If all you have is a hammer, everything looks like a nail!

This was something which my tech lead used to tell me when I was so obsessed with nosql databases a few years back. I would try to find problems to solve that has a use case for nosql databases or even try to convince me(I didn’t realise it back then) that I need to use nosql db for this new idea that I have, without really thinking deep enough whether the data in question is better represented using an sql schema or not.

Now, leading a team of young developers, I come across similar suggestions from few of my team members who just discovered this new and shiny tech and want to use it in production projects.

While I am not against new and shiny, it’s not a good practice to jump right in to it without exploring it deep enough or considering all the shortcomings. The most important question to ask is, whether some of the problems you are trying to solve can be solved with the current stack.

Modifying your stack requires more than just a week’s experience of playing around with the getting started guide and stack overflow replies. This is something which need to be carefully considered after taking inputs from the people who would be supporting it, that include operations, sysadmins and teams that are gonna interface with your stack indirectly.

I am not talking about delaying adoption by waiting for long list of approvals to get some thing that would bring immediate value, but a carefully orchestrated plan for why and how to migrate to a new stack.

Just because one of the tech giants made a move to a new stack and wrote about it in their engineering blog doesn’t mean that you need to make a switch in the same direction. Take a moment to analyse the possible reasons that motivated them to do it, ask yourself if your organisation is struggling with the exact same problems, observe how others facing the same issue are addressing it, and then make an informed decision.
Collect enough data to support your proposal.
Ask yourself again if you are the one holding the hammer.
If the answer is no, forge ahead!

I have been gone a while. Sorry. Workplace no longer allows phones on the lab and I work exclusively in the lab. Anyway here is a thing that pissed me off:

Systems Engineer (SE) 1 : 😐 So we have this file from the customer.
Me: 😑 Neat.
SE1: 😐 It passes on our system.
Me: 😑 *see prior*
Inner Me (IM): 🙄 is it taught in systems engineer school to talk one sentence at a time? It sounds exhausting.
SE1: but when we test it on your system, it fails. And we share the same algorithms.
Me: 😮 neat.
IM: 😮neat, 😥 wait what the fuck?
Me: 😎 I will totally look into that . . .
IM: 😨 . . . Thing that is absolutely not supposed to happen.

*Le me tracking down the thing and fixing it. Total work time 30 hours*

Me: 😃 So I found the problem and fixed it. All that needs to happen is for review board to approve the issue ticket.
SE1: 😀 cool. What was the problem?
Me: 😌 simple. See, if the user kicked off a rerun of the algorithm, we took your inputs, processed them, and put them in the algorithm. However, we erroneously subtracted 1 twice, where you only subtract 1 once.
SE1: 🙂 makes sense to me, since an erroneous minus 1 only effects 0.0001% of cases.

*le into review board*

Me: 😐 . . . so in conclusion this only happens in 0.0001% of cases. It has never affected a field test and if this user had followed the user training this would never have been revealed.
SE2: 🤨 So you're saying this has been in the software for how long?
Me: 😐 6 years. Literally the lifespan of this product.
SE2: 🤨 How do you know it's not fielded?
Me: 😐 It is fielded.
SE2: 🤨 how do you know that this problem hasn't been seen in the field?
Me: 😐 it hasn't been seen in 6 years?
IM: 😡 see literally all of the goddamn words I have said this entire fucking meeting!!!
SE2: 😐 I would like to see an analysis of this to see if it is getting sent to the final files.
Me: 🙄 it is if they rerun the algorithm from our product. It's a total rerun, output included. It's just never been a problem til this one super edge case that should have been thrown out anyway.
SE2: 🤨 I would still like to have SE3 run an analysis.
Me: 🙄 k.
IM: 😡 FUUUUUUUUUCK YOOOOOU

*SE3 run analysis*

SE3: 😐 getting the same results that Me is seeing.
Me: 😒 see? I do my due diligence.
SE2: 😐 Can you run that analysis on this file again that is somehow different, plus these 5 unrelated files?
SE3: 😎 sure. What's your program's account so I can bill it?
IM: 😍 did you ever knooooow that your my heeeerooooooo.

*SE3 runs analysis*

SE3: 😐 only the case that was broken is breaking.
SE2: 😐 Good.
IM: 🤬🤬🤬🤐 . . . 🤯WHY!?!?
Me: 😠 Why?
SE2: 😑 Because it confirms my thoughts. Me, I am inviting you to this algorithm meeting we have.
Me/IM: 😑/😡 what . . . the fuck?

*in algorithm meeting*

Me: 😑 *recaps all of the above* we subtract 1 one too many times from a number that spans from 10000 to -10000.
Software people/my boss/SE1/SE3: 🤔 makes sense.
SE2:🤨 I have slides that have an analysis of what Me just said. They will only take an hour to get through.
Me: 😑 that's cool but you need to give me your program's account number, because this has been fixed in our baseline for a week and at this point you're the only program that still cares. Actually I need the account to charge for the last couple times you interrupted me for some bullshit.

*we are let go.*

And this is how I spent 40+ useless hours against a program that is currently overrunning for no reason 🤣🤣🤣

Moral: never involve math guys in arithmetic situations. And if you ever feel like you're wasting your time, at least waste someone else's money.

Let the student use their own laptops. Even buy them one instead of having computers on site that no one uses for coding but only for some multiple choice tests and to browse Facebook.

Teach them 10 finger typing. (Don't be too strict and allow for personal preferences.)

Teach them text navigation and editing shortcuts. They should be able to scroll per page, jump to the beginning or end of the line or jump word by word. (I am not talking vi bindings or emacs magic.) And no, key repeat is an antifeature.

Teach them VCS before their first group assignment. Let's be honest, VCS means git nowadays. Yet teach them git != GitHub.

Teach git through the command line. They are allowed to use a gui once they aren't afraid to resolve a merge conflict or to rebase their feature branch against master. Just committing and pushing is not enough.

Teach them test-driven development ASAP. You can even give them assignments with a codebase of failing tests and their job is to make them pass in the beginning. Later require them to write tests themselves.

Don't teach the language, teach concepts. (No, if else and for loops aren't concepts you god-damn amateur! That's just syntax!)

When teaching object oriented programming, I'd smack you if do inane examples with vehicles, cars, bikes and a Mercedes Benz. Or animal, cat and dog for that matter. (I came from a self-taught imperative background. Those examples obfuscate more than they help.) Also, inheritance is overrated in oop teachings.

Functional programming concepts should be taught earlier as its concepts of avoiding side effects and pure functions can benefit even oop code bases. (Also great way to introduce testing, as pure functions take certain inputs and produce one output.)

Focus on one language in the beginning, it need not be Java, but don't confuse students with Java, Python and Ruby in their first year. (Bonus point if the language supports both oop and functional programming.)

And for the love of gawd: let them have a strictly typed language. Why would you teach with JavaScript!?

Use industry standards. Notepad, atom and eclipse might be open source and free; yet JetBrains community editions still best them.

For grades, don't your dare demand for them to write code on paper. (Pseudocode is fine.)

Don't let your students play compiler in their heads. It's not their job to know exactly what exception will be thrown by your contrived example. That's the compilers job to complain about. Rather teach them how to find solutions to these errors.

Teach them advanced google searches.

Teach them how to write a issue for a library on GitHub and similar sites.

Teach them how to ask a good stackoverflow question :>

Inappropriate experience at work: One of our project managers got arrested one day for fraud. Apparently an employee had been in the middle of an online purchase and walked away from their desk. He happened to see the unmasked entry of the CC info (this was before websites cared about masking sensitive form inputs). I guess the temptation was too great…and he was too stupid to realize he’d get caught…and he jotted it all down. He made thousands of dollars in purchases which, naturally, eventually led back to him.

The same guy, before he got arrested, had made a joke when someone in an office team email said “Feel free to have some cake in the break room.” He replied “No need to do anything to me for the cake.” His first name was “Free”.

I often read articles describing developer epiphanies, where they realized, that it was not Eclipse at fault for a bad coding experience, but rather their lack of knowledge and lack of IDE optimization.

No. Just NO.

Eclipse is just horrendous garbage, nothing else. Here are some examples, where you can optimize Eclipse and your workflow all you like and still Eclipse demonstrates how bad of an IDE it is:

- There is a compilation error in the codebase. Eclipse knows this, as it marks the error. Yet in the Problems tab there is absolutely nothing. Not even after clean. Sometimes it logs errors in the problems tab, sometimes t doesn't. Why? Only the lord knows.

- Apart from the fact that navigating multiple Eclipse windows is plain laughable - why is it that to this day eclipse cannot properly manage windows on multi-desktop setups, e.g. via workspace settings? Example: Use 3 monitors, maximize Eclipse windows of one Eclipse instance on all three. Minimize. Then maximize. The windows are no longer maximized, but spread somehow over the monitors. After reboot it is even more laughable. Windows will be just randomly scrabled and stacked on top of each other. But the fact alone that you cannot navigate individual windows of one instance.. is this 2003?

- When you use a window with e.g. class code on a second monitor and your primary Eclipse window is on the first monitor, then some shortcuts won't trigger. E.g. attempting to select, then run a specific configuration via ALT+R, N, select via arrows, ALT+R won't work. Eclipse cannot deal with ALT+R, as it won't be able to focus the window, where the context menus are. One may think, this has to do with Eclipse requiring specific perspectives for specific shortcuts, as shortcuts are associated with perspectives - but no. Because the perspective for both windows is the same, namely Java. It is just that even though Shortcuts in Eclipse are perspective-bound, but they are also context-sensitive, meaning they require specific IDE inputs to work, regarldless of their perspective settings. Is that not provided, then the shortcut will do absolutely nothing and Eclipse won't tell you why.

- The fact alone that shortcut-workarounds are required to terminate launches, even though there is a button mapping this very functionality. Yes this is the only aspect in this list, where optimizing and adjusting the IDE solves the problem, because I can bind a shortcut for launch selection and then can reliably select ant trigger CTRL+F2. Despite that, how I need to first customize shortcuts and bind one that was not specified prior, just to achieve this most basic functionality - teminating a launch - is beyond me.

Eclipse is just overengineered and horrendous garbage. One could think it is being developed by people using Windows XP and a single 1024x768 desktop, as there is NO WAY these issues don't become apparent when regularily working with the IDE.

In 2015 I sent an email to Google labs describing how pareidolia could be implemented algorithmically.
The basis is that a noise function put through a discriminator, could be used to train a generative function.
And now we have transformers.

I also told them if they looked back at the research they would very likely discover that dendrites were analog hubs, not just individual switches. Thats turned out to be true to.

I wrote to them in an email as far back as 2009 that attention was an under-researched topic. In 2017 someone finally got around to writing "attention is all you need."

I wrote that there were very likely basic correlates in the human brain for things like numbers, and simple concepts like color, shape, and basic relationships, that the brain used to bootstrap learning. We found out years later based on research, that this is the case.

I wrote almost a decade ago that personality systems were a means that genes could use to value-seek for efficient behaviors in unknowable environments, a form of adaption. We later found out that is probably true as well.

I came up with the "winning lottery ticket" hypothesis back in 2011, for why certain subgraphs of networks seemed to naturally learn faster than others. I didn't call it that though, it was just a question that arose because of all the "architecture thrashing" I saw in the research, why there were apparent large or marginal gains in slightly different architectures, when we had an explosion of different approaches. It seemed to me the most important difference between countless architectures, was initialization.

This thinking flowed naturally from some ideas about network sparsity (namely that it made no sense that networks should be fully connected, and we could probably train networks by intentionally dropping connections).

All the way back in 2007 I thought this was comparable to masking inputs in training, or a bottleneck architecture, though I didn't think to put an encoder and decoder back to back.

Nevertheless it goes to show, if you follow research real closely, how much low hanging fruit is actually out there to be discovered and worked on.

And to this day, google never fucking once got back to me.

I wonder if anyone ever actually read those emails...

Wait till they figure out "attention is all you need" isn't actually all you need.

p.s. something I read recently got me thinking. Decoders can also be viewed as resolving a manifold closer to an ideal form for some joint distribution. Think of it like your data as points on a balloon (the output of the bottleneck), and decoding as the process of expanding the balloon. In absolute terms, as the balloon expands, your points grow apart, but as long as the datapoints are not uniformly distributed, then *some* points will grow closer together *relatively* even as the surface expands and pushes points apart in the absolute.

In other words, for some symmetry, the encoder and bottleneck introduces an isotropy, and this step also happens to tease out anisotropy, information that was missed or produced by the encoder, which is distortions introduced by the architecture/approach, features of the data that got passed on through the bottleneck, or essentially hidden features.

Angular material is shit.

Angular material has literally no documentation. There is no search on their shit docs. And there is absolutely no info in their docs about some shit they use.
So when you need to know what the fuck does smth like MatOption mean (what inputs does this piece of fuck takes?) you will need to dig into the code.

Got one right now, no idea if it’s the “most” unrealistic, because I’ve been doing this for a while now.

Until recently, I was rewriting a very old, very brittle legacy codebase - we’re talking garbage code from two generations of complete dumbfucks, and hands down the most awful codebase I’ve ever seen. The code itself is quite difficult to describe without seeing it for yourself, but it was written over a period of about a decade by a certifiably insane person, and then maintained and arguably made much worse by a try-hard moron whose only success was making things exponentially harder for his successor to comprehend and maintain. No documentation whatsoever either. One small example of just how fucking stupid these guys were - every function is wrapped in a try catch with an empty catch, variables are declared and redeclared ten times, but never used. Hard coded credentials, hard coded widths and sizes, weird shit like the entire application 500ing if you move a button to another part of the page, or change its width by a pixel, unsanitized inputs, you name it, if it’s a textbook fuck up, it’s in there, and then some.

Because the code is so damn old as well (MySQL 8.0, C#4, and ASP.NET 3), and utterly eschews the vaguest tenets of structured, organized programming - I decided after a month of a disproportionate effort:success ratio, to just extract the SQL queries, sanitize them, and create a new back end and front end that would jointly get things where they need to be, and most importantly, make the application secure, stable, and maintainable. I’m the only developer, but one of the senior employees wrote most of the SQL queries, so I asked for his help in extracting them, to save time. He basically refused, and then told me to make my peace with God if I missed that deadline. Very helpful.

I was making really good time on it too, nearly complete after 60 days of working on it, along with supporting and maintaining the dumpster fire that is the legacy application. Suddenly my phone rings, and I’m told that management wants me to implement a payment processing feature on the site, and because I’ve been so effective at fixing problems thus far, they want to see it inside of a week. I am surprised, because I’ve been regularly communicating my progress and immediate focus to management, so I explain that I might be able to ship the feature by end of Q1, because rather than shoehorn the processor onto the decrepit piece of shit legacy app, it would be far better to just include it in the replacement. I add that PCI compliance is another matter that we must account for, and so there’s not a great chance of shipping this in a week. They tell me that I have a month to do it…and then the Marketing person asks to see my progress and ends up bitching about everything, despite the front end being a pixel perfect reproduction. Despite my making everything mobile responsive, iframe free, secure and encrypted, fast, and void of unpredictable behaviors. I tell her that this is what I was asked to do, and that there should have been no surprises at all, especially since I’ve been sending out weekly updates via email. I guess it needed more suck? But either way, fuck me and my two months of hard work. I mean really, no ego, I made a true enterprise grade app for them.

Short version, I stopped working on the rebuild, and I’m nearly done writing the payment processor as a microservice that I’ll just embed as an iframe, since the legacy build is full of those anyway, and I’m being asked to make bricks without straw. I’m probably glossing over a lot of finer points here too, just because it’s been such an epic of disappointment. The deadline is coming up, and I’m definitely going to make it, now that I have accordingly reduced the scope of work, but this whole thing has just totally pissed me off, and left a bad taste about the organization.

How do you get over the bad times? I keep having to work with shitty legacy systems that were written in perl and flash in the 90s, but my boss keeps telling me "No" on redoing some of the bigger stuff even though it is really needed. I mean, that is your goal here, right? Rebuilding this POS? FFS you still stored passwords in plain text twoo weeks ago! But no, you's rather dig around in Perl than upset some random user because his fucking interface looks different.

But then I also have to work with another system that I could redo in Cake/Laravel in two weeks (it's literally getting and writing data to one table, so two views and user auth), and the previous dev just... made a huge mess. I mean, why would you need to post data asynchronously when it's this one stupid form ? Just do a regular form submit? And the system is really not suitable for extending, because everything is in the database, EVERYTHING! Like, html form inputs? So to add a simple input to the template I have to create a new input type in the types table and then add that to the form structure table? Only to have the input checked by fucking regex? REGEX! Why? Seriously, this is not some high end CMS that needs this level of code reusability No. This is a simple fucking form.

And I can't get it to work. No documentation of course. No comments, either. All of this makes me feel like I'm just the shittiest dev ever. I feel dumb, and useless. Haven't turned on my private PC in weeks because I see no reason to work on any of my own stuff.

I used to have a job, working with Magento and Wordpress. And yeah, it was horrible, it was chaos, but it was fun and I was great at it. I bent that motherfucking system to fit my needs. People respected my opinion, they were convinced I could program this and that, and I proved them right. Did I make mistakes? Hell yeah. Did I give up? Fuck no!

But now, I just feel like I can't even write a simple fucking form any more. I'm just so close to giving up on development as a whole, even though I love it so much.

All day I wrote pretty error mesages for all sorts of nonsensical inputs to a program that only I will ever use.

But now I still need exceptions for the typos in my exceptions for the typos in my code for... what am I trying to do?

How do you switch from testing while debugging (functional) to TDD unit tests?

Usually I test while coding by just running the use case and making sure while coding, bad inputs are caught/handled.

But most times I start with a general idea of the structure and what the about should be (which essentially would be the functional test case?)

I don't think about how you can break each part or the functions I need until I need them. Then usually start simple and then refactor. And until I'm sure each time I refactor would require changing the tests?

Finally finished an algo to check an image for grouping of pixels that will form a rectangular area. I got the grouping to work on one image, but found it was utterly failing on another. I went through every step of the algo and still could not find the solution. The 128x128 image was working, but the 128x16 image was not. I knew it had something to do with the dimensions. Started thinking it was overflowing a buffer somewhere. So I started putting asserts in the functions that abstracted the buffer access. None of the numbers exceeded the proper bounds. It was close to bedtime so I finally gave up. I was tired. Then I realized it wouldn't be until the next evening when I could look at this again. So I got up again and started looking at the code again. I had a loop to check the output of my algo that I did the memory access of the buffer. It too was not fully filling my temp image to show how the algo was working. WTF!

Then I finally realized the flaw:
buffer[x+y*height]

And my test loop to test the algo:
buffer[x+y*ymax]

I kept overlooking the error because I was sure it was right. Also my asserts for the functions to access the buffers? They only checked the inputs x and y. So it didn't help that the math was wrong for reading and writing the buffers. It also worked fine on 128x128 images because the width and height were the same.

It is funny that I struggled with this part. The algo was actually surprisingly easy to formulate. I just looked through every point and checked a buffer to see if that point was used. If not then I would attempt to grow in the x and y direction the shaped of that point based upon pixel color. This was saved in a structure while growing that point. Then when that rectangle could not be grown further the inner loop would continue checking used points again.

I still have work to do to use the data this algo produces. I need to now figure out how to parent the rectangular areas to each other. I will probably use my check buffer to keep track of these rects by an index. Then do adjacent checks to determine parenting. Eventually I will have to extend this algo to 3 dimensions, but that should not be difficult.

Let me just say that I've been playing whack a mole with a new feature for while now. And it's becoming very tiring.

TLDR; CTO is changing the way we're going to implement this, every other day.

June 1st,
CEO: let's implement feature AAA,
CTO: we're going to have a call with Andy to tell us all about his product that will make this super easy, call will be June 4th.

Days before June 4th,
Me: Researchs product X, makes demo works flawlessly.

June 4th,
Call all good, few tips from Andy. We come to the pricing section of Product X

CTO: this will not work, pricing doesn't fit on our budget, fair enough.

June 7th -11th
Me: research altenative approach. Makes second demo.

CTO: Works good, seems to have too many moving parts, let's have call with Bob to check Product Y. It should make our lifes easier.

ME: Geee, ok let's check it out.

June 14th,

Call with Bob, all good, product has a fair price, stuff is experimental.

CTO: let's use Product Y, and just use what we get from their api now, and worry about changes later.

Me: Hmmm, that's a bit risky, but ok, you the boss, right?, starts again new demo. API doesn't work as documented.

Lots of trial and error to figure out how the api is working now, finally demo works well,

June 17th,

API changed, now it works as documented, (expected as it is experimental), previous demo doesn't work anymore.

June 18th,

Redoing research. inputs are completely different from Product Y now, need to redo all that is working and do and a lot more of research.

Go live is scheduled for end of next week, I hope that the API is stable now, and that I get to go live on schedule.

It is funny to see, that it would probably been the same if we just waited on the API to stabilize, and check the pricing section before choosing a product? Who knows.

Anyways, I actually feel happy that over the years I developed the patience to work with ever changing situations like this one.

Can somebody explain to me why developers (especially web) have to micromanage every single thing into it's own f*ing component.

Story time: I have an input form with some tabs. I discovered that the UI Library (Devextreme) has a nice little component that handles forms, (including tabs, groups, etc.). So I make a page, configure tabs, inputs and whatnot.

Now, I already knew that my coworkers can't handle html that is bigger than a page. So instead of putting the configs in the frontend, I made nice files where I store those, to keep them nicely clean and seperated.

Me feeling very good, went off to have a nice lunch break.

I come back read the message from my coworker, asking me to make every tab it's own component and form and load them into a separate Tab-Component, instead of using the built in configuration

......

WHAT?

Like seriously. I have a f*ing library that handles that, why the f*ck do I need to reinvent the wheel here!?

Supposedly it's to make it more maintainable, easier to find bugs, flatten the hierarchy.

Here's a little wake up call you morons: Nesting hundreds of components into each other does *not* help you with that.
It just creates a rabbit-hole of confusing containers that you have to navigate and dissect every time you try to find something.

"Can I fix the bug in the detail Page? Sure I'll tell you tomorrow when I find out which fucking component the bug results from".

Components are there to be *reused*. It's using inheritance for reusing code all over again, but worse.

But maybe I'm just old fashioned, and conservative. Maybe I'm just a really bad software engineer, because nowadays everything seems to result in architectures spreading hundreds of folders, thousands of files with nothing but arbitrary cut-offs with no real benefit, that I don't see the value in.

The frontend developers in my company are the reason why I have anxiety. Here are few things that grinds my knees:
1) for a long time in projects, they deleted the auth token from their storage without integrating the logout api. They thought why use an API for that. :)
2) most of them had no clue that form fields could accept javascript as inputs and work as XSS vulnerabilities. This actually happened with a client, he got so fucking pissed.
3) One of them asked me to convert a PATCH request to DELETE cos fuck REST and HTTP methods.

For fuck’s sake. I need to get out of this place.

How do you think about unit testing/TDD when writing apps? (I'm working this at 3am so might be a bit messy... Just a thought I woke up to).

Whenever I write an app, I don't write unit tests but as I'm developing I may create test functions for specific parts that I run to validate a specific component is working before moving onto the next.

So first, when I get a problem, break it up into components based on the requirements. It's usually sort of input, processor, output sequence.

Where the processor is essentially the core app. And so I start coding it, referring to the input thru an interface, model objects, adding fields as I go along (assume no matter what the input, I will get these before the logic is called). I may add some more interfaces as well for other data I may need but I know won't be going in the first input.

So I write all the logic, functions needed to get a basic app to run that does what I am writing the app for.

Only then do I write a test functions passing in different parameters to make sure the logic and response is what I want and making fixes as necessary. At that point I basically have the simplest version of the app.
(I guess this is sort of like mocking?)

Then build outwards implementing and testing components as I go along and may do some simple refactoring/redesign. (I guess all these tests are functional then, have to start the whole app).

And finally when I have the basic requirements fully complete I will add the "nice to haves" on top via refactoring of specific logic in specific components. Again testing by running the app maybe with simple inputs.

I guess now I'm thinking how do you write unit tests/TDD if the app keeps changing (via adhoc refactorings) as you are creating it?

I’m trying to update a job posting so that it’s not complete BS and deters juniors from applying... but honestly this is so tough... no wonder these posting get so much bs in them...

Maybe devRant community can help be tackle this conundrum.

I am looking for a junior ml engineer. Basically somebody I can offload a bunch of easy menial tasks like “helping data scientists debug their docker containers”, “integrating with 3rd party REST APIs some of our models for governance”, “extend/debug our ci”, “write some preprocessing functions for raw data”. I’m not expecting the person to know any of the tech we are using, but they should at least be competent enough to google what “docker is” or how GitHub actions work. I’ll be reviewing their work anyhow. Also the person should be able to speak to data scientists on topics relating to accuracy metrics and mode inputs/outputs (not so much the deep-end of how the models work).

In my opinion i need either a “mathy person who loves to code” (like me) or a “techy person who’s interested in data science”.

What do you think is a reasonable request for credentials/experience?

You know what's terrifying? When you hear a contracts person say "we already know what we're going to bid, so we don't need engineering inputs."

How do you implement TDD in reality?

Say you have a system that is TDD ready, not too sure what that means exactly but you can go write and run any unit tests.

And for example, you need to generate a report that uses 2 database tables so:

1. Read/Query
2. Processor logic
3. Output to file

So 1 and 3 are fairly straightforward, they don't change much, just mock the inputs.

But what about #2. There's going to be a lot of functions doing calculations, grouping/merging the data. And from my experience the code gets refactored a lot. Changing requirements, optimization (first round is somewhat just make it work) so entire functions and classes maybe deleted. Even the input data may change. So with TDD wouldn't you end up writing a lot of throwaway code?

A lot of times I don't know exactly what I want or need other than I need a class that can do something like this... but then I might end up throwing the whole thing out and writing a new one one I get a clearer idea of what i or the user wants or needs.

Last week I was building a new REST API, the parameters and usage changed like 3 times. And even now the code is in feasibility/POC testing just to figure out what needs to be used. Do I need more, less parameters, what should they be. I've moved and rewritten a lot of code because "oh this way won't work, need to try this way instead"

All I start with is my boss telling me I need an API that lets users to ... (Very general requirements).

I wrote my first proper promise today

I'm building a State-driven, ajax fed Order/Invoice creation UI which Sales Reps use to place purchases for customers over the phone. The backend is a mutated PHP OSCommerce catalog which I've been making strides in refactoring towards OOP/eliminating spahgetti code and the need for a massive bootstrapper file which includes a ton of nonsense (I started by isolating the session and several crucial classes dealing with currency, language and the cart)

I'm using raw JS and jquery with copious reorganization.

I like state driven design, so I write all my data objects as classes using a base class with a simple attribute setter, and then extend the class and define it's attributes as an array which is passed to the parent setter in the construct.

I have also populateFromJson method in the parent class which allows me to match the attribute names to database fields in the backend which returns via ajax.

I achieve the state tracking by placing these objects into an array which underscore.js Observe watches, and that triggers methods to update the DOM or other objects.

Sure, I could do this in react but
1) It's in an admin area where the sales reps using it have to use edge/chrome/Firefox
2) I'm still climbing the react learning curve, so I can rapid prototype in jquery faster instead of getting hung up on something I don't understand
3) said admin area already uses jquery anyway
4) I like a challenge

Implementing promises is quickly turning messy jquery ajax calls into neat organized promise based operations that fit into my state tracking paradigm, so all jquery is responsible for is user interaction events.

The big flaw I want to address is that I'm still making html elements as JS strings to generate inputs/fields into the pseudo-forms.

Can anyone point me in the direction of a library or practice that allows me to generate Dom elements in a template-style manner.

!Rant
#serious inputs needed

I need help identifying that one big problem that all devs face. Irrespective of the language, domain, etc.. And I wanna solve it. I need your inputs on what bothers you the most, and what tech you work with.. Then we can upvote the comments that are applicable to all of us.. I'll setup a github once we come to a conclusion

Any cool buttons and inputs CSS library?
I just need cool / elegant / nice buttons and inputs. If it has some image niceties or something else it's ok.
I did little Googling, ngl; I'll look a bit more later in the day, but I figured maybe someone here may have something interesting to comment.

I need to solve 4x4 tic tac toe using alpha beta pruning, I have solved 3x3 tic tac toe using minimax, and it works very well, but in 4x4 minimax takes lot of time as search space is too big, so I am applying alpha beta pruning but I am not getting clear idea, that upto what depth I should go and what evaluation function should I write in order to give perfect game play, heuristic may also lead to false results and not perfect game play, so how do I ensure perfect game play for 4x4 using fast approach?? Any suggestions or approach will help me a lot. Looking forward to get some inputs.

I'm fucking tired of my computer having random
2 seconds latency on any basic action and being slow as fuck regardless of powerful processor, ssd and 32GB RAM. Music via bluetooth is basically unusable since every few seconds the music stops for a 0.2s then plays again. I installed this system (opensuse tumbleweed) in February this year and it's just sad that I have reinstall again (any ideas for distro) ?

I made a dummy mistake of buying a CPU without internal graphics and this resulted in having to buy a GPU. So I got myself Nvidia(another mistake) since i though i would be using CUDA on the university. Turnes out CUDA cannot be installed for some retarded reason.
With Nvidia GPU the screens on my two monitors are swapping every time I use a hdmi switch to use other computer. On AMD GPU this problem does not exist. AMD GPU pro drivers are impossible to install. Computers barely fucking work, change my mind. Shit is breaking all the time. Everything is so half assed.

The music player that i use sometimes swaps ui with whatever was below it like for example the desktop background and i need to kill the process and start again to use the program. WTF.

Bluetooth seems to hate me. I check the bluetooth connected devices on my computer, it says headphones connected. BULLSHIT. The headphones are fucking turned OFF. How the fuck can they be connected you dumbass motherfucker computer. So I turn on the headphones. And I cannot connect them since the system thinks that they are already connected. So I have to unpair them and pair them again. WTF. Who fucking invents this bullshit?
Let's say i have headphones connected to the computer. I want to connect them to phone. I click connect from the phone settings. Nothing happens. Bullshit non telling error "could not connect". So I have to unpair from computer to pair to phone. Which takes fucking minutes, because reasons. VERY fucking convenient technology.
The stupid bluetooth headphones have a loud EARRAPE voice when turning them on "POWER ON!!! PAIRING", "CONNECTED", "DISCONNECT". Loudness of this cannot be modified. The 3 navigation buttons are fucking unrecognizable so i always take few seconds to make sure i click the correct button.

Fucking keyboard sometimes forgets that I remapped esc key to caps lock and then both keys don't work so i need to reconnect the keyboard cable. At least it's not fucking bluetooth.

The only reason why hdmi switches exist is because monitor's navigation menus have terrible ui and/or infrared activated, non-mechanical buttons.
Imagine the world where monitors have a button for each of it's inputs. I click hdmi button it switches it's input to hdmi. I click display port button - it switches to display port. But nooo, you have to go through the OSD menu.

My ~ directory has hundred of files that I never put there. Doesn't feel like home, more like a crackhead crib.

My other laptop (also tumbleweed) I click on hibernate option and it shuts down. WTF. Or sometimes I open the lid and screen is black and when i click keyboard nothing happens so i have to hold power button and restart.

We've been having computers for 20 + years and they still are slow, unreliable and barely working.

Is there a cure? I'm starting to think the reason why everything is working so shitty and unreliable, is because the foundations are rotten. The systems that we use are built with c, ridden with cryptic abbreviated code, undefined behavior and security vulnerabilities. The more I've written c programs the more convinced I am, that we should have abandoned it for something better long ago. Why haven't we? And honestly what would be better? Everything fucking sucks. The rust seems to be light in the tunnel but I don't know if this is only hype or is it really better. I'm sure it can't be worse than c or c++. Either we do something with the foundations or we're doomed.

I'm beginning to feel like any kind of specific approximation via neural networks is a myth. That if you can't reduce output to simple categorical values that can be broadly interpreted between two points, that it doesn't work.

I have some questions and they don't seem to be getting answered about the design of the net. How many layers should I use ? How many neurons per layer ? How does this relate to the number of desired quantitive scalar outputs I'm looking to create, even if they are normalized, they can vary GREATLY and will if I'm approximating the out of several mathematical expressions. Based on this and the expected error ranges of these numbers and how many possible major digits could be produced within the domain of the variable inputs being introduced, how many neurons per layer ? What does having more layers do ? In pytorch there don't seem to be a lot of layer types per say, but there are a crap ton of activation functions, and should I just be using these at the tail end or should they actually be inserted between layers so the input of the next layer passes through another series of actiavtion functions ? what does this do to the range of output ?

do I need to be a mathematician to do this ?

remembered successes removed quantifiable scalars entirely from output, meaning that I could interpret successful results from ranges of decimal points.

but i've had no success with actual multi variable regression as of yet, even when those input variables are only 2 and on limited value ranges eg [0,100] and [0, 2pi]

and then there are training epochs to avoid overfitting, and reasonable expectation of batches till quality results will start to form.

Upgrading my tech skills.. Once again I feel my personal my personal dev environment and told are much more up-to-date than what I use at work.... Though the book Kim reading is on TDD and was written 3 years ago.

Maybe I should read another on in cloud services and ML... but don't have any motivation for these topics.

I need TDD for work because now we're emphasizing unit test coverage...

I usually only use manual functional tests to verify the final outputs as either the testing framework is broken (JS) or I don't have time to relearn the frameworks for the particular language...

Anyway got off topic... So questions after:

1. Do you ever feel your technologically always more ahead than what you do at work and essentially you bring skills to the job but you don't learn much out of it?

2. How do you test? I actually got into a bit of a argument/discussion with my colleagues about how to implement unit tests. Apparently there are 2 ways to test? Black box vs WhiteBox. She said she tests only Public methods using mock inputs, dependencies. She read online and seems there is an opinion that should only test public functions and if you can't then your app is designed incorrectly, not separated enough.

For me I test the private functions individually (WhiteBox/Java reflection) because the public one is like generateReport and as a whole is like a Pachinko machine, too many unique paths that would need a test case for.

So thoughts? Yes sorry for turning it into a remake I guess...

Why is the 1password login process so shitty? You need 4 different inputs, and every time you login it forces you to download the same PDF you have saved 10 times. there's no skip option, and it doesn't memorize that you have already freaking downloaded it.
on top of that, my company has restricted the accounts to not reset their password on their own. How the hell this helps a company? Why this option even exist?
I mean I have to DM my manager: Sorry dude, I'm so clumsy that I have accidentally deleted my password, can you initiate a reset password process so I can have access to 1password?

Guys is that the case, that it is such a hassle to work with forms in React-Redux application?! Hell, it takes a lot of time to just create a simple form with like 3 lines of inputs.

Everytime I need to setup bunch of those Actions that will fire on a field value change, than selectors to pick from the state and send to the backend with redux-saga. OMG OMG OMG.

Redux-forms kind a struggle to setup too at first, but I guess I have to go for it anyway?

because bluetooth isn't designed for ultra-high-gaming-inputs. and you don't need higher polling rates unless you're a very competetive gamer. in which case you're using wired, anyway, because latency.

Any Elixir devs out there ?

I started learning Elixir to explore distributed computing.

Need more inputs.