec 28, 2010; Honest to god thinks Half Life 2 is a bad game. But at least he likes Monster Hunter.

Jaded Burnout posted:

What's the state of system languages these days? Is Go still bad because generics? Is Rust still for weirdos? Did JBlow release his language yet?

Educate a poor Ruby dev please.

C++ still going strong.

Don�t think Blow has released his language, but I think he�s still making YT vids of its development.

# ? Jan 31, 2018 14:53

taqueso: Mar 8, 2004

Rust seems to be staying mostly cool and good.

# ? Jan 31, 2018 14:55

Jaded Burnout: Jul 10, 2004

leper khan posted:

C++ still going strong.

I did scoop a handful of C++ and it seemed like an acquired taste.

taqueso posted:

Rust seems to be staying mostly cool and good.

What do you feel makes it cool and/or good?

# ? Jan 31, 2018 14:57

taqueso: Mar 8, 2004

Jaded Burnout posted:

What do you feel makes it cool and/or good?

I guess I'll start with the bad - It forces you to define how the data is allowed to be accessed and who owns the data, which can be easy if you aren't doing anything special, and a bit tricky if you are. This is the hardest part of learning the language, or at least it has been difficult for me. However, forcing you to define that ensures you can only access the data in a safe manner. It allows you to write code at a C level (or C++) while maintaining safety. And you get modern abstractions and language features. It is incredibly liberating to know the C-type footguns I'm used to dealing with aren't things I need to worry about.

The abstractions are chosen to fit nicely at a lowish level and tend to be very lightweight. Heavyweight abstractions are provided by libraries. The interface system makes a lot of sense and is easy to reason about. The people making the language are responsive and really seem to be trying to make a good, useful language based on real-world user feedback. The tooling is excellent. The docs are excellent. The community is excellent.

# ? Jan 31, 2018 15:23

Love Stole the Day: Nov 4, 2012; Please give me free quality professional advice so I can be a baby about it and insult you

Recently visited a friend who had that Amazon Alexa thing and had one of those hacker seizures where a million ideas appear out of nowhere. Does the Alexa thing and similar products use a common API for its NLP stuff that I can go read up on and try hacking with or is it all proprietary where you need to pay to hack and to already have bought one of their things beforehand?

One such crazy idea was to do a Reboot-style "Glitch! Command line!" and perform bang searches on DuckDuckGo or whatever, then reroute the results to a screen of my choice. Could use voice commands to search for legislation on Thomas and display the text on TV or tablet... or to an !endic search on DuckDuckGo to translate a word to Korean.

Love Stole the Day fucked around with this message at 22:18 on Jan 31, 2018

# ? Jan 31, 2018 22:13

The Fool: Oct 16, 2003

Love Stole the Day posted:

Recently visited a friend who had that Amazon Alexa thing and had one of those hacker seizures where a million ideas appear out of nowhere. Does the Alexa thing and similar products use a common API for its NLP stuff that I can go read up on and try hacking with or is it all proprietary where you need to pay to hack and to already have bought one of their things beforehand?

One such crazy idea was to do a Reboot-style "Glitch! Command line!" and perform bang searches on DuckDuckGo or whatever, then reroute the results to a screen of my choice. Could use voice commands to search for legislation on Thomas and display the text on TV or tablet... or to an !endic search on DuckDuckGo to translate a word to Korean.

Start here: https://developer.amazon.com/alexa-skills-kit

e: all of your ideas sound possible, but you will likely just be using an alexa skill to forward the command to a rPi and doing the heavy lifting there.

The Fool fucked around with this message at 22:29 on Jan 31, 2018

# ? Jan 31, 2018 22:24

TooMuchAbstraction: Oct 14, 2012; I spent four years making
Waves of Steel
Hell yes I'm going to turn my avatar into an ad for it.; Fun Shoe

I'm using an HTML canvas with a Javascript script for visualization of the output of a procedural map generator. However, for some reason I only can have one map in a given file. Can anyone tell me why this HTML doesn't draw anything on the second canvas?

# ? Feb 2, 2018 15:35

nielsm: Jun 1, 2009

I'm not sure exactly why it fails, but if I wrap each canvas in a block element (like <p>) it displays all of them correctly.

https://jsfiddle.net/cec2x4xp/

# ? Feb 2, 2018 15:45

TooMuchAbstraction: Oct 14, 2012; I spent four years making
Waves of Steel
Hell yes I'm going to turn my avatar into an ad for it.; Fun Shoe

Interesting, I wonder why that happens? Excellent to have a workaround though, so thank you very much!

# ? Feb 2, 2018 16:30

Volguus: Mar 3, 2009

TooMuchAbstraction posted:

I'm using an HTML canvas with a Javascript script for visualization of the output of a procedural map generator. However, for some reason I only can have one map in a given file. Can anyone tell me why this HTML doesn't draw anything on the second canvas?

This:

code:

<canvas id="2" width="82" height="62" stroke="black"/>

should become this:

code:

<canvas id="2" width="82" height="62" stroke="black">
</canvas>

everywhere.

# ? Feb 2, 2018 17:14

nielsm: Jun 1, 2009

That's a good point, <canvas /> is only valid as open + close if you're actually in XML mode. In standard HTML mode the / is just a stray character the parser ignores for you.

# ? Feb 2, 2018 17:22

TooMuchAbstraction: Oct 14, 2012; I spent four years making
Waves of Steel
Hell yes I'm going to turn my avatar into an ad for it.; Fun Shoe

Aha, that'd do it. So I guess in my original setup, the parser was opening a canvas, then leaving it open until the end of document. That'd explain why it wouldn't let me open a second canvas.

# ? Feb 2, 2018 17:26

shovelbum: Oct 21, 2010; Fun Shoe

Is MUMPS still used outside of Epic? Like did everyone who works in this language learn it there?

# ? Feb 7, 2018 16:50

csammis: Aug 26, 2003; Mental Institution

I've never heard about it outside the context of Epic, and I worked in and around healthcare EMR vendors for several years. Doesn't mean it's not out there though. I'd be surprised if there weren't ex-Epic employees going around offering consulting services in and about MUMPS.

# ? Feb 7, 2018 20:08

School of How: Jul 6, 2013; quite frankly I don't believe this talk about the market

Imagine there is a slot machine. The cost to pull the handle is $15. The slot machine only has 3 results, you get either prize #1 worth $13, prize #2 worth $20, or prize #3 which is worth $100. The machine needs to be designed to assign prizes based on a probability that the machine can be operated profitably. In order for the machine to be profitable, prize #1 must be the most common, and prize #3 must be the least common prize. How can I calculate the probability of each prize in order to maintain break even? Is there such a formula I can plug the prize values into to get each probability? Maybe this is more of a math problem than a computer problem, but I need some python code that calculates this probability.

# ? Feb 8, 2018 20:26

nielsm: Jun 1, 2009

As the machine owner, you get a profit of $2 for every time prize #1 is won, a loss of $5 when prize #2 is won, and a loss of $85 when prize #3 is won. If there were just prizes #1 and #2, how would you calculate how many times you need to produce result #1 to make up for one result #2 ? Try writing it as an equation to solve.

# ? Feb 8, 2018 20:55

LOOK I AM A TURTLE: May 22, 2003; "I'm actually a tortoise."; Grimey Drawer

School of How posted:

Imagine there is a slot machine. The cost to pull the handle is $15. The slot machine only has 3 results, you get either prize #1 worth $13, prize #2 worth $20, or prize #3 which is worth $100. The machine needs to be designed to assign prizes based on a probability that the machine can be operated profitably. In order for the machine to be profitable, prize #1 must be the most common, and prize #3 must be the least common prize. How can I calculate the probability of each prize in order to maintain break even? Is there such a formula I can plug the prize values into to get each probability? Maybe this is more of a math problem than a computer problem, but I need some python code that calculates this probability.

There's an infinite number of different probabilities that will satisfy this. The general case of what you're looking would be something like this:

Given a set of prizes in ascending order of value: n₀, n₁, ..., n_n, find probabilities p₀, p₁, ..., p_i, where the sum of all p_n is in the range [0, 1] and p_n < p_n-1 for all n, such that p₀n₀ + p₁n₁ + ... + p_in_i = b, where b is the size of the bet. Since all p_n are unknown, there's no single solution to this equation.

In your particular case, one example solution that breaks even would be that $13 happens with probability 40%, $20 happens with probability 20%, and $100 happens with probability 5.8%. To calculate this I just made up some probabilities for prize 1 and 2, and then solved for p in the following equation to find the break-even probability for prize 3: 13*0.4 + 20*0.2 + 100*p = 15. Another solution would be 50%, 15% and 5.5%.

# ? Feb 8, 2018 21:20

Adbot: ADBOT LOVES YOU

# ? Jun 7, 2024 22:17

nielsm: Jun 1, 2009

LOOK I AM A TURTLE posted:

Another solution would be 50%, 15% and 5.5%.

What happens the remaining 29.5% of the time?

# ? Feb 8, 2018 21:42

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > The Cavern of COBOL > Ask General Programming Questions Not Worth Their Own Thread

«‹›672 »